Home

RapidMiner Studio Manual - RapidMiner Documentation

1. B process gt H old Figure 2 20 Representation of the execution order This order is unfavourable however since more data sets have to be handled at the same time der of the operators and to change it Automatic size Changes the size of the white working area in such a manner that all operators currently positioned have just enough space This is 43 2 First steps a Process X 32 a E B Process gt 2 0 Ea Figure 2 21 New order after some changes particularly practical for automatic downsizing size optimisation Update projected meta data If clicked the projected meta data information at the ports will be updated to match the real data after operator execution Additionally the context menu allows to export the process to PDF and other 44 2 3 Design Perspective formats and to print it 2 3 5 Parameters View W lt new process gt RapidMiner Studio 60 000 rapid 010 P W Erm File Edit Process Tools View Help 2 QG0B9 ea SODPUB a s 2B Operators i Process E Parameters P PR e t Apress Ht BIS varg Process Control 37 IE Process 2a 5 amp Repository Access 6 logverbosity init gt Import 26 inp a a Export 17 Joos l a Data Transformation 114 GQ Modeling 118 resuitfile g GQ Evaluation 29 a i random seed 2001 J send mail
2. lt Process XML qe m t g HtH d inp res res Generate Sale Generate Attri Generate Attributes example set output example set output Meta data Data Table Number of examples 100 9 attributes Generated by Generate Attributes example set output Generate Sales Data output Role Name Type Range Missings Comment id transacti integer unkown 0 A store_id nominal Store 0 0 custome nominal Custo 0 product_id integer 10000 0 product_ nominal Books 0 date date Unbound 0 amount integer 1 10 0 single_p integer 10 100 0 total_price real unkown 0 v Press F3 for focus Figure 3 9 The meta data contains the complete path of the object and re mains with the exception of the newly added attribute total price unchanged It is easy to see in the line Generated by that the last thing the object stemmed from now is the operator Generate Attributes and was previously the opera tor Generate Sales Data In addition hardly anything has changed both the number of the examples and the eight original attributes have stayed the same However a ninth attribute has been added Our newly defined attribute 64 3 3 The First Analysis Process total_price can also now be found in the table And our process has still not been executed as you can see just by looking at the status
3. Samples imone DB amp Local Repository Christian Pels F 9 Project 1 Christian Pels E Project 2 christian Pels Results Christian Pels Data Christian Pels E Gj Processes Christian Pels Analysis Christian Pels Reporting Christian Pels Q ETL Christian Pels me Figure 3 2 A repository structured into projects and each of those structured according to data processes and results 55 3 Design of Analysis Processes It is recommended that you create new directories in the repository for individual analysis projects and name these accordingly It will never hurt to structure fur ther within the projects e g structuring into further subdirectories for project specific data different phases of data transformation and analysis or for results A repository could thus have the structure like it is shown in Figure 3 2 3 3 The First Analysis Process After the creation of the process RapidMiner Studio automatically switches to the Design Perspective and you can start with the process design In later chap ters we will talk in detail about how to load data in RapidMiner Studio and store it in your repository In this section however the basic execution of processes is more important to us and we will therefore wait a short while before analysing real data na er gt ie g i CEE PLE m e e Biprocoss a9 HtHS FCBEBFRRB ont J process inp ine 7 g M ogie Ts Mode Ev
4. Jason Whitmore trom EartyBira Ventures Whatis an application wizard ova ica are ane cass dive deseo make he poorer elitair F predcve anaes accessinie to everyone retary business managers and business Tutorials analysts Rapiavine 6 0 pronaes our specie appleation waaras te start ncucng Learn the basics of RapidMiner MEREST A Drectmarkeimg Sentiment anais S aa matnance Hook for alterting size en aoa 3 Log View Nov 22 2013 1 20 17 PM SEVERE Cannot execute Vastng Proce smo gvon for rasultfie using s Figure 2 5 Size changes between views You see the familiar Home View and the new Log View at the bottom If you move the mouse into the highlighted area between them the cursor changes shape and indicates that you can change the sizes of the views by dragging so by holding the mouse button down and moving the cursor Feel free to try it out 24 2 2 Perspectives and Views As already suggested you can also change the position of the views as you like In order to do this simply move the cursor onto the name area of the view and drag the view to another position The position in which the view would be arranged after releasing the mouse button is highlighted by a transparent gray area 19GB SF na ere a m Te IN New Process News Create your own analysis Rapiaminer v6 0 Makes Proactive AnaWics Nore Accessinie t0 AN a The primaryfocus in RapidMiner 8 is the additi
5. except that the set of nominal values is a superset of the empty set A trivial statement but one that is nevertheless correct The meta data cannot be fully determined in all cases as early as at the time of development This is generally the case whenever the meta data is dependent on the actual data as it is here In this case RapidMiner Studio tries to obtain as much information as possible about the data 110 i a Global leader in predictive analytics software D rapidminer Boston London Dortmund Budapest www rapidminer com
6. to predict the result of such a coin toss The physical formulas for this are all known in any case We shall now look at another scenario only this time we can predict the outcome of the situation A glass will break if it falls from a certain height onto a certain type of ground We even know in the fractions of the second when the glass is falling There will be broken glass How are we able to achieve this rather amazing feat We have never seen the glass which is falling in this instant break before and the physical formulas that describe the breakage of glass are a complete mystery for most of us at least Of course the glass may stay intact by chance in individual cases but this is not likely For what it s worth the glass not breaking would be just as non coincidental since this result also follows physical laws For example the energy of the impact is transferred to the ground better in this case So how do we humans know what exactly will happen next in some cases and in other cases for example that of the toss of a coin what will not The most frequent explanation used by laymen in this case is the description of the one scenario as coincidental and the other as non coincidental We shall not go into the interesting yet nonetheless rather philosophical discussions on this topic but we are putting forward the following thesis The vast majority of processes in our perceptible environment are not a result of coi
7. which can be left again using the arrow pointing upwards or via the navigation bar Defining the Execution Order In nearly all cases RapidMiner succeeds in automatically determining the cor rect execution order of the operators In order to do this RapidMiner uses the connection information and the fact that an operator the result of which is to be used by another operator must obviously be executed before the latter However there are cases where the order cannot be automatically defined such as completely parallel subprocesses or where the automatic order is not correct for example because a macro must first be computed before it can be used as a parameter in a later operator But there are also other reasons that often play a big part such as more efficient data handling or an exact order desired for execution for reporting for example 41 2 First steps For this purpose RapidMiner offers an elegant method for indicating the order of the operators and even for editing the execution order comfortably Please click on icon with the double arrow pointing upwards and downwards with the question mark in the toolbar of the Process View Fig 2 18 and the process view shows the order definition of the operators Instead of the icon for each operator the number of its execution will now be shown The transparent orange line connects the operators in this order as shown in Figure 2 20 To change such an execution order you
8. 0 0 e e a ee ee ee 54 3 3 The First Analysis Process oaao 56 3 3 1 Transforming Meta Data aaas 58 3 4 Executing Processes nna Tace do do a aa E eese anra iutL t 68 3 4 1 Looking at Results aaae 69 3 4 2 Breakpoints aoaaa 00000 ee eee eee 70 Data and Result Visualization 75 4 1 Result Visualization ws sa uasa n a e a a a i aa a aA 75 4 1 1 Sources for Displaying Results 76 4 2 About Data Copies and Views oaoa a 79 4 3 Display Formats o sisi l t r a Eea a oan a e e E a E aT 80 4al Description asos dine a anssi ia ee A E a a es 81 AS2 Vables ceo 20 6 a MER PEPE ELS 81 A323 Charts Sc Geb amp h auan aa be Pots ab eed dod a Boe 86 ASA Graphs sraid Menke De Baw oe Be ei yd eS ot ee 89 4 3 5 Special Views ia ane m a 6 0 Bee eRe aS 92 4 4 Result Overview 0 0202 ee ee ee 92 Repository 95 5 1 The RapidMiner Studio Repository 95 5 1 1 Creating a New Repository 0 97 5 2 Using the Repository 2 0 0 0 2 00000 99 5 2 1 Processes and Relative Repository Descriptions 99 5 2 2 Importing Data and Objects into the Repository 100 5 2 3 Access to and Administration of the Repository 103 5 2 4 The Process Context 022005 104 5 3 Dataand Meta Data 02 2 0 Ae 106 5 3 1 Propagating Meta Data from the Repository and through the ProCess i gt ss ihe bo AS Go ee ee en a es 108 1 Motivation and Fundamental T
9. B ExampleSet Set Role ExampleSet 100 examples 1 special attribute 3 regular attributes Filter 100 100 examples all s ri RowNo transaction store_id product_cat total_price Data 4 4 Store01 Toys 270 739 A 2 2 Store 15 Movies 121 173 3 3 Store 12 Movies 483 063 Statistics A 4 Store 05 Books 84 813 ae 5 5 Store 01 Clothing 326 077 Charts 6 6 Store 11 Sports 169 424 T 7 Store 10 Health 80 619 l 8 8 Store 10 Health 469 522 Advanced 9 9 Store 11 Health 141 737 Charts 10 10 Store 14 Toys 341 352 j EJ 11 11 Store 14 Health 409 243 Annotation 12 12 Store 12 Sports 234 470 13 13 Store 08 Electronics 713 422 14 14 Store 06 Home Garde 284 513 15 15 Store 01 Home Garde 324 730 Figure 4 6 Colour codings and alternating line backgrounds make navigation eas ier within tables This colour coding is also transferred to the meta data attributes with special roles also have a consistently light yellow background and regular attributes have an alternating light blue and white background However this colour scheme can be completely different for other objects like in Fig 4 7 In the case of a correlation matrix for example individual cells can also be coloured the darker they are the stronger is the correlation between these attributes 82 4 3 Display Formats File Edt Process Tools Vi x uB nw SPNB 3370 amp Result Overview B ExamploSet Genera
10. Health 27 04 2007 Figure 4 9 Actions like selecting lines or columns sorting contents by columns or adjusting column widths are available in a context menu Copying Table Contents Just as with the text view above you can also highlight individual cells within tables using the mouse or highlight the complete table by clicking in the table and using Ctrl A Actions are also available in the context menu for highlighting whole lines or columns You can then copy the selected area onto the clipboard by means of Ctrl C and paste it into other applications Please note that the table structure stays as it is if for example you paste into applications such as Microsoft Excel which support tabular data 4 3 3 Charts One of the strongest features of RapidMiner Studio are the numerous visualisation methods for data other tables models and results offered in the Charts View and Advanced Charts View 86 4 3 Display Formats Configuring Charts The structure of all charts in RapidMiner Studio is basically the same There is a configuration area on the left hand side which consists of several familiar elements amp Result Overview E ExampleSet Generate Sales Data 4 re as amount 1 000 9 000 ri i gt Scatter Store 15 oo Data this gt date Store 14 o eo Statistics C Log Scale Store 13 e y Axis 5 Store 12 e 0 store_id Charts T Log Scale Store 11 eo eo tl Color Colum
11. Transformation are usually only executed on regular attributes so on those with out a special role However the operators offer an option called include special attributes for this meaning that the changes are also applied to those with a special role 3 4 Executing Processes Now we are ready and want to execute the process just created for the first time The status indicators of all operators should now be yellow and there should be no entries in the Problems View In such a case it should be possible to execute our process consisting of the three operators for generating data computing the total turnover for each transaction and filtering attributes without any problems You have the following options for starting the process 1 Press the large play button in the toolbar of RapidMiner 2 Select the menu entry Process Run 3 Press F11 gt iB Figure 3 13 The play button starts the process you can stop the process in be tween with the pause button and stop aborts the process completely While a process is running the status indicator of the operator being executed in each case transforms into a small green play icon This way you can see what point the process is currently at After an operator has been successfully executed the status indicator then changes and stays green until for example you change a parameter for this operator Then the status indicator will be yellow The same applies for
12. a shadow like it is shown in Figure 2 16 In this way you can set the course for a correct connection early during the insertion Tip It is recommended that you activate the option for automatic connection for the input ports at least Even if the connection algorithm based on the meta data occasionally creates a wrong connection you still save yourself a lot of work for all cases where the correct connection is automatically recognised Connections 2 Manually You can also interconnect the operators manually and this is even necessary for more complex processes In order to do this click on an output port You will now draw an orange strand like it is shown in Figure 2 17 Click on an input port in order to connect the selected output port with this input port In order to cancel the process hold the mouse still and click using the right hand mouse button The orange strand will disappear and you can continue working as normal Connections 3 Fully automatically Sometimes numerous operators are already in a sub process and are not yet connected In this case the options Auto Wire and Re Wire can serve you 36 2 3 Design Perspective lt Process B GiPerocess gt rHt od Read Excel 2 Leg cel 2 fil out L e Figure 2 16 The second operator is highlighted during the dragging process frame plus shade and is preferably connected with the new op erator if the latter is now dropped and
13. additional tab in the large area on the left hand side 1 Automatic Opening We have already seen that the final results of a process i e objects which are supplied to the result ports on the right hand side in the process are displayed automatically The same also applies for the results at connected ports in the case of a breakpoint You can simply collect at the result ports all process results that you wish to see at the end of an analysis process and they are all shown together in the tabs of the Results Perspective 2 Results from Repositories The second option for displaying results is loading results from one of your repos itories You can do this via the context menu of a repository entry or simply by double clicking on an entry Of course this process is not only recommended for 77 4 Data and Result Visualization reviewing results but also for comparing with earlier results 3 Results from Ports A third possibility for looking at results and even intermediate results is display ing results which are still at ports RapidMiner tries to store the results which were supplied by individual operators at the relevant ports for a while longer If there are still results at a port these can be selected and looked at via the context menu of the port amp Show ExampleSet Result E Store ExampleSet in Repository Z Disconnect Port Figure 4 3 Display of results which are still at ports You may kno
14. all operators that follow This means you can see very quickly on which operators a change could have an effect 68 3 4 Executing Processes The process defined above only has a short runtime and so you will hardly have the opportunity to pause the running process In principle however you can briefly stop a running process with the pause symbol e g in order to see an intermediate result The operator currently being executed is then finished and the process is then stopped You can recognise a process that is still running but currently paused by the fact that the colour of the play icon changes from blue to green Press the play button again to continue executing the process further If you do not wish to merely pause the process but to abort it completely then you can press the stop button Just like when pausing the operator currently being executed is finished and the process fully aborted immediately after Please note that you can switch to the Design Perspective immediately after aborting the process and make changes to processes even if the execution of the current operator is being finished in the background You can even start further processes and do not need to wait for the first process to be completed Note It was explained above that the operator being executed is always com pleted if you abort This is necessary to ensure a sound execution of operators However completing an operator may need much more time in individu
15. and admittedly very simple process which generates some data and displays the result in the Result Perspective configuration was very simple in this specific case It was not at all necessary to set any operator parameter A red status indicator and entries in the Problems View would have indicated such a configuration need 3 3 1 Transforming Meta Data We will now deal with one of the most fascinating aspects of RapidMiner Studio namely the ability to compute the output of an operator or process beforehand and to even do this during the design time so without having to load the actual data or even perform the process This is made possible by the so called meta data transformation of RapidMiner Studio Of course each operator defines the way in which the received input data is transformed This is its task at the end of the day The special thing about RapidMiner Studio however is that this can not only be done for actual data but also for the meta data about this data This is typically much less voluminous 58 3 3 The First Analysis Process than the data itself and gives an excellent idea of which characteristics a particular data set has The meta data in RapidMiner Studio essentially equates to the concept descriptions we discussed previously It contains the attribute names of the example set as well as the value types and the roles of the attributes and even some fundamental statistics So much for the theory but what doe
16. can click anywhere on an operator to select it The path leading to this operator can now not be changed but clicking again on another operator will attempt to change the order in such a way that the second operator is executed as soon as possible after the first While you move the mouse over the remaining operators you will see the current choice in orange up to this operator and in grey starting from this operator A choice that is not possible is symbolised by a red number You can cancel a current selection by right clicking In this way you can as shown in Fig 2 21 change the order of the process described above to the following with only a few clicks 2 3 4 Further Options of the Process View After having discussed nearly all options of this central element of the RapidMiner Design Perspective we will now describe the remaining actions in the toolbar which can be seen in Figure 2 18 as well as further possibilities of the Process View The five icons on the right hand side of the Process View toolbar perform the following actions Auto wire and Re wire connections The plug symbol allows to auto wire and re wire the connections between operators Automatic arrangement Rearranges all operators of the current process accord ing to the connections and the current execution order Show and alter execution order This action allows you to see the execution or 42 2 3 Design Perspective a2 Process 52 lt 3 Bi
17. expects an example set well which are hidden behind the plug symbol in the Process View This works particularly well if a relatively sequential approach was taken when the process was created and the operators were properly lined up one behind the other i e the previous operator was always marked by a frame and shadow during insertion It is always wise however to perform a manual examination following the fully automatic connection since unintended connections can occur especially in the case of more complex processes 37 2 First steps lt Process m B FProcess gt ril ff Dd Figure 2 17 Click on an output port in order to connect right click to cancel Selecting Operators On order to edit parameters you must select an individual operator You will recognise the operator currently selected by its orange frame as well as its shadow If you wish to perform an action for several operators at the same time for example moving or deleting please select the relevant operators by dragging a frame around these In order to add individual operators to the current selection or exclude individual operators from the current selection please hold the CTRL key down while you click on the relevant operators or add further operators by dragging a frame 38 2 3 Design Perspective Moving Operators Select one or more operators as described above Now move the cursor onto one of the selected operators and drag
18. form of a so called tooltip Unlike in other programs this help information is much more powerful than normal By pressing key F3 you can turn such a tooltip into a proper dialog which you can move around and change in size as you wish In addition these RapidMiner Studio tooltips are also able to include elements other than textual information with the meta data such as tables for example Please note that the meta information does not necessarily have to be available immediately You may have to first initiate the loading of the meta data by click ing once on a link within the tooltip Doing this means that should the tooltips of the repository entries be inadvertently looked at the possibly quite large meta 106 5 3 Data and Meta Data data is prevented from having to be loaded immediately causing RapidMiner Studio to slow down Tip Hold the cursor over a repository entry for a short time in order to look at the meta data or load it first If the entry is an intermediate result for example you can easily recognise what pre processing has already taken place The following picture shows what the meta data for the golf data set from the ex ample directory in the Sample repository provided with RapidMiner Studio looks like Fig 5 6 First you will see that the data set contains 14 examples Num ber of examples and 5 attributes Number of attributes The attribute with the name Outlook is nominal and takes
19. in Rapid Miner Studio also for the operator descriptions for example which are shown as 59 3 Design of Analysis Processes lt Process XML B Process gt g i tv EEA Generate Sale 4a Generate Sales Data output output Meta data Data Table Number of examples 100 8 attributes Generated by Generate Sales Data output Role Name Type Range Missings Comment id transactio integer unkown 0 store_id nominal Store 0 0 customer nominal Custom 0 product_id integer 10000 0 product_c nominal Books 0 date date Unbound 0 amount integer 1 10 0 single_pr integer 10 100 0 Press F3 for focus Figure 3 5 The meta data of the output port of the operator Generate Sales Data a tooltip in the Operators View You can take time to read the tooltip and also adjust it in terms of size if you press the key F3 beforehand Please note that the meta data can often only represent an estimation and that an exact indication is not always possible This is explained by the fact that parts of the meta data are unknown or can only be indicated inaccurately e g with 60 3 3 The First Analysis Process the indication lt 100 Examples for the number of the examples Nevertheless the meta data is a valuable source of help both for forthcoming design decisions and for the automatic recognition of problems as well as the suggestions for their solu
20. never 7 encoding SYSTEM Parameters View LB Repositories la s asaoe je 9E a H Hep Comment E amp Local Repository Cristian Pe E amp Main Repository Christian Pe Synopsis Dy Problems Blo The root operator which is the outer most operator of every SB No problems found process Message Fixes Location Description Each process must contain li exactly one operator of this class and it must be the root j operator of the process This operator provides a set of Figure 2 22 Parameters of the currently selected operator are set in the param eter view Figure 2 22 shows the Parameters View of RapidMiner Numerous operators require one or several parameters to be indicated for a correct functionality For example operators that read data from files require the file path to be indicated Much more frequently however parameters are not absolutely necessary although the execution of the operator can be controlled by indicating certain parameter values and in the case of modelling also frequently be optimised After an operator offering parameters has been selected in the Process View its parameters are shown in the Parameters View Like the other views this view 45 2 First steps also has its own toolbar which is described in the following Under the toolbar you will find the icon and name of the operator currently selected followed by the actual paramete
21. repository However RapidMiner Studio can also import numerous other formats within processes You will find operators for this in the group Import However caution is required when using these operators The availability of meta data is not guaranteed for these operators which can lead for example to processes that assume the existence of certain attribute values only noticing any errors in the runtime of the process Nevertheless using these file formats is sometimes not avoidable e g for the regular execution of ETL processes The goal of these processes should be however to transfer the data into the repository with a subsequent store operator so that it can be used by the actual analysis processes that follow The operators of the Import group have numerous parameters tailored to the respective format Please see the respective operator documentation for their description 102 5 2 Using the Repository Storing Objects from the Result or Process View After you have executed a process the Results Perspective with the tab of the same name is presented to you in the basic setting On the right hand side of its toolbar there is a button with which you can store the result currently selected in the repository A dialog will also appear here allowing you to select a folder and a name If your process contains intermediate results which are not or no longer indicated in the Results Perspective you can also store the
22. simply drag the highlighted area in this overview to the desired place and the Process View will adjust automatically 49 2 First steps 2 3 8 Problems and Log View Problems View A further very central element and valuable source of help during the design of your analysis processes is the Problems View Any warnings and error messages are clearly indicated in a table here Fig 2 26 AS Problems Log 2 Errors Message Fixes Location 3 Gaussian Process cannot handle polynominal attributes Convert attributes to numerical Q Gaussian Process training set Gaussian Process cannot handle polynominal label Choose among 4 available quick fixes Gaussian Process training set Figure 2 26 Representation of all current problems In the first column with the name Message you will find a short summary of the problem In this case the data mining method Gaussian Process is not able to handle polynomial multivalued categorical attributes The last column named location shows you the place where the problem arises in the form of the operator name and the name of the input port concerned The icon on the left side of the Problems View toolbar activates a filter that displays only the problems of the currently selected operator This is practical for larger process with several error sources There is also the possibility of suggested solutions for such problems and of im plementing them directly These soluti
23. the result output of the total process The screen ought to then look roughly like in Fig 3 6 Tip Instead of dragging an operator into the Process View and reconnecting the 61 3 Design of Analysis Processes lt Process XML e Y Lp v t H 2 EX E Figure 3 6 The data is generated first and then a new attribute is produced ports you can also drag the operator onto an existing connection If you move the cursor position exactly onto the connection the latter will be highlighted and the new operator will be inserted directly into the connection Even if this process would work now which is visible from the yellow status indicators and the empty Problems View then the second operator would not compute anything without a further configuration and the final result would only be the same as that after the first operator We therefore choose the new operator Generate Attributes and select it in this way The display in the parameter view changes accordingly and the parameters of this operator are shown The substantial parameter has the name function descriptions and is configured on the associated button with one click as can be seen in Fig 3 7 After you have pressed the button with the name Edit List 0 a dialog will open giving you the opportunity to enter the desired computation in Fig 3 8 You can add further entries in such lists of individual parameters with the two actions A
24. this rule of thumb and would soon reach a decision as to the most probable future of the falling article Analogy reasoning and the creation of rules are two first examples of how humans and also data mining methods are able to anticipate the outcome of new and unknown situations Our description of what goes on in our heads and also in most data mining methods on the computer reveals yet another interesting insight The analogy reasoning described does at no time require the knowledge of any physical formula to say why the glass will now break The same applies for the rule of thumb described above So even without knowing the complete physical description of a process we and the data mining method are equally able to generate an 1 2 Fundamental Terms estimation of situations or even predictions Not only was the causal relationship itself not described here but even the data acquisition was merely superficial and rough and only a few factors such as the material of the falling article glass and the falling height approx 2m were indicated and relatively inaccurately at that Causal chains therefore exist whether we know them or not In the latter case we are often inclined to refer to them as coincidental And it is equally amazing that describing the further course is possible even for an unknown causal chain and even in situations where the past facts are incomplete and only described inaccurately This section has give
25. well as operators in order to apply the generated models to new data sets Evaluation Operators which can compute the quality of a model and thus for new data e g cross validations bootstrapping etc You can select operators within the Operators View and add them in the desired place in the process by drag and drop You connect the operators by drawing a line between the output and input ports of the operators You have the choice whether you want the operators to be connected automatically when inserted Select the plug symbol on the left hand side of the toolbar of the view in Figure 2 11 and define whether outgoing and or incoming connections are to be created automatically amp Operators Repositories P gt E Figure 2 11 Actions and filters for the Operators View In order to make the work as easy for you as possible the Operators View also supports filter besides which can be used to search for parts of the operator name or the complete operator name Just enter the search word into the filter field As soon as there are less than 10 search hits altogether the tree is opened up to reveal all search hits This means you do not need to navigate through the complete hierarchy each time Clicking on the red cross next to the search field erases what is currently entered and closes up the tree again The icons right beside the search field can filter out deprecated operators and sort the operators according to t
26. would probably produce the best results and for which customers In the following chapters we will focus on further uses of data mining and at the same time practise transferring concepts such as customers business processes or products into attributes examples and data sets This will train the eye to detect further possibilities of application tremendously and will make analyst life much easier for you later on First though we would like to spend a little time on RapidMiner and give a small introduction to its use so that you can implement the following examples immediately 17 2 First steps RapidMiner Studio combines technology and applicability to serve a user friendly integration of the latest as well as established data mining techniques Defining analysis processes with RapidMiner Studio is done by drag and drop of operators setting parameters and combining operators As we will see in the following processes can be produced from a large number of almost randomly nestable operators and finally be represented by a so called process graph flow design The process structure is described internally by XML and developed by means of a graphical user interface In the background RapidMiner Studio constantly checks the process currently being developed for syntax conformity and automatically makes suggestions in case of problems This is made possible by the so called meta data transformation which transforms the underlying meta da
27. Books s4813 Pona a s 5 Store01 Clothing 326 077 B Labor epcttatons rone Charts 6 e Sloe 11 Soos 169 424 iJ Market Data none e B Polynomial vore 7 7 Sorem Heam s0519 preka z Sem ees ENED pail Advanced 9 s Store11 Heath 141 737 iB anani Charts J weionng irons 10 1 Store 14 Tos 341352 4 Ss n n Store 14 Heath 409 243 04_Learner rers SS G 02 Preprocessing rons annotation we SRE CATS E 02 Validation 2 2 Store 02 Elecronios 719 422 a u 4 Store 00 Home Garde 284 513 1 35 Store01 Homercare 324 730 Mabie aed se 1 Sore00 Tos ua 8 Goa omer pars Boe 7 v tore 06 Electronics 203 100 a EA 18 18 Sweo0 Movies 173 487 E tain Repostory 13 1 Soera sooks 321 105 2 EJ Store09 Books 418 413 a a Store 14 Woes 400 004 ey ns Gj Processes 2 2 Seos Tos 348 053 ance a Soreog Elecronies 97582 amp Reporting a 24 Storao2 Toys 262 566 GET icnrsian Fe 25 25 Storeo3 Heath 72982 20 2 Sweta Books 143 575 z a stoe10 Cloming 112195 2 2 Store05 ovos 680 004 2 2 Storeo3 HomelGarde 47 574 20 a Store 14 Home Garae 285 179 a a stoe10 spons stana 2 2 Storeco Heath 298 596 33 3 Storeo5S Home Garde 90 224 Figure 3 14 After a process has been successfully executed you can look at the results in the Result Perspective chapter we will talk in detail about the possibilities of the Result Perspective If you wish to return to the Design Perspective then you can do this at any time using the switching methods you are familiar with Tip After
28. HA S Ber R a Gi Process games wa j E impor 26 iw T e Enan loonie E T Data Transformation 114 E Modeling 118 resulte i E Evaluanon 29 ones mane 3 on svete J E Repositories Help amp B3 8Sae Comment View 2 OSS IH Local Repository chrsien Pe Wain Repository cistio P The root operator which is the outer most operator or every process Description Each process must contain exactly one operator of this class and it must be the root operator of the process This operator provides a set of parameters that are of global relevance to tne process like logging and initialization of parameters of the random number generator Figure 2 24 Help texts are shown both for currently selected operators in the Operators View and for currently selected operators in the Process View shown description includes a short synopsis which summarises the function of the oper ator in one or a few sentences a detailed description of the functionality of the operator and a list of all parameters including a short description of the param eter the default value if available the indication as to whether this parameter is an expert parameter as well as an indication of parameter dependencies Comment View Unlike the Help View the Comment View is not dedicated to pre defined de scriptions but rather to your own comments on individual steps of the process Simp
29. RapidMiner Studio Manual J rapidminer 2014 by RapidMiner All rights reserved No part of this publication may be reproduced stored in a retrieval system or transmitted in any form or by means electronic mechanical photocopying or otherwise without prior written permission of RapidMiner Contents 1 Fundamental Terms 1 1 Coincidence or not 2 2 2 0 2 ee 1 2 Fundamental Terms 2 2 200002 ae 1 2 1 Attributes and Target Attributes 0 1 2 2 Concepts and Examples 1 2 3 Attribute Roles 2 0202000 L24 Values Typis en 4tk EE he he a Roe ole ee ee 1 2 5 Data and Meta Data aoaaa aa 1 2 6 Modelling sscan 2 First steps 2 1 Installation and First Repository soo 2 2 Perspectives and Views oaoa aaa e 2 3 Design Perspectives moeces a eaa iaee e yy e he A A 2 3 1 Operators and Repositories View 29 2 Process View a minniy 4 wi ee Bet E E ee A 2 3 3 Operators and Processes ooa a 2 3 4 Further Options of the Process View 2 3 5 Parameters View nc ee ck Sk ee ae e 2 3 6 Help and Comment View 2 3 7 Overview View ou lies dah es 2 Ge Rea ew e 2 3 8 Problems and Log View 4 3 Design of Analysis Processes 3 1 Creating a New Process 10 11 14 15 19 20 21 27 28 31 31 42 45 47 49 50 53 53 Contents VI 3 2 Repository Actions 0
30. accurately as possible such as material properties or the nature of the ground is important And one can even try to find an answer to the question as to whether this book will help you by recording the characteristics of yourself the reader and aligning them with the results of a survey of some of the past readers These measured reader characteristics could be for example the educational background of the person concerned the liking of statistics preferences with other possibly similar books and further features which we could also measure as part of our survey If we now knew such characteristics of 100 readers and had the indication as to whether you like the book or not in addition then the further process would be almost trivial We would also ask you the questions from our survey and measure the same features in this way and then for example using analogy reasoning as described above generate a reliable prediction of your personal taste Customers who bought this book also bought This probably rings a bell 1 2 1 Attributes and Target Attributes Whether coins or other falling articles or even humans there is as previously mentioned the question in all scenarios as to the characteristics or features of the respective situation We will always speak of attributes in the following when we mean such describing factors of a scenario This is also the term that is always used in the RapidMiner software when such describing featur
31. ae 10 10 Store14 Toys 341 352 amp 11 1 Store 14 Health 409 243 Anneaation 12 12 Store 12 Sports 234 470 13 13 Store 08 Electronics 713 422 14 14 Store 06 Home Garde 284 513 15 15 Store 01 Home Garde 324 730 Figure 4 4 The views Data currently shown Statistics Charts and Ad vanced Charts exist for a data set For data sets for example there are three views i e the display of the data itself Data View meta data and statistics Statistics View the display of different visualisations Charts View and the advanced display of different visualisations In the example above you can see the data view of a data set in 80 4 3 Display Formats the form of a table Besides such tables further standard display formats are available which we would like to explain in the following 4 3 1 Description The most fundamental form of visualisation is that in text form Some models as well as numerous other results can be displayed in textual form This is typically done within the so called Description View which you can select if there are several views for this object using the buttons on the left side of the tab In RapidMiner Studio you can always highlight such texts with the mouse and copy onto the clipboard with Ctrl C The results are then available in other applications also You can also highlight longer texts completely by clicking on the text area followed by Ctrl A and then
32. ajority of work during a data analysis is time spent transferring the data into a format suitable for data mining These transformations are therefore dealt with in detail in later chapters 1 2 6 Modelling Once we have the data regarding our customers available in a well structured format we can then finally replace the unknown values of our target attribute with the prediction of the most probable value by means of a data mining method We have numerous methods available here many of which just like the analogy reasoning described at the beginning or the generating of rules of thumb are based on human behaviour We call the use of a data mining method model and 15 1 Fundamental Terms Number Number Prototype Number f f 0 ur of pur positively Name Address Sector Subsector of em i p J chases chases received ployees group 1 group 2 Pi 76 Any Street T Doe Systems _ bending yes Sunnyville Mechanics gt 1000 5 0 Inc ma Massachusetts chines 4456 Parkway Tele 600 John Paper Blvd Salt Lake IT communi 1000 3 7 City Utah cations 5500 Park Williams amp Street Hart no a reb Sar Trade Textiles lt 100 1 ll Sons ford Connecti cut Table 1 1 An example scenario 16 1 2 Fundamental Terms the result of such a method i e the prediction instruction is a model Just as data mining can be used for different issues this also appl
33. al cases and also require other resources such as storage space So when you are aborting very complex operators you can see this taking hours and requiring additional resources then your only option is to restart the application 3 4 1 Looking at Results After the process was terminated RapidMiner Studio should have switched to the Result Perspective Fig 3 14 If this was not the case then you probably did not connect the output port of the last operator with one of the result ports of the process on the right hand side Check this and also check for other possible errors taking the notes in the Problems View into consideration Feel free to spend a little time with the results Since the process above has not yet performed any modelling but only transformed data the result only consists of an example set You can look at the meta data of this data set and try out the table plus some of the visualisations in Charts or Advanced Charts In the next 69 3 Design of Analysis Processes lp i9GB9 na erie 3 93B F Resultovenisw 8 Exemplasat Select Attibutas B Repositorios ExamploSot 100 examples 1 special attibuto 3 regular atributes Fiter 1001100 examples at a aga RowNo transact store 1d product cat toal_pice B O Samples oon Data 1 1 Stoet Toys 270739 O FR anoe J Deals gt 2 2 Store 15 Movies 121 173 PR x 3 Store12 Movias 483 063 BOM one Statistics E Gol Tasteat nor 4 4 Store05
34. aluation 28 A 4nigaen epert parameters Oran Synopsis The root operator which is the outer most operator of every process Description Each process must contain exactly one operator of this class and it must be tne root operator of the process This operator provides a set of parameters e Figure 3 3 The preset Design Perspective immediately after the creation of a new process 56 3 3 The First Analysis Process As long as you have not changed the selection and positions of the individual views for the Design Perspective your screen should more or less look like the one in Figure 3 3 We will now begin our new process starting with the generating of data which we can work on As already said We will see in later chapters how we can use data from the repository or even import it directly from other data sources such as databases or files by using operators But for the moment we will put this aside and generate a small synthetic data set Please expand the group Utility in the Operators View and then the group Data Generation The numbers in brackets next to the individual groups indi cate the number of operators for this group You should now see several operators that can be used for generating an artificial data set This includes the operator Generate Sales Data Now drag this operator onto the white area while hold ing down the mouse button and release it there The oper
35. ator will be inserted and also directly connected depending on the automatic connection setting in the Operators View If this does not happen you can manually connect the output port of the new operator with the first result port of the entire process on the right hand side of the white work area Alternatively it would of course have also been possible to insert the operator by using the New Operator dialog as described in the previous chapter Either way the result ought to look roughly like Figure 3 4 As you have surely noticed the full name of this operator Generate Sales Data is too long and is cut off after the first few letters Move the mouse onto this operator and stay there for a few moments The name will now be shown in full in a small animation Of course you could also rename the operator and give it a shorter name As you can see the status indicator of the operator at the bottom left hand side is yellow This means that the operator has not produced any errors but it has also not yet been successfully executed So you have only fully configured the operator thus far but this does in no way mean it has been directly executed You can easily see that from the fact that the status indicator then turns green Had you not noticed that you have already configured the operator Indeed the 57 3 Design of Analysis Processes Process XML 4 B FProcess gt 2 Er oH od Figure 3 4 An initial
36. ators Multiply with Materialize Data can be a way out You as an analyst can explicitly define your wish to have a copy of the data by first multiplying the reference to the data set by means of Multiply and then explicitly re creating both virtual data sets as tables by means of Materialize Data 79 4 Data and Result Visualization No analyst will seriously do this much work just to be able to access the results via the ports But such interconnections can arise from time to time even in parallel strands of processes and then be resolved depending on the size of the data set by means of views or even explicit copies 4 3 Display Formats However the results got into the Result Perspective each result is displayed within its own file card And in addition there are other different ways of displaying a large number of results which are also referred to as views within RapidMiner Studio g Result Overview 8 ExampleSet Set Role ExampleSet 100 examples 1 special attribute 3 regular attributes Filter 100 100 examples all hd J g RowNo transaction store_id product_cat total_price Data 1 1 Store01 Toys 270 739 ra gt 2 2 Store 15 Movies 121 173 3 3 Store 12 Movies 483 063 aiaia 4 4 Store 05 Books 84 813 5 5 5 Store 01 Clothing 326 077 Charts 6 6 Store 11 Sports 169 424 uf 7 Store 10 Health 80 619 al 8 8 Store 10 Health 469 522 Advanced 9 9 Store 11 Health 141 737
37. clude them from the selection Nodes that are currently selected can be moved while holding down the mouse button You will find further notes on handling graphs in these two modes in the help dialog which is shown if you click on the button Help in the configuration area of the graph Further Settings You can set whether the captions for nodes and edges are to be shown or not The most important setting not necessarily for trees but for other graphs is the choice of a suitable layout which can be made in the selection box directly underneath the mode box The different algorithms have different strengths and weaknesses and you usually have to try and see which display gives the best result for the graph at hand 91 4 Data and Result Visualization 4 3 5 Special Views Alongside the views description table chart and graph described above there are also occasionally further display components which are rarer however and which should be self explanatory For Frequent Itemsets for example there is another special kind of table or graph for the related association rules 4 4 Result Overview We already mentioned the Result Overview Figure 4 14 at the beginning which can always be found as a kind of placeholder in the place where the remaining results are also indicated F4 Result Overview E ExampleSet Retrieve Deals New Process 1 results Process results Completed Nov 22 2013 1 20 38 PM execution ti
38. copy Result Overview Kernel Model SVM Kernel Model Description 4 Total number of Support Vectors 114 te Bias offset 0 058 Weight Table Feature weight calculation only possible for two class learning problems Please use the operator SVMWeighting instead 20w number of classes 3 Figure 4 5 Some models such as models are displayed in textual form Numerous other objects also offer a display in form of a readable text 4 3 2 Tables One of the most frequent display formats of information within RapidMiner is tabular form This is hardly surprising for a software solution with the primary goal of analysing data in tabular structures However tables are not only used for displaying data sets but also for displaying meta data weightings of influence factors matrices like the correlations between all attributes and for many more 81 4 Data and Result Visualization things These views frequently have the term Table in their name especially if confusions are to be feared Otherwise such tables are simply referred to with terms like Data View or Statistics View Colour Schemes Nearly all tables in RapidMiner use certain colour codings which enhance the overview With data sets for example the lines are shown alternately in different colours Attributes with a special role are given a light yellow background here and regular attributes a light blue one z Result Overview
39. d E E gd Group By Column by TE Statistics Stack Column product_category Legend Column Charts one ee Ny None a Value Column Advanced FeO Charts amount C Absolute Values a Aggregation P ee C Use Only Distinct Rotate Labels vertical Figure 4 11 Change of the chart configuration depending on the chart type 4 3 4 Graphs Graphs are a further display format which are found relatively frequently in RapidMiner Graphs basically mean all visualisations which show nodes and their relationships These can be nodes within a hierarchical clustering or the nodes of a decision tree as in Figure 4 13 Graphs like that of the decision tree are mostly referred to as a Graph View and are available under this name Zooming Using the mouse wheel if there is one you can zoom in on and out from the graphs Alternatively you also have two buttons on the top left hand side of the configuration area with which you can increase and reduce the zoom level of your graph 89 4 Data and Result Visualization TE Result Overview E ExampleSet Retrieve Iris 2 E Exampleset Discretize lahel tis setosa_ Iris versicolor risvirginica a Be Data Point Color gt label X Statisties Mer d Matrix Charts Umati 7 4 ae Advanced itendscope ____ ini Net width Net Height 5 40 eo Annotation Training Rounds 25 Adaptation Radius15 Bai O Calculate Figure 4 12 Complex
40. d these supports will therefore not be offered Log View During the design and in particular during the execution of processes numerous messages are written at the same time and can provide information particularly in the event of an error as to how the error can be eliminated by a changed process design amp Problems Log PAR a WOV 22 ZUTS 12 47 SO r MNO Saving resus PA Nov 22 2013 12 47 48 PM INFO Process Samples processes 01_Learner 05_NaiveBayes finished successfully after 0 s Nov 22 2013 12 47 49 PM INFO No filename given for result file using stdout for logging results Nov 22 2013 12 47 49 PM INFO Loading initial data Nov 22 2013 12 47 49 PM INFO Process Samples processes 01_Learner 05_NaiveBayes starts Nov 22 2013 12 47 49 PM WARNING Join Special attribute id already exist skipping Nov 22 2013 12 47 49 PM WARNING Join Special attribute label already exist skipping Nov 22 2013 12 47 49 PM INFO Saving results Nov 22 2013 12 47 49 PM INFO Process Samples processes 01_Learner 05_NaiveBayes finished successfully after 0 s Nov 22 2013 12 48 30 PM INFO No filename given for result file using stdout for logging results Nov 22 2013 12 48 30 PM INFO Loading initial data Nov 22 2013 12 48 30 PM INFO Process Samples processes 01_Learner 05_NaiveBayes starts Nov 22 2013 12 48 30 PM SEVERE Process failed No data was delivered at port Join right disconnected v Figure 2 27 Fu
41. dd Entry and Remove Entry and also delete selected entries The 62 3 3 The First Analysis Process Parameters Seer B Ha Generate Attributes function descriptions Edit List 0 2 hidden expert parameters Figure 3 7 The parameters of the operator Generate Attributes W Edit Parameter List function descriptions x Edit Parameter List function descriptions List of functions to generate attribute name function expressions total_price amount single_price Figure 3 8 Computation of the new attribute total_price as a product of amount and single_price names of the desired parameters are in the table heading Add a row enter the name of the new attribute on the left and enter the function on the right which computes this new attribute In this case it is simply the product of two other attributes Confirm your input with Apply and the dialog will close The button that says Edit List ought to show a 1 in brackets meaning that you can see how many entries the parameter list has and therefore in this case how many new attributes are generated We can now observe what effect the addition 63 3 Design of Analysis Processes of the operator Generate Attributes has on the meta data RapidMiner has already transformed the meta data in the background and you can see the new meta data as a tooltip via the output port of the operator Fig 3 9
42. displayed on the left hand side of the perspective and can be maximized again or looked at briefly from there 4 Detach Detaches the view from the current perspective and shows it within its own window which can be moved to wherever you want Now have a little go at arranging the two views in different ways Sometimes a little practice is required in order to drop the views in exactly the desired place It is worthwhile experimenting a little with the arrangements however because other settings may make your work far more efficient depending on screen resolution and personal preferences Sometimes you may inadvertently delete a view or the perspective is uninten tionally moved into particularly unfavourable positions In this case the View menu can help because apart from the possibility of reopening closed views via Show View the original state can also be recovered at any time via Restore Default Perspective Additionally you have the option of saving your own perspectives under a freely selectable name with the action New Perspective Fig 2 4 You can switch between the saved and pre defined perspectives either in the View menu or on the right side of the toolbar 26 2 3 Design Perspective 2 3 Design Perspective As already mentioned at the beginning you will find an icon for each pre defined perspective within the right hand area of the toolbar oe Home Design F8 Result
43. e Data 64 64 Store 01 Customer 1t 61867 Books 30 12 2007 8 95 811 Fas b3 78 78 Store 01 Customer 8 44627 Books 05 09 2006 3 98 397 5 5 Store 01 Customer 4 80069 Clothing 06 01 2008 5 65 215 PP 65 65 Store 01 Customer 7t 31254 Health 17 11 2007 1 44 732 Sg 15 15 Store 01 Customer 1 90837 Home Garden 28 04 2008 7 46 390 Charts 1 1 Store 01 Customer 1 53642 Toys 01 04 2007 3 90 246 55 55 Store 01 Customer 2 93469 Toys 15 11 2006 5 35 761 a 79 79 Store 02 Customer 1 61688 Books 16 04 2006 2 86 820 Advanced 56 56 Store 02 Customer 7 55677 Electronics 30 09 2005 8 75 169 _ 53 53 Store 02 Customer 2 36959 Movies 18 07 2008 6 67 188 92 92 Store 02 Customer 1 60439 Sports 07 01 2006 1 57 898 Annotation 24 24 Store 02 Customer 1 59553 Toys 07 12 2007 3 87 555 72 72 Store 02 Customer 1 13316 Toys 14 04 2007 6 29 414 25 25 Store 03 Customer 7 94389 Health 27 04 2007 1 72 882 Figure 4 8 Sorting was first performed in this table in ascending order according to the attribute store_id and then according to product category within the store ID blocks also in ascending order Note Sorting can be time consuming It is therefore deactivated for large tables so that no sorting is started inadvertently and the program cannot be used in this time You can set the threshold value at which sorting is deactivated in the settings under Tools Preferences Moving Columns You can change the order of columns in most tables by click
44. e colouring You can see a detailed view of the results by clicking on an entry In the case above the result consists of an example set and a decision tree A further click on the entry will close it again You can of course also open several entries at the same time and compare the results comfortably by doing so Two actions are available for each entry on the top right e To restore the process belonging to an entry in this form and e To delete the entry from the Result Overview In addition you have the option to delete the complete overview in the context menus of the Overview and of the individual entries 93 5 Managing Data The Repository Tables databases collections of texts log files websites measured values this and the like is at the beginning of every data mining process Data is prepared converted merged and at the end you will receive new or differently represented data models or reports In this chapter you will find out how to handle all these objects with RapidMiner Studio 5 1 The RapidMiner Studio Repository As soon as your collection of processes and associated files exceeds a certain size you will see it is wise to organise those in a consistent and structured manner One possibility is the organisation of projects on file level Files are grouped into projects and a directory is created in each case for output data intermediate results reports etc While creating organised project structur
45. e most similar events that have already taken place for example Did the falling article break in these cases or not We must first find the events with the greatest similarity which represents a kind of optimisation We use the term optimisation here since it is actually unimportant whether we are now maximising a similarity or the sales figures of one enterprise or any other the variable concerned so similarity here is always optimised The analogy reasoning described then tells us that the majority of glasses we have already looked at broke and this very estimation then becomes our prediction This may sound complicated but this kind of analogy reasoning is basically the foundation for almost every human learning process and is done at a staggering speed The interesting thing about this is that we have just been acting as a human data mining method since data analysis usually involves matters such as the repre sentation of events or conditions and the data resulting from this the definition of events similarities and of the optimisation of these similarities However the described procedure of analogy reasoning is not possible with the toss of a coin It is usually insufficient at the first step and the data for factors 1 Fundamental Terms such as material properties or ground unevenness cannot be recorded Therefore we cannot have these ready for later analogy reasoning This does in no way mean however that the event o
46. eactivated Their connections are detached and they are no longer executed Deactivated operators are shown in gray Operators can also be de activated within their context menu in the Process View as well as via the menu entry Edit Enable Operator Rename One of the ways to rename an operator Further ways are pressing F2 after selection selecting Rename in the context menu of the operator in the Process View as well as the menu entry Edit Rename Delete One of the ways to delete an operator Further ways are pressing DELETE after selection selecting Delete in the context menu of the operator in the Process View as well as the menu entry Edit Delete Toggle Breakpoints Breakpoints can be set here both before and after the exe cution of the operator where the process execution stops and intermediate results can be examined There is also this possibility in the context menu of the operator in the Process View as well as in the Edit menu A break point after operator execution can also be activated and deactivated with F7 2 3 6 Help and Comment View Help View Each time you select an operator in the Operators View or in the Process View the help window within the Help View shows a description of this operator This 47 2 First steps 7 iain o 5 Ele Em pocss Tools vew Help A98RF eax OPIS a Ss 2B Process Ef Parameters gt 5 a t Hiroes g Htr
47. ed above since the provision of results is guaranteed here 4 2 About Data Copies and Views The fact that no unnecessary data copies are created is sometimes a source for confusion This applies in particular for the second possibility of displaying re sults mentioned above i e via the context menu of ports Let us assume you have a data set and add an operator for normalisation In its presetting the nor malisation operator changes the underlying data Even if you look at the data set at a port which is before the normalisation in the process flow but chronolog ically after the normalisation was already performed then the data at the port will have also changed beforehand This behaviour should actually be quite clear as previously mentioned no copy of the data was created either and the same data set was changed further And yet this strange behaviour of uncontrolled data changes leads to confusion from time to time However you do have two ways of influencing this behaviour 1 Use of views Numerous operators for data transformations offer a param eter create view which instead of causing a change to the data merely causes a further view to be put on the data which changes the data on the fly so during data access These computations do then not affect previous ports or even ports in other parallel strands of the process 2 Explicit copies Especially for smaller data sets the combination of the oper
48. eing designed and all output ports are already connected Or if the intermediate result is deep inside an intricate subprocess There is of course a sophisticated solution in RapidMiner Studio for this too which does not make any process redesign necessary You can simply insert a so called breakpoint by selecting one of the options Breakpoint Before or Breakpoint After from the context menu of an operator as shown in Fig 3 15 If a breakpoint was inserted after an operator for example then the execution of the process will be interrupted here and the results of all connected output ports will be indicated in the Results Perspective This means you can look at these results without having to make further changes to the process design A breakpoint before an operator functions similarly to a breakpoint after an operator In this case the process will be interrupted before the execution of this operator and the objects next to the connected input ports of this operator are indicated The fact that a breakpoint is next to an operator is indicated by a small red symbol at the lower edge of the operator Fig 3 16 Tip The use of Breakpoint After in particular is relatively frequent which is why this action also has a keyboard shortcut You can add a breakpoint after the operator currently selected or remove all breakpoints currently present by pressing key F7 RapidMiner Studio automatically switches to the Result Perspective i
49. ere the results in the repository are to be saved to In the Context View you have the possibility of placing data from a repository at the entry ports and of writing outputs back into the repository You can give such an indication for each port This has two advantages 1 You can forget about the operators for Retrieve and Store which often makes your process somewhat clearer 105 5 Repository 2 Using the context is also practical for testing processes which are to be integrated by means of the operator Execute Process The data at this operator will overwrite the values defined in the process context 5 3 Data and Meta Data Apart from the actual data RapidMiner Studio also stores other information in the repository Data about the data so called meta data Such meta data is available for each type of object and it can be particularly useful for models and data sets The meta information stored for data sets includes for example The number of examples The number of attributes The types names and roles of the attributes e The ranges of values of the attributes or some fundamental statistics e plus the number of missing values per attribute This information can be seen in the repository without loading the data set beforehand which can take some time depending on size Simply move the cursor over a repository entry and stay on the entry for a few seconds The meta data will be presented to you in the
50. erms In this chapter we would like to give you a small incentive for using data mining and at the same time also give you an introduction to the most important terms Whether you are already an experienced data mining expert or not this chapter is worth reading in order for you to know and have a command of the terms used both here and in RapidMiner 1 1 Coincidence or not Before we get properly started let us try a small experiment e Think of a number between 1 and 10 e Multiply this number by 9 e Work out the checksum of the result i e the sum of the numbers Multiply the result by 4 e Divide the result by 3 e Deduct 10 The result is 2 1 Fundamental Terms Do you believe in coincidence As an analyst you will probably learn to answer this question in the negative or even do so already Let us take for example what is probably the simplest random event you could imagine i e the toss of a coin Ah you may think but that is a random event and nobody can predict which side of the coin will be showing after it is tossed That may be correct but the fact that nobody can predict it does in no way mean that it is impossible in principle If all influence factors such as the throwing speed and rotation angle material properties of the coin and those of the ground mass distributions and even the strength and direction of the wind were all known exactly then we would be quite able with some time and effort
51. es arise There are many synonyms for this term and depending on your own background you will have already come across different terms instead of attribute for example e Characteristic Feature e Influence factor or just factor e Indicator Variable or Signal 1 2 Fundamental Terms We have seen that description by attributes is possible for processes and also for situations This is necessary for the description of technical processes for example and the thought of the falling glass is not too far off here If it is possible to predict the outcome of such a situation then why not also the quality of a produced component Or the imminent failure of a machine Other processes or situations which have no technical reference can also be described in the same way How can I predict the success of a sales or marketing promotion Which article will a customer buy next How many more accidents will an insurance company probably have to cover for a particular customer or customer group We shall use such a customer scenario in order to introduce the remaining im portant terms Firstly because humans are famously better at understanding examples about other humans And secondly because each enterprise proba bly has information i e attributes regarding their customers and most readers can therefore relate to the examples immediately The attributes available as a minimum which just about every enterprise keeps about its cu
52. es is sensible using the normal file sys tem is recommended only in the rarest cases and is hardly sufficient for the needs of a data mining solution Different reasons such as confidentiality or limited storage space can make creating files on the local computer impossible If a pro cess created on the local computer is to be executed on a remote server this requires manual interventions like copying the process and adapting paths The collaborative creation of processes manipulation of data and evaluation of results 95 5 Repository requires an external rights and version administration Files stored in different formats require the correct setting of parameters such as separators and coding for each new loading Intermediate results and process variants soon grow to a considerable number meaning that one can lose track easily Loading and look ing at data in order to regain an overview requires a loading process that may be lengthy or even the running of an external application Annotations of files which can make this easier are not supported by normal file systems RapidMiner s answer to all these problems is the repository which takes up all data and processes Although data can also be introduced into processes from outside the repository which is necessary for the execution of ETL processes for example using the repository offers a number of advantages which you will not want to miss e Data processes results and rep
53. ese steps can be defined and detached again Finally it is even possible to define the order of the steps in this perspective The next sections show you how to use the Process View 2 3 3 Operators and Processes Working with RapidMiner Studio fundamentally consists in defining analysis pro cesses by indicating a succession of individual work steps In RapidMiner Studio 31 2 First steps Process Control 37 a E Repository Access 6 amp import 26 amp Export 17 E Data Transformation 114 GQ Modeling 118 Evaluation 29 E Repositories la s a sao 2 OE DB F Local Repository Chistian Pel E amp Main Repository c 2 gt F m t Process Process View Problems 3B No problems found Log Message Fixes Location IE Process logverbosity init ioone C m C sro ey 7 Hep Comment Process Synopsis The root operator which is the outer most operator of every process Description Each process must contain exactly one operator of this class and it must be the root operator of the process This operator provides a set of W lt new process gt RapidMiner Studio 6 0 000 rapid F0lo P W Jj ess Eile Edit Process Tools View Help GHC 97 ea Od S Jee Operators g Process Parameters Figure 2 12 In the Process View the components of RapidMiner the so called operator
54. ess completely as usual by pressing stop 73 4 Data and Result Visualization In the previous sections we have seen how the graphical user interface of Rapid Miner Studio is built up and how you can define and execute analysis processes with it At the end of such a process the results of the process can then be indicated in the Results Perspective Switch now to this Results Perspective by clicking once in the toolbar This will be dealt with in detail within this chap ter Depending on whether you have already produced representable results you should now see in the original settings at least roughly the screen shown in 4 1 Otherwise you can recreate this preset perspective under View Restore De fault Perspective as always The Result Perspective is the second central work ing environment of RapidMiner Studio alongside the Design Perspective already discussed We have already discussed the Repositories View on the right In this chapter we will therefore focus on the remaining component of the perspective 4 1 Result Visualization We have already seen that objects which are placed at the result ports at the right hand side of a process are automatically displayed in the Results Perspective after the process is completed The large area on the top left hand side is used here where the Results Overview is also already displayed which we will discuss at the end of this chapter 79 4 Data and Result Visualizati
55. f a coin toss is coincidental but merely shows that we humans are not able to measure these influence factors and describe the process In other cases we may be quite able to measure the influence factors but we are not able to correlate these purposefully meaning that computing similarity or even describing the processes is impossible for us It is by no means the case that analogy reasoning is the only way of deducing forecasts for new situations from already known information If the observer of a falling glass is asked how he knows that the glass will break then the answer will often include things like every time I have seen a glass fall from a height of more than 1 5 metres it has broken There are two interesting points here The relation to past experiences using the term always as well as the deduction of a rule from these experiences If the falling article is made of glass and the falling height is more than 1 5 metres then the glass will break The introduction of a threshold value like 1 5 metres is a fascinating aspect of this rule formation For although not every glass will break immediately if greater heights are used and will not necessarily remain intact in the case of lower heights introducing this threshold value transforms the rule into a rule of thumb which may not always but will mostly lead to a correct estimate of the situation Instead of therefore reasoning by analogy straight away one could now use
56. f a comment has been entered for this operator then this is indicated by this icon Subprocess This is a very important indication since some operators have one or more subprocesses It is shown by this indication whether there is such a subprocess You can double click on the operator concerned to go down into the subprocesses If several operators are interconnected then we speak of an analysis process or process for short Such a succession of steps can for example load a data set transform the data compute a model and apply the model to another data set Such a process may be in RapidMiner Studio like it is shown in Figure 2 15 Such processes can easily grow to several hundred operators in size in RapidMiner Studio and spread over several levels or subprocesses The process inspections continually performed in the background as well as the process navigation aids shown below ensure that you do not lose track and that you define correct pro cesses even for more complex tasks 34 2 3 Design Perspective Read Excel Select Attribu Set Role Decision Tree tra mod Q exa Apply Model Read Excel 2 Figure 2 15 An analysis process consisting of several operators The colour cod ing of the data flows shows the type of object passed on Inserting Operators You can insert new operators into the process in different ways Here are the details of the different ways e Via drag amp drop
57. f value type text in the case of free text of the value type numerical in the case of numbers and of the value type nominal in the case of only few values being possible like with the two possibilities yes and no for the target attribute Please note that in the above example the number of employees although really of numerical type would rather be defined as nominal since a size class i e gt 1000 was used instead of an exact indication like 1250 employees 12 1 2 Fundamental Terms The following table will give you an overview of all value types supported by RapidMiner RapidMiner Value type Use name Categorical non numerical values Nominal nominal usually used for finite quantities of different characteristics Numerical values numeric For numerical values in general Whole numbers positive and nega Integers integer tive Real numbers real Real numbers positive and negative Text text Random free text without structure Special case of nominal where only 2 value nominal binominal i two different values are permitted J Special case of nominal where more multi value nomi polynominal than two different values are permit na ted Date Time date_time Date as well as time Date date Only date Time time Only time 13 1 Fundamental Terms 1 2 5 Data and Meta Data We want to summarise our initial situation one more time We ha
58. forts from the beginning which are unlikely to lead to a purchase But how can that be done The need for alternative solutions and thus the interest in the new product arose within the customer study on a subset of your customers Performing this study for all your customers is much too costly and so this option is closed to you And this is exactly where data mining can help Let us first look at a possible selection of attributes regarding your customers e Name e Address Sector Subsector Number of employees Number of purchases in product group 1 e Number of purchases in product group 2 The number of purchases in the different product groups means the transactions in your product groups which you have already made with this customer in the past There can of course be more or less or even entirely different attributes in your case but this is irrelevant at this stage Let us assume that you have the information available regarding these attributes for every one of your customers Then there is another attribute which we can look at for our concrete scenario The fact whether the customer likes the prototype or not This attribute is of course only available for the 100 customers from the study the information on this attribute is simply unknown for the others Nevertheless we also include the attribute in the list of our attributes e Prototype positively received e Name e Address 1 2 Fundamental Terms Sector S
59. from the Operators View as described above e Via double click on an operator in the Operators View e Via dialog which is opened by the menu entry Edit New Operator Ctrl D e Via context menu in a free area of the white process area and there via the submenu New Operator and the selection of an operator In each case new operators are depending on the setting in the Operators View either automatically connected with suitable operators or the connections have to be made or corrected manually by the user 35 2 First steps Connecting Operators After you have inserted new operators you can interconnect the operators in serted There are basically three ways available to you which will be described in the following Connections 1 Automatically when inserting If you have activated the option for automatic connection under the plug symbol in the Operators View then RapidMiner will try to connect the operator with suitable output ports after inserting If for example the new operator has an input port which requires an example set then RapidMiner will try to find an operator that could already produce such an example set If there is only one option then this choice is clear and the operator is connected If there are several options however RapidMiner will try to select the option which is the closest on the left above the current mouse position The associated operator is marked with a frame and
60. hat designates the weight of the example with regard to the label In this case the role has the name Weight Attributes without a special role i e those which simply describe the examples are also called regular attributes and just leave out the role desig nation in most cases Apart from that you have the option in RapidMiner of allocating your own roles and of therefore identifying your attributes separately in their meaning 1 2 4 Value Types As well as the different roles of an attribute there is also a second characteristic of attributes which is worth looking at more closely The example of Doe Systems above defined the respective values for the different attributes for example Doe Systems Inc for the attribute Name and the value 5 for the number of past purchases in product group 1 Regarding the attribute Name the concrete value for this example is therefore random free text to a certain extent for the attribute number of purchases in product group 1 on the other hand the indication of a number must correspond We call the indication whether the values of an attribute must be in text or numbers the Value Type of an attribute In later chapters we will become acquainted with many different value types and see how these can also be transformed into other types For the moment we just 11 1 Fundamental Terms need to know that there are different value types for attributes and that we speak o
61. he most used operators 30 2 3 Design Perspective Tip Professionals will know the names of the necessary operators more and more frequently as time goes on Apart from the search for the complete name the search field also supports a search based on the initial letters so called camel case search Just try REx for Read Excel or DN for Date to Nominal and Date to Numerical this speeds up the search enormously Repositories View The repository is a central component of RapidMiner Studio which was intro duced in Version 5 It is used for the management and structuring of your anal ysis processes into projects and at the same time as both a source of data as well as of the associated meta data In the coming chapters we will give a detailed description of how to use the repository so we shall just say the following at this stage Warning Since the majority of the RapidMiner Studio supports make use of meta data for the process design we strongly recommend you to use the RapidMiner repository since otherwise for example in the case of data being directly read from files or databases the meta data will not be available meaning that numerous supports will not be offered 2 3 2 Process View The Process View Fig 2 12 shows the individual steps within the analysis process as well as their interconnections New steps can be added to the current process in several ways Connections between th
62. ies for models They can be easy to understand and explain the underlying processes in a simple manner Or they can be good to use for prediction in the case of unknown situations Sometimes both apply such as with the following model for example which a data mining method could have supplied for our scenario If the customer comes from urban areas has more than 500 employees and if at least 3 purchases were transacted in product group 1 then the probability of this customer being interested in the new product is high Such a model can be easily understood and may provide a deeper insight into the underlying data and decision processes of your customers And in addition it is an operational model i e a model which can be used directly for making a prediction for further customers The company John Paper for example satisfies the conditions of the rule above and is therefore bound to be interested in the new product at least there is a high probability of this Your goal would therefore have been reached and by using data mining you would have generated a model which you could use for increasing your marketing efficiency Instead of just contacting all existing customers and other candidates without looking you could now concentrate your marketing efforts on promising customers and would therefore have a substantially higher success rate with less time and effort Or you could even go a step further and analyse which sales channels
63. indicators which are still yellow You may now ask yourself And So I know the result beforehand and without process execution What do I get from that Well rather a lot You can now see at a glance what a particular operator or sub process is doing with the input data Since the meta data is also considerably smaller than the complete data sets this examination can also be performed much faster than on the complete data This way you get feedback in the shortest time as to whether there is a problem which may make further data transformation necessary and not only after an analysis process lasting several hours has aborted with an error And last but not least RapidMiner Studio can continue processing the information from the meta data and continue supporting you in the design of the process e g through only all attributes that are still available and newly generated being displayed on the graphical user interface while attributes are being filtered Now try the following for example Open the group Data Transformation At tribute Set Reduction and Transformation Selection and drag the operator named Select Attributes into the process ideally directly onto the connection after the last operator Remember that the connection must be highlighted be fore you drop the operator then it will be correctly reconnected immediately You should have now defined the process as in Fig 3 10 Select the new operat
64. ing customers would be more likely to 14 1 2 Fundamental Terms receive the prototype positively or reject it The search for such a prediction instruction is one of the tasks which can be performed with data mining However it is important here that the information for the attributes of the indi vidual examples is in an ordered form so that the data mining method can access it by means of a computer What would be more obvious here than a table Each of the attributes defines a column and each example with the different attribute values corresponds to a row of this table For our scenario this could look like in table 1 1 for example We call such a table an Example Set since this table contains the data for all the attributes of our examples In the following and also within RapidMiner we will use the terms Data Data Set and Example Set synonymously A table with the appropriate entries for the attribute values of the current examples is always meant in this case It is also such data tables which have lent their name to data analysis or data mining Note Data describes the objects of a concept Meta Data describes the characteristics of a concept and therefore also of the data Most data mining methods expect the examples to be given in such an attribute value table Fortunately this is the case here and we can spare ourselves any further data transformations In practice however this is completely different and the m
65. ing on the column heading and dragging the column to a new position while holding down the mouse button This can be practical if you wish to compare the contents of two columns with one another in large tables Adjusting Column Widths You can adjust the width of columns by holding the cursor over the area between two columns and changing the width of the column to the left of the separation 84 4 3 Display Formats area while holding down the mouse button Alternatively you can also double click on this gap which causes the width of the column to the left of the gap to be automatically adjusted to the necessary minimum size Last but not least you can also hold the Ctrl key down when you double click on a gap causing the size of all columns to be adapted automatically Tip You should note this combination CTRL double click on a gap in the column heading area so that you can quickly adjust column widths Actions in the Context Menu In most tables you can open a context menu with further actions by right clicking on a table cell The details of these actions are 1 10 11 Select Row Selecting a line Select Column Selecting a column Fit Column Width Adjusting the width of the selected column Fit all Column Widths Adjusting all column widths Equal Column Widths Using same standard width for all columns Sort by Column Ascending Sorting by this column in ascending order Sor
66. inting right the arrow pointing upwards and the navigation bar breadcrumb Process 4 f Process gt 72 EYEDA Figure 2 18 Actions in the Process View The individual actions Arrow pointing left Returns to the last place of editing in a similar way to the navigation which is familiar from internet browsers Individual steps can also be skipped via the pop up menu Arrow pointing right Returns to the most recent editing places in the history in a similar way to the navigation which is familiar from internet browsers Individual steps can also be skipped via the pop up menu Arrow pointing upwards Leave the current subprocess and return to the greater process Navigation bar The navigation bar shows the way from the main process to the current subprocess via all levels passed through Clicking once on one of the operators will show the process concerned You can navigate further downwards using the small arrows pointing right In order to therefore descend into a subprocess you need to double click on an operator with the subprocess icon at the bottom on the right In order to go a level up again you can navigate upwards using the arrow The current path 40 2 3 Design Perspective is shown by the navigation bar Fig 2 19 which can alternatively be used to navigate in both directions lt Process v v T lab per f exa f per e Figure 2 19 A subprocess named Validation
67. ion for storing the process After a process has been stored in the repository all references to repository entries set as parameters of operators are resolved in relation to the location of the process What does that mean Entries in the repository are designated as follows RepositoryName Folder Subfolder File The two slashes at the beginning indicate that the name of a repository will follow first Then further folder names and finally a file name We call such details absolute In the following description Folder Subfolder File the repository designation is missing at the front This description is therefore repository relative It refers to the file described in the same repository where the process in which this description is used is located The slash at the front indicates an absolute path description If this is also missing the description relative is resolved 99 5 Repository RelativeFolder File designates for example a file in the folder RelativeFolder which we reach by moving up a directory from the file containing the current process and looking for the folder RelativeFolder there So if the process is located for example in the file MyRepository ProjectA Processes ProcessB this description leads to MyRepository ProjectA RelativeFolder File Note The descriptions above probably sound more complicated than they really are in practice As long as before anything else y
68. iple you are completely free in how you structure your repository In the context menu of the entries in the repository browser and also in the repository view you will find all necessary entries for the administration of your data and processes as you can see them in Fig 3 1 Repositories E a SITH 3 Samples none H DB E amp Local Repository Christian Pels Q data Christisn Pels mee aea Store Process Here Rename F2 amp Create Folder E Copy Strg C Paste Strg V E Copy Location to Clipboard Delete Entf G Refresh F5 tq Open in file browser Figure 3 1 The context menu of the repository entries both in the repository browser and in the repository view offers all necessary options for administration Store Process Here Stores the current process in the given location 54 3 2 Repository Actions Rename Renames the entry or the directory Create Folder Creates a new directory in this place Copy Copies the selected entry for later insertion in other places Paste Pastes a previously copied entry to this place Copy Location to Clipboard Copies a clear identifier for this entry to the clip board meaning you can use this as a parameter for operators in web interfaces or the like Delete Deletes the selected repository entry or directory Refresh Updates the display Open in file browser Shows the repository in a file browser This is not recom mended Repositories E a SIH
69. is name has been added in the Home Perspective 23 2 First steps View Run Tools Help Perspectives gt C New Perspective a Show View gt C3 Restore Default Perspective v amp Expert Mode F4 Figure 2 4 View menu Ele Edt Process Tools wew Help TERS ea SPIE B ee SB 2 Home Home Screen New Process News Create your own analysis RapidMiner v6 0 Makes Predictive Analytics More Accessible to All The primary focus in RapidMiner 6 is the addition of application wizards a new cless offeature designed to make the power of predictive analyics accessible to everyone particularly business manage s and business analysts aplaMiner A Potential Game Changer Open David Noris Practice Leader Analytics at Bloor Research published an oventew on Repidiner x on Monday November 16 2013 In E he called Rapiiner a potential game change Open previous analysis pe has improved their products dramaticaly forn the very technical frameworks Inat existed bu voroven aboutto change we now have the delivery of Rapidliner as a serious contender with the financial backing to mako an impact on the incumbents road mors Application Wizard Rapid rebrands as Rapidtiner Data mining in three simple steps On November 4 Rapid announced t wil ve now known as Rapidviner retlecing Me company s shitin fecus TecnCninch coveredithe story in cept takang wih one ot me company s investors
70. itories View The root operator which is the outer most operator of every process Description Each process must contain exactly one operator of this class and it must be process like logging and initialization of parameters of the random number generator Figure 2 10 Design Operators of RapidMiner Without extensions you will find at least the following groups of operators in the tree structure Process Control Operators such as loops or conditional branches which can control the process flow Utility Auxiliary operators which alongside the operator Subprocess for group ing subprocesses also contain the important macro operators as well as the operators for logging Repository Access Contains operators for read and write access in repositories Import Contains a large number of operators in order to read data and objects from external formats such as files databases etc Export Contains a large number of operators for writing data and objects into external formats such as files databases etc 29 2 First steps Data Transformation Probably the most important group in the analysis in terms of size and relevance All operators are located here for transforming both data and meta data Modeling Contains the actual data mining process such as classification meth ods regression methods clustering weightings methods for association rules correlation and similarity analyses as
71. l find further ways of accessing the repository by right clicking once on entries in the repository You will be familiar with these possibilities from the file 103 5 Repository management of your computer These actions are also available via the toolbar of the repository view and are largely self explanatory Store Process here Stores the current process to the location indicated Rename Renames the entry or the directory Create Folder Creates a new folder here Delete Deletes the selected repository entry or directory Copy Copies the selected entry so it can be pasted in other places later on Paste Pastes a previously copied entry to this place Copy Location to Clipboard Copies a clear identifier for this entry onto the clip board meaning you can use this as a parameter for operators in web in terfaces or the like Open Process If you have selected a process the current process will be closed and the selected one loaded Refresh If the repository is located on a shared file system or if you use Rapid Miner Server meaning data can be changed at the same time by other users you can refresh the view of the repository with this 5 2 4 The Process Context We have already used the output ports of the process on the right hand side of the Process View previously e g in order to make the results of the process visible in the Result Perspective In addition to the output ports of the process there are also input ports which yo
72. ly select an operator and write any text on it in the comment field This will then be saved together with your process definition and can be useful for tracing individual steps in the design later on The fact that a comment is available for 48 2 3 Design Perspective an operator is indicated by a small text icon at the lower edge of the operator 2 3 7 Overview View Particularly in the case of extensive processes the white work area will no longer be sufficient and will be enlarged either via the context menu of the Process View by means of the key combinations of Ctrl and the arrow pointing left right upwards and downwards or simply by dragging an operator to the edge In this case however the entire work area will no longer be visible at the same time and navigation within the process will be made more difficult In order to improve the overview and provide a comfortable way of navigating at the same time RapidMiner Studio offers the Overview View Fig 2 25 which shows the entire work area and highlights the currently displayed section with a small box p Overview _ 3 2 2 gt 3 4 z gt A 5 hel x a s a Fi E ThaL u 4 3 4 a K y Figure 2 25 The Overview View shows the entire process and highlights the vis ible section You will see that the section moves within the Process View when scrolling now using the scrollbar or simply by dragging an operator to the edge of the section Or you can
73. m eon Tutorials eci manteing 1 Seamer anes Learn the basics of RapidMiner Predidive maintenance Howie predictive analytics like hockoy now known as Rapiduiner refectino me TNE TNO SUY m pth Tang W one OF re from Eariyaird Ventures What ie an application wizard Action mizards are a new class ot feature desioned to make he nomer o Rantaine creche anaes jone panic paene aponcanon Evert to describe what you do ata cocktail party Or to your neighbor Check out our Pragicve Analiics for he Win deo Figure 2 3 Home Perspective of RapidMiner Studio On the right hand side of the toolbar inside the upper section of RapidMiner Studio you will find four icons which switch between the individual RapidMiner Studio perspectives A perspective consists of a freely configurable selection of individual user interface elements the so called views Those can be arranged however you like In the Home Perspective there is only one view one preset at least namely the home screen which you are looking at now You can activate further views by accessing the View menu Fig 2 4 In the subitem Show View you will find all available views of RapidMiner Studio Views which are now visible in the current perspective are marked with a tick Activate a further view by making a selection for example the view with the name Log You will now see in Fig 2 5 that a second view with th
74. me 0 s New Process 1 results Process results Completed Nov 22 2013 1 23 30 PM execution time 0 s New Process 1 results Process results Completed Nov 22 2013 1 24 33 PM execution time 0 s New Process 2 results Process results Completed Nov 22 2013 1 25 36 PM execution time 0 s New Process 1 results Process results Completed Nov 22 2013 1 29 00 PM execution time 0 s New Process 1 results Process results Completed Nov 22 2013 1 31 26 PM execution time 0 s New Process 2 results Process results Completed Nov 22 2013 1 32 10 PM execution time 0 s Tree Decision Tree ExampleSet Retrieve Deals Data Table 4 attributes Role Name Age Gender New Process 2 results Process results Completed Nov 22 2013 1 32 19 PM execution time 0 s Figure 4 14 The Result Overview indicates the results of the last processes 92 Source Samplesidata Deals Number of examples 1000 Type Range Missings no D7 missing mean values integer no binominal _ female missing J analysis 4 4 Result Overview The Result Overview serves as a compact overview of all process executions of the current RapidMiner session Each two line entry consists of the name of the process the number of the results as well as information on when the process was completed and how long it ran for Each block displaying results of the same process have an alternat
75. n Sokin e o o eo o Advanced amount Charts Store 09 iet bd ns C Log Scale z S Jitter g Store 08 Annotation i ay 3 EF Store 07 4 o W Rotate Labels Store 06 e o b aiin i Store 05 oo e k e Store 04 g o o Store 03 o e o Store 02 e e kd Store 01 oo eo o Figure 4 10 Visualisation of a data set and the chart configuration on the left hand side The most important setting can be found left at the top and corresponds to the type of visualisation More than 30 different 2D 3D and even high dimensional visualisation methods are available for displaying your data and results In the image above you will see a plot of the type Scatter Depending on the type of the selected chart all further setting fields change With a scatter plot for example you indicate the attributes for the x axis and for the y axis and can use a third attribute for colouring the points You can do further things specific to the scatter plot such as indicate whether the axes are to be scaled logarithmically 87 4 Data and Result Visualization Tip The Jitter function is very helpful especially for data sets which do not only contain numbers but also nominal values You indicate whether and how far the points are to be moved away from their original position in a random direction You can therefore make points which would otherwise be covered by other points easily visible Many charts also allow f
76. n aos oog EE system Montor Figure 4 7 Tables in RapidMiner often indicate interesting information with colours In this case darker backgrounds highlight stronger corre lations between attributes Sorting Most tables can be sorted in RapidMiner with a simple click Move the cursor roughly into the centre of the column heading and click on the heading A small triangle will now indicate the sorting order A further click will change the sorting order and a third click will deactivate the sorting again You can sort also according to several columns at the same time i e sort by one column first and then by up to two further columns within this sorting In order to do this start by sorting the first column and sort in the desired order Now press and hold the Ctrl key while you add further columns to the sorting In the following example we have sorted the transactions according to the ID of the store first of all and then by the category of the article The order of the columns within this sorting is symbolised by triangles of different sizes ranging 83 4 Data and Result Visualization from large to small Fig 4 8 g Result Overview E ExampleSet Generate Sales Data mn ExampleSet 100 examples 1 special attribute 7 regular attributes Filter 100 100 examples all X mi RowNo transaction_id store id customer_id product_id product _categon date amount single_pric
77. n be created The Operator Store If you have an ETL process or another process the result of which you would like to store in the repository you can do this by integrating the operator Store into your process E Parametrs Process gt 2 OLFEAJTTDIDPRR E E Repository Access 1 Store ea _ Se Ls I Repositories gag ague 205 Brew 0B 1 amp Local Repostiny E Main Repositoy B Store Rapiartines Studio Core Synopsis This operator stores an IO Object in the data story This operator stores an IO Object at a location in the data repository The location of the object ResoWve relative to Local Repositoryiprocesses 10 be stored is specified through he repository entry parameter The stored abject can be used by other processes by using the Retrieve operator Please see the attached Example Processes to understand the basic working of nis operator The Stare operator Is used 10 store an ExampleSet and a model in the Example Processes Location New Process Taput gy inu This port expects an IO Object In the attached Example Processes an ExampsSet and a model are provided as e Figure 5 4 The operator Store can be used to store any data and objects in the repository The dialog shows the repository browser so that the storage location can be specified and appears in the parameters of the
78. n the case of a breakpoint and shows the intermediate results You can see that you are in a 71 3 Design of Analysis Processes amp Show Operator Info F1 v Enable Operator Strg E P Rename F2 Replace Operator gt a Save as Building Block fe Cut Strg X E Copy Strg C Paste Strg E Delete Entf ig Breakpoint Before Breakpoint After F X All Breakpoints Debug Mode Figure 3 15 You can stop the process cycle using breakpoints and examine in termediate results Generate Attri Figure 3 16 A breakpoint is defined before or after this operator breakpoint at this time and not for example at the end of the process by looking at two indicators First of all the status indicator in the bottom left hand corner of the main window of RapidMiner shows a red light i e a process is running but is not being actively executed at present If no process at all is running at present then this indication would just be gray The second indicator for a breakpoint is the play symbol which is now green instead of blue Fig 3 17 72 3 4 Executing Processes gt is Figure 3 17 The green play symbol indicates that the process is currently in a breakpoint and can continue being executed if pressed The process can now be started again simply by pressing the green play symbol and continue being executed until completion or until the next breakpoint You can of course abort the proc
79. n you an idea of the kind of problems we wish to address in this book We will be dealing with numerous influence factors some of which can only be measured insufficiently or not at all At the same time there are often so many of these factors that we risk losing track In addition we also have to deal with the events which have already taken place which we wish to use for modelling and the number of which easily goes into millions or billions Last but not least we must ask ourselves whether describing the process is the goal or whether analogy reasoning is already sufficient to make a prediction And in addition this must all take place in a dynamic environment under constantly changing conditions and preferably as soon as possible Impossible for humans Correct But not impossible for data mining methods 1 2 Fundamental Terms We are now going to introduce some fundamental terms which will make dealing with the problems described easier for us You will come across these terms again and again in the RapidMiner software too meaning it is worth becoming acquainted with the terms used even if you are an experienced data analyst First of all we can see what the two examples looked at in the previous section namely the toss of a coin and the falling glass have in common In our discussion on whether we are able to predict the end of the respective situation we realised 1 Fundamental Terms that knowing the influence factors as
80. ncidences The reason for our inability to describe and extrapolate the processes precisely is rather down to the fact that we are not able to recognise or measure the necessary influence factors or correlate these 1 1 Coincidence or not In the case of the falling glass we quickly recognised the most important char acteristics such as the material falling height and nature of the ground and can already estimate in the shortest time the probability of the glass breaking by analogy reasoning from similar experiences However it is just that we cannot do with the toss of a coin We can watch as many tosses of a coin as we like we will never manage to recognise the necessary factors fast enough and extrapolate them accordingly in the case of a random throw So what were we doing in our heads when we made the prediction for the state of the glass after the impact We measured the characteristics of this event You could also say that we collected data describing the fall of the glass We then reasoned very quickly by analogy i e we made a comparison with earlier falling glasses cups porcelain figurines or similar articles based on a similarity measure Two things are necessary for this firstly we need to also have the data of earlier events available and secondly we need to be aware of how a similarity between the current and past data is defined at all Ultimately we are able to make an estimation or prediction by having looked at th
81. ng the repository you will be welcomed into the Home Perspective Fig 2 3 The right section shows current news about RapidMiner if you are connected to the Internet The list in the centre shows the typical actions which you will perform frequently after starting RapidMiner Studio Here are the details of those 1 New Process Opens the design perspective and creates a new analysis process 2 Open Opens a repository browser if you click on the button You can choose and open an existing process in the design perspective If you click on the arrow button on the right side a list of recently opened processes appears You can select one and it will be opened in the design perspective 21 2 First steps ro New Repository Alias NewLocalRepository Root directory Use standard location a Enter parameters to create a new local or RapidMiner Server repository CACEN Figure 2 2 Definition of a new local repository for storing your data and analysis 22 processes It is advisable to create a new directory as a basis Either way RapidMiner Studio will then automatically switch to the Design Perspective Application Wizard You can use the Application Wizard to solve typical data mining problems with your data in three steps The Direct Marketing Wizard allows you to find marketing actions with the highest conversion rates The Predictive Maintenance Wizard predicts necessary maintenance activitie
82. none w1 Z amp processes none Q 01_Learner none amp 02_Preprocessing none GB 03_Validation none E 04_Attributes none GQ 05_Visualisation none a SE _Meta e 01_ParameterOptimization none w1 98 02_ParameterSetter none w1 w 03_RandomOptimization none w1 63 04_LearningCurve none w1 e 05_MultipleLabelLearning none w1 w O6_OperatorEnabler none w1 4 p 07_EvolutionaryParameterOptimization none 09_ParialExampleSetLearner none w1 B 10_DifferentStructures none w1 11_SubsetlterationAndMacros none w1 E e Clustering none E Gy 08_Other none LocalRepository boec v1 Figure 5 1 The repository view with an opened example directory 5 1 1 Creating a New Repository In order to be able to use the repository you must first create one RapidMiner Studio asks you to do this when it is started for the first time You can later 97 5 Repository add further repositories by using the first button in the toolbar of the repository view The following pictures show the simple procedure If you do not use the RapidMiner Server select the first option to create a local repository and then choose Next Now give your repository a name and choose a directory for it to be created in Close the dialog with Finish You can now use your repository ay New Repository a Enter parameters to create a new local or RapidMiner Server repository New local
83. on File Edit Process Tools View Help 29BRF ea Orla a i EE Result Ovemew in Desn ra Figure 4 1 Result Perspective of RapidMiner Each currently opened and indicated result is displayed as an additional tab in this area like in Figure 4 2 Strictly speaking each result is also a view which you can move to anywhere you wish as usual In this way it is possible to look at several results at the same time You can of course also close individual views i e tabs by clicking once on the cross on the tab The other functionalities of views such as maximisation are also completely available to you here RapidMiner Studio will close the old results before the new results are displayed 4 1 1 Sources for Displaying Results There are several sources from which you can have results displayed We will present all ways to you in the following 76 4 1 Result Visualization Edt Process Tools View Help 2G0ORS ra ODUS a ga 2B E Resun venon B ExampleSet Generais Aoves UB EranpieSat Selec aarus B revostones ExampleSet 100 examples 1 special attribute 8 regular afributes Fiter 100 100 examples an Jai sae BE e aaa Data 1 1 StoreO1 Customer 1 53842 Toys 01042007 3 90246 270739 Q F amp a 5 2 swore 15 Customer 1i 90945 Moves 15022005 2 osas 121173 _ enteaol lle 3 3 Siow 12 Customer 18548 Moves 27 092007 5 asia 482083 or Santee tos customers ato outs orosaos 5 16903 aasta cares
84. on af application wizards a now class offeature designed to make the power ofpredicive analytics accessible to everyone particularly business managers and business analysts Ropidliner A Potentiai Game Changer Open David Norns Practce Leader analytes at Bloor Research published an aveniew on Rapidliner NE on Monday November 18 2013 Init he called Rapiclner a potental game changer Open previous analysis Everyone has improved thelr products dramatically rom te vary technical frameworks that existed E atthe end of he last century Eut as good as these improvements have been ther have failed to realy delver the power speed and ease of use thats tequiredto break the mould and unleash he potential of on demand data mining in the hands of a businessfocussed analyst That could all be boul io chenge we now have te delivery of RapidMiner as a serious contendet wit tne Manca aaron ra ar pac unm semen DEEE 08 Application Wizard apd reba Bapdaner Data mining in three simple steps On Novembar 4 Rapid announced twill be now known as RapidNiner refecting the company shiftn focus TechCrunch covered the story in depth talking with one ofthe company s investors Jason Whitmore tom Eariyird Ventures Whats an application wizard Doce sueriweste ste aN Gaa TEA target rae me PO REI i prodicve analytes ccossibleo avoryono particularly business managers and business Tutorials analysts Repalinr v6 0 provides four spec apalca
85. on methods are called Quick Fixes The second column gives an overview of such possible solutions either directly as text if there is only one possibility of solution or as an indication of how many differ ent possibilities exist to solve the problem In the example above there are two different possibilities for handling the second problem But why is this solution suggestion called quick fix Just try double clicking on the relevant quick fix field in the table in such a case In the first case the solution suggestion would be directly executed and a relevant operator automatically configured and inserted 50 2 3 Design Perspective in such a way that the necessary pre processing is performed In the second case with several solution possibilities a dialog would appear asking you to select the desired solution method In this case one or more necessary operators would be configured and inserted in such a way that the problem no longer arises In this way you can recognise problems very early on and with just a few clicks very comfortably eliminate them during the design process Note The determination of potential problems as well as the generation of quick fixes are among the functions of RapidMiner Studio that are dependent on meta data being supplied correctly We strongly recommend you use the repository since otherwise e g in the case of direct reading of data from files or databases the meta data will not be available an
86. operator if the Directory button is clicked on Using the operator Generate Data the example process in this picture gener ates a data set which is to be stored in the repository The Store operator only has one parameter repository_location If you press the button with the 101 5 Repository folder next to this parameter you will get a dialog in which you can first assign a folder in the repository and then a name for the data set Figure 5 4 If you execute the process you will see that a new entry will appear in the repository containing the generated data set The store operator is therefore particularly useful for data integration and transformation processes which are to be per formed automatically or regularly for example within the process scheduler of the RapidMiner Server Using the wizard as described above is definitely the more frequently used way to ensure a one off and fairly interactive integration of data Note You can not only connect data sets with the store operator but also models and all other RapidMiner Studio objects You can therefore also store any results in your repository Importing other formats with operators The repository stores data sets in a format which contains all data and meta data needed by RapidMiner Studio Your data will probably be in another format at the beginning CSV Excel SQL databases etc As described above you can transfer these files into your
87. or and select the option subset in its parameters for the parameter attribute filter type Please note that a further parameter named attributes has now appeared This is in bold so you would need to define it before you could perform the process You can also see this from the red status indicator of the operator as well as from the entry in the Problems View You could now choose the quick fix in the Problems View by double clicking or simply configure the parameter attributes Again by clicking on a button this time the one that says Select Attributes The parameters should be like in Fig 3 11 Now press the button that says Select Attributes and select the attributes product_category store_id and total_price from the list in the dialog Fig 65 3 Design of Analysis Processes lt Process XML BH Process gt 2 E Y amp A Generate Sale Generate Attri Select Attribu Figure 3 10 Generation of data generation of a new attribute selection of a subset of attributes Parameters E SBS eeork amp B Select Attributes attribute filter type attributes Q Select Attributes _ invert selection include special attributes Figure 3 11 The parameter attributes only appears if subset has been chosen as the filter type 66 3 3 The First Analysis Process 3 12 that appears either by do
88. orts are stored in locations indicated relative to one another in a mechanism that is transparent for the user e Opening or loading the files requires no further settings Data can be opened looked at or incorporated into the process with a single click You will get an overview of the stored data its characteristics and remarks made by yourself at any time without having to open the file separately e All input and output data plus intermediate results are annotated with meta information This ensures the consistency and integrity of your data and makes validating processes possible at the time of development as well as the provision of context sensitive assistants The repository can either be on a local or shared file system or be made available by the external RapidMiner Server The following picture shows the repository view which displays the content of the repository RapidMiner Studio provides a set of example processes and example data which you will find in the repository initially created Some of these can be seen in Figure 5 1 96 5 1 The RapidMiner Studio Repository Repositories g3 amp Operators a E T Samples none cS Q data none F Golf none v1 ig GolfTestset none w1 ig IriS none v1 Labor Negotiations none v1 8 Market Data none v1 ig Polynomial none w1 E Ripley Set none v1 F Sonar none v1 E Transactions none v1 ig Weighting
89. ou define a location within the repository for each new process and then simply use the repository browser for all operator parameters requiring an entry in the repository RapidMiner Studio will ensure fully automatically that relative data is always used as far as possible This makes repository restructuring and making copies for other users easier in particular which would be difficult with absolute descriptions 5 2 2 Importing Data and Objects into the Repository There are numerous ways to import data and other objects such as models into the repository We will now describe the most important ones Importing Example Sets with Wizards If you have data in a certain format and wish to use it in a RapidMiner Studio process so called wizards are available to you for many file formats and databases A wizard is a dialog which guides you step by step through the loading process With all wizards you can assign certain meta data such as attribute types ranges of values and roles for the individual columns In the upper area of the repository you will find an icon which starts the appropriate wizard for the selected file type You will find the same action in the File menu of RapidMiner Studio Finally 100 5 2 Using the Repository there is another particularly simple way to import files Simply drag the file to be imported into the process view while holding down the mouse button If possible an appropriate operator will the
90. ple a bluish colour indicates that an example set is required If the upper half of the port and the name of the port are red then this indicates a problem This problem is easy to see for the operator in figure 2 13 it is not connected and the input ports still need a connection to a suitable source Output ports are white if the result is unclear or cannot yet be supplied in the current configuration As soon as all necessary configurations are complete i e all necessary parameters are defined and all necessary input ports connected then the output ports are coloured according to their type Figure 2 14 Status indicators of operators 33 2 First steps But not only the ports can visualise their status by means of different status indicators but also the complete operator Fig 2 14 These are given from left to right by Status light Indicates whether there is a problem like parameters that have not yet been set or unconnected input ports red whether the configuration is basically complete but the operator has not yet been implemented since then yellow or whether everything is OK and the operator has also already been implemented successfully green Warning triangle Indicates when there are status messages for this operator Breakpoint Indicates whether process execution is to be stopped before or af ter this operator in order to give the analyst the opportunity to examine intermediate results Comment I
91. propriate installation package for your operating system and install RapidMiner Studio according to the instructions on the website All usual Windows versions are supported as well as Macintosh Linux or Unix systems Please note that an up to date Java Runtime at least version 7 is needed for the latter If you are starting RapidMiner Studio for the first time you will be asked to create a new repository Fig 2 1 We will limit ourselves to a local repository on your computer first of all later on you can then define repositories in the network which you can also share with others For a local repository you just need to specify a name alias and define any directory on your hard drive Fig 2 2 You can select the directory directly by clicking on the folder icon on the right It is advisable to create a new directory in a convenient place within the file dialog that then appears and then use this new directory as a basis for your local repository This repository serves as a central storage location for your data and analysis processes and will accompany you in the near future 20 2 2 Perspectives and Views New Repository a Enter parameters to create a new local or RapidMiner Server repository New local repository New RapidMiner Server repository Figure 2 1 Create a local repository on your computer to begin with the first use of RapidMiner Studio 2 2 Perspectives and Views After choosi
92. repository New RapidMiner Server repository e gt gt nist 9G cancer Figure 5 2 You can use a repository on a shared RapidMiner Server or select a local repository E New Repository g Enter parameters to create a new local or RapidMiner Server repository Alias NewLocalRepository Root directory Use standard location IKI CCE Figure 5 3 RapidMiner Studio asks for the name and directory for a newly cre ated local repository 98 5 2 Using the Repository 5 2 Using the Repository It makes sense to use a uniform directory structure for projects for example a project folder with the name of the project and a folder each for processes input data and results All examples in this manual follow this structure You can create directories using the context menu in the repository view or using the button in the toolbar at the top of this view 5 2 1 Processes and Relative Repository Descriptions Before discussing in the following sections how you can store data and processes in the repository and access these again we would like to first give some fundamental tips on referencing these objects within the repository You can store processes in the repository by selecting the entry Store Process in the context menu or by selecting the appropriate entry in the File menu In the latter case the repository browser opens where you can indicate the locat
93. rocess t v Hted Retrieve inp wg out q res 2 oiscretize example set output example set output ae Meta data Data Table Number of examples 150 6 attributes Generated by Discretize example set output Normalize example set output Retrieve output Role Name Type Range Missings Comment al real 4 300 0 a4 real 0 100 0 id id nominal 2lid_1 id 0 label label nominal lris set 0 a3 real 0 1 0 a2 nominal range1 0 Press F3 for focus Figure 5 7 Meta data transformation in RapidMiner 109 5 Repository You are surely wondering why the parameter range name type had to be set to short See for yourself and set it to long If you execute the process you will see that the nominal values now give more information They additionally contain the limits of the intervals created This is handy but insignificant for the process The information on the interval limits is not available however as long as the discretization has not actually been performed Therefore it cannot be considered for the meta data display at the time of process development It can then only be indicated in the meta data that the range of values of the discretized attribute is a superset of the empty set i e that it is not empty This means that the meta data is not fully known So in this case we can say virtually nothing at all about the expected meta data
94. rs Bold font means that the parameter must absolutely be defined and has no default value Italic font means that the parameter is classified as an expert parameter and should not necessarily be changed by beginners to data analysis Figure 2 23 shows the Parameters View in detail Parameters Seer B W Nominal to Date attribute name date type date v date format C J7 ke time zone SYSTEM v C keep old attribute Figure 2 23 The parameters of the operator nominal to date Please note that some parameters are only indicated when other parameters have a certain value For example an absolute number of desired examples can only be indicated for the operator sampling when absolute has been selected as the type of sampling The actions of the toolbar refer just like the parameters to the operator currently selected Expert Mode The icon on the left switches between expert mode and beginner 46 2 3 Design Perspective mode Only in the expert mode are all parameters shown in the beginner mode the parameters classified as expert parameters are not shown Operator Info Display of some fundamental information about this operator such as expected inputs or a description This dialog is also displayed by pressing F1 after selection via the context menu in the Process View as well as via the menu entry Edit Show Operator Info Enable Disable Operators can be temporarily d
95. rther information particularly on process execution and in the event of an error can be found in the Log View You can copy the text within the Log View Fig 2 27 as usual and process it 51 2 First steps further in other applications You can also save the text in a file delete the entire contents or search through the text using the actions in the toolbar 52 3 Design of Analysis Processes We became acquainted with the fundamental elements of the graphical user in terface of RapidMiner Studio in the last chapter such as perspectives and views and discussed the most important aspects of the Design Perspective of Rapid Miner Studio We would now like to make use of the new possibilities in order to define and execute an initial simple analysis process You will soon realise how practical it is that with RapidMiner Studio you do not need to perform the process again for every change in order to determine the effect of the change But more about it later 3 1 Creating a New Process Whether you now select the action New from the Home Perspective the New icon on the left side of the main toolbar of RapidMiner Studio or the associated entry in the File menu A new analysis process is created in each case which you can work on in the following Tip We recommend to save the process in your repository when it is created 53 3 Design of Analysis Processes 3 2 Repository Actions In princ
96. rticles in that customers will typically have no material properties and falling articles will only rarely buy in product group 1 It is important that for each of the problems in this book or even those in your own practice you first define which concepts you are actually dealing with and which attributes these are defined by We implicitly defined above by indicating the attributes name address sector 1 Fundamental Terms etc and in particular the purchase transactions in the individual product groups that objects of the concept customer are described by these attributes Yet this concept has remained relatively abstract so far and no life has been injected into it yet Although we now know in what way we can describe customers we have not yet performed this for specific customers Let us look at the attributes of the following customer for example e Prototype positively received yes e Name Doe Systems Inc e Address 76 Any Street Sunnyville Massachusetts e Sector Mechanics e Subsector Pipe bending machines e Number of employees gt 1000 e Number of purchases in product group 1 5 e Number of purchases in product group 2 0 We say that this specific customer is an example for our concept customer Each example can be characterised by its attributes and has concrete values for these attributes which can be compared with those of other examples In the case described above Doe Systems Inc is al
97. s F9 Wizard Figure 2 8 Toolbar Icons for Perspectives The icons shown here take you to the following perspectives Home Perspective The Welcome Perspective already described above which RapidMiner welcomes you with after starting the program Design Perspective This is the central RapidMiner Studio perspective where all analysis processes are created edited and managed Result Perspective If a process supplies results in the form of data models or the like then RapidMiner Studio takes you to this perspective It provides statistics charts advanced charts and more Wizard Perspective This is the perspective which shows you the Application Wizard to apply typical data mining problems on your data You can switch to the desired perspective by clicking inside the toolbar or alter natively via the menu entry View Perspectives followed by the selection of the target perspective RapidMiner Studio will eventually switch to another perspective if it seems a good idea e g to the Result Perspective on completing an analysis process Now switch to the Design Perspective by clicking in the toolbar It will be dealt with in detail in this section The Result Perspective is the topic of chapter 4 You should now see the screen in Figure 2 9 Since the Design Perspective is the central working environment of RapidMiner Studio we will discuss all parts of the Design Perspective separately in the fol 27 2 Fir
98. s The Churn Analysis Wizard allows you to identify which cus tomers are most likely to churn and why The Sentiment Analysis Wizard analyses a social media stream and gives you an insight into customers thinking Tutorials Starts a tutorial window which shows several available tutori als from creating the first analysis process to data transformation Each tutorial can be used directly within RapidMiner Studio and gives an intro duction to some data mining concepts using a selection of analysis processes 2 2 Perspectives and Views 498k S AA ab hma Perspectives 4 2 A Snom Views Actions New Process Q Dayar E Rapilliner v6 0 Makes Predictive Analytics More Accessible to AN Tne primary fous in Rapichiner ie te addition ot appiicaten waar a new class sttochaa dsipodla mica tie PINAR U PENSEES BEANA SOARS IO WATAN particularly business managers and business analysts Repidbtiner A Potential Game Changer Dana Nems Peace Lana tous Rasare pubian an ovanan t RapidMiner vS on Monday November 18 2073 Init ne called Rapiatiiner Paat gone changer Open crenene hasimproed fet pr is dam forthe veyed Open previous analysis fra jat eisted atthe end of tie last century but as qood as these hot Ter hate eledi a mazas ar Uen beat MAG and intenna 5 on veran EH Application Wizard Data mining in three simple steps oer ona usmess Nat REPA AO NaS Varde isiat nua ctu
99. s are connected these process components are called operators An operator is defined by several things e The description of the expected inputs e The description of the supplied outputs e The action performed by the operator on the inputs which ultimately leads to the supply of the outputs e A number of parameters which can control the action performed The inputs and outputs of operators are generated or consumed via ports A port expects a specific type of input We will see that an operator in RapidMiner 32 2 3 Design Perspective Studio is represented by a module in the following form where input ports are placed on the left side and output ports are placed on the right side Apply Model Figure 2 13 An operator can be connected via its input ports left and output ports right Such an operator can for example import data from the repository a database or from files In this case it would have no input ports although it would have a parameter at least specifying the location of the data Other operators transform their inputs and return an object of the same type Operators that transform data belong in this group And other operators still consume their input and transform it into a completely new object many data mining methods come under this category and supply a model for the given input data for example The colour of the ports indicates the input type a port must be supplied with For exam
100. s the meta data look like in practice i e in RapidMiner Studio In RapidMiner Studio the meta data are provided at the ports Just go over the output port of the recently inserted operator with the cursor and see in Figure 3 5 what happens A tooltip will appear which describes the expected output of the port First the name of the operator and of the port followed by the kind of meta data In this case we are dealing with the meta data of an example set The number of the examples can also be inferred 100 as well as the number of the attributes 8 Then there comes a description of the path the object would have taken through the process during an execution In this case the path has only one station i e the port of the generating operator However the most important part of the meta data at least for an example set is the table which describes the meta data of the individual attributes The individual columns are Role The role of the attribute If nothing is indicated then it is a regular at tribute Name The name of the attribute Type The value type of the attribute Range The value range of the attribute so the minimum and maximum in the case of numerical attributes and an excerpt of possible values in the case of nominal attributes Missings The number of examples where the value of this attribute is unknown Comment A comment depending on the attribute Tip There are such tooltips of a higher complexity in several places
101. se from the Process View In order to do this click on a port where data is present using the right hand mouse button This is the case at the output ports of all operators that have already been executed You will recognise this from the darker colour and an appropriate entry in the context help You select the menu entry Store in Repository here to store the object Please bear in mind however that the data at the ports may be released again after some time in order to save memory and is therefore not guaranteed to stay at the ports for any amount of time Please also see the explanations in the previous chapter 5 2 3 Access to and Administration of the Repository Once you have imported your data into the repository you can use it in your processes with the retrieve operator You can drag the operator from the Oper ators View into the process as usual and define the parameter for the repository entry there But it gets easier still Simply drag an entry in the repository e g a data set onto the Process View using the mouse A configured operator with a reference to this entry will now be inserted automatically here If the entry is an object a new operator of the Retrieve type will be created and configured accordingly If the repository entry is a process however then a new operator of the Execute Process type will be created and its parameter will automatically refer to the selected process from the repository You wil
102. ser 5 5 Sweor Customer 30059 Cloting 06012008 5 es215 328077 B Labor iepotstons ble window area Dustomer 7t 55848 Sports 03 06 2006 3 56 475 169 424 J Market Data 7 7 siore 10 Customer7 11762 Hem 20092006 3 2687380619 es 8 Sowo Customer ti 75857 Hea 29112008 7 er07s 480522 muncea 9 s Swen Customer 97291 Heath 21032006 3 aas wr Peers 1 1 Sore 14 Customers 39580 Tos 05092005 6 aw 341382 b aene g Sorena Customer S8638 Heal 05062008 5 na 40920 heen 12 Store 12 Customer e4853 Spots 00122007 8 29309234470 n B sieve 08 Customer 20805 Electonics 25052008 6 wm maz u Store 06 Customer 1 19424 Home Gard 11 052007 4 T128 284513 r be _ i 15 store Ot Customer 90837 HomelGardk 28042008 7 46390 324730 E 07 Custenng 6 6 Store 09 Customer 65759 Tos 25022005 1 um uau a 08_Oner i Store 08 Customer 11 49125 Elecronics 01 052008 4 sors 202100 O i w 8 Store 06 Customers 68351 Modes 27122008 3 sra 173487 E Man Reposton 1 0 Sorena Customer t 98903 Books 18062005 6 sase 321105 2 2 tore 09 Customer 71 48122 Books 29012007 8 saw aan 2 a swore 14 Customer 21 94524 Moves 25042008 7 esris asooo 2 2 Simoa Customers 89794 Tos 02082007 5 sasn aosa a a ore 08 Customers 25809 Electonics 15072005 1 oren 97682 2 u Sowo Customer tt 59553 Tos 07122007 3 arses 262666 25 2 Soreoa Customer7 9389 Heal 27042007 1 maz 72002 2 2 Store 14 Customer a 21470 Books 14022007 7 nsn 43575 u Bi System Monitor Figure 4 2 Each open result is displayed as an
103. so an example of a customer who participated in our study There is therefore a value available for our target attribute prototype positively received Doe Systems was happy and has yes as an attribute value here thus we also speak of a positive example Logically there are also negative examples and examples which do not allow us to make any statement about the target attribute 1 2 3 Attribute Roles We have now already become acquainted with two different kinds of attributes i e those which simply describe the examples and those which identify the examples 10 1 2 Fundamental Terms separately Attributes can thus adopt different roles We have already introduced the role label for attributes which identify the examples in any way and which must be predicted for new examples that are not yet characterised in such a manner In our scenario described above the label still describes if present the characteristic of whether the prototype was received positively Likewise there are for example roles the associated attribute of which serves for clearly identifying the example concerned In this case the attribute adopts the role of an identifier and is called ID for short You will find such attributes iden tified with this role in the RapidMiner software also In our customer scenario the attribute name could adopt the role of such an identifier There are even more roles such as those with an attribute t
104. some time you will want to switch frequently between the Design Per spective and the Result Perspective Instead of using the icon or the menu entries you can also use keyboard commands F8 to switch to the Design Perspective and F9 to switch to the Result Perspective 3 4 2 Breakpoints Meta data transformation is a very powerful tool for supporting the design of analysis processes and making it much more comfortable There is simply no longer the necessity to perform the process more often than needed for test pur poses during design In fact the expected result can already be estimated on the 70 3 4 Executing Processes basis of the meta data Thus meta data transformation and propagation ought to revolutionise the world of data analysis a little instead of having to perform each step separately as before in order to configure the next operator the results of several transformations can now be foreseen directly without any execution This is of course an enormous breakthrough in particular for the analysis of large data sets Nevertheless the necessity arises in some cases of going beyond the meta data and seeing a specific result in all its details When the design is running it is usually no problem to place the desired intermediate result at a result port of the process and to execute the process very simply The desired results are then shown in the Result Perspective But what can you do if the process has already finished b
105. st steps lowing and discuss the fundamental functionalities of the associated views A 4niggen expert parameters Oren The root operator which is the cuter most operator of every process Description Each process must contain exactly one operator of this class and it must be tne root operator of the process This operator provides a set of parameters e Figure 2 9 Design Perspective of RapidMiner 2 3 1 Operators and Repositories View There are two very meaningful views in this area at least in the standard setting which are described in the following Operators View All work steps operators available in RapidMiner Studio are presented in groups here and can therefore be included in the current process You can navigate within the groups in a simple manner and browse in the operators provided to your heart s desire If RapidMiner Studio has been extended with one of the available extensions then the additional operators can also be found here 28 2 3 Design Perspective Ele Edt Process Tools ew Help GTGkhs oa OPIS A Sis 2B ot Process Ef Paametrs J gt ER a e t Broess e OHFES Tmava Ei Process 1ogvervosiy wat 7 atte Daat E Modeling 110 resulte a her S evatuaten 29 SSS poaa 2004 Operators View iene i roang asa LE Repositories agae DB IH Local Repository c tain Repository Repos
106. stomers are for example geographical data and information as to which products or services the customer has already purchased You would be surprised what forecasts can be made even from such a small amount of attributes Let us look at an admittedly somewhat contrived example Let us assume that you work in an enterprise that would like to offer its customers products in future which are better tailored to their needs Within a customer study of only 100 of your customers some needs became clear which 62 of these 100 customers share all the same Your research and development department got straight to work and developed a new product within the shortest time which would satisfy these new needs better Most of the 62 customers with the relevant needs profile are impressed by the prototype in any case although most of the remaining participants of the study only show a small interest as expected Still a total of 54 of the 100 customers in the study said that they found the new product useful The prototype is therefore evaluated as successful and goes into production now only the question remains as to how from your existing customers or even from other potential customers you are going to pick out exactly the customers with whom the subsequent marketing and sales efforts promise the greatest success You would therefore like to optimise your efficiency in this area which means in 1 Fundamental Terms particular ruling out such ef
107. t by Column Descending Sorting by this column in descending order Add to Sorting Columns Ascending Adding to the sorting columns as cending Add to Sorting Columns Descending Adding to the sorting columns de scending Sort Columns by Names Reordering columns by sorting the column head ings in alphabetical order Restore Column Order Restoring the original column order 85 4 Data and Result Visualization a Result Overview ig ExampleSet Generate Sales Data ExampleSet 100 examples 1 special attribute 7 regular attributes Filter 100 eri Row No transaction_id store_id customer_id product_id product_category date Data 64 64 Store01 Customer 1t 61867 Books 30 12 2007 gt gt 78 78 Store 01 44627 Rooks 05 09 2008 5 5 Store 01 Custome r Copy sts Statistics EH Select Row 65 65 Store 01 Custome H Select Column 15 15 Store 01 Custome E Fit Column Width Charts 1 1 Store01 Custome E3 Fit all Column Widths f 55 55 Store 01 Custome Ei Equal Column Widths l 79 79 Store02 Custome ff Sort by Column Ascending Advanced 56 56 Store 02 Custome 3 Sort by Column Descending Charts 53 53 Store 02 Custome Ti Add to Sorting Columns Ascending kS 92 92 Store 02 pool Add to Sorting Columns Descending l 24 24 Smad Cume EA Sort Columns by Names Annotation HA Restore Column Order 72 72 Store 02 Custome gp Ha Move column gt 25 25 Store 03 Customer 7 94389
108. ta at the design stage in such a way that the form of the re sult can already be foreseen and solutions can be identified in case of unsuitable operator combinations quick fixes In addition RapidMiner Studio offers the possibility of defining breakpoints and of therefore inspecting virtually every in termediate result Successful operator combinations can be pooled into building blocks and are therefore available again in later processes RapidMiner Studio contains more than 1500 operations altogether for all tasks of professional data analysis from data partitioning to market based analysis to attribute generation it includes all the tools you need to make your data work for you But also methods of text mining web mining the automatic sentiment analysis from Internet discussion forums sentiment analysis opinion mining as well as the time series analysis and prediction are available RapidMiner 19 2 First steps Studio enables us to use strong visualisations like 3 D graphs scatter matrices and self organizing maps It allows you to turn your data into fully customizable exportable charts with support for zooming panning and rescaling for maximum visual impact 2 1 Installation and First Repository Before we can work with RapidMiner Studio you of course need to download and install the software first You will find it in the download area of the RapidMiner website http www rapidminer com Download the ap
109. te Data E Correlation Matric Correlation Matra Repositories ine lnc ace a ee ae ae cee a OTA att 1 0 009 0 157 0 056 0 099 0 052 0 028 0 200 0 038 0119 0 025 0 134 0 022 0 083 0 026 samples T at 2056 0 032 0 080 i 0 014 0 179 9436 0015 0 019 0 073 0 083 0414 0108 0114 0 06 8 Deals Testset none el oeo or am Be oe a e cme Bao Parwse atg oos2 w109 oos om9 ooz F 021 0027 ooe oo23 ooas 0153 0122 0201 0 06 IB ColtTestset nore siis a7 0 028 01630167 01350078 0021 4 0164 oaos 0003 0100 0023 0 080 0 045 00 B tis rone atia 0200 0 132 0 038 0 015 0 251 0 027 0164 4 0 186 0 018 0 102 0 011 0 056 0072 0 04 E Lanor Negotiations nsn j mm om oo ow jam om owm oon vom oos omo ooo H oar owa oni t ieee w e e Sa em a on e a ianea atts 0 083 0 000 0 000 0114 0 075 0201 0 045 0072 0 043 0 070 0 266 0o44 0 108 4 0 00 E G processes inore at15 0 024 0 036 0 042 0 007 0 028 0068 0018 0 041 0159 0 153 0 015 0 219 0 058 0 009 KI QQ 01_Leamer inon atii 0 198 0 187 0 029 0 131 0 035 0 021 0 080 0 063 0 087 0 138 0 042 0 159 0 115 0 088 0 00 A 02_Preprocessing inon att 0 248 0 100 0 134 0 065 0 083 0 223 0 062 0 025 0 015 0 094 0 045 0 400 0217 0 057 0 18 03_Validation ams aoa aa oos oms ooe oza oos 183 ons oom oors oos ka oan ati24 0 082 0 021 0 117 0 045 0 048 0 014 0073 0 077 0 044 0 002 0 065 0 128 0 033 0 064 0 17 8 amp Main Repository chvistian Pat ataa 032 0408 odat 0407 ooes oom7 oorge oomo loose oos aose 0120 ao
110. the mouse while holding down the button All selected operators will now be moved to a new place depending on where you move the mouse If in the course of this movement you reach the edge of the white area then this will be automatically enlarged accordingly If you should reach the edge of the visible area then this will also be moved along automatically Copying Operators Select one or more operators as described above Now press Ctrl C to copy the selected operators and press Ctrl V to paste them All selected operators will now be placed to a new place next to the original operators where you can move them further Deleting Operators Select one or more operators as described above You can now delete the selected operators by e Pressing the DELETE key e Selecting the action Delete in the context menu of one of the selected operators e By means of the menu entry Edit Delete Deleting Connections Connections can be deleted by clicking on one of the two ports while pressing the ALT key at the same time Alternatively you can also delete a connection via 39 2 First steps the context menu of the ports concerned Navigating Within the Process If we look at the toolbar of the Process View then we can see that we have only made use of one action so far In this section we will discuss the following four elements on the left side of the toolbar the arrow pointing left the arrow po
111. the three values overcast rain and sunny The attribute Temperature on the other hand is numerical and takes values ranging from 64 to 85 given in Fahrenheit of course Finally the attribute Play is nominal again but still has a special role It is marked as label The role is in italics and is given before the attribute name Golf Data Table Number of examples 14 5 attributes Role Name Type Range Missings Comment Outlook nominal overcas Temperat integer 64 35 Humidity integer 65 96 Wind nominal ffalse tr Play nominal no yes Figure 5 6 The meta data of the golf data set from the example directory of the Sample repository provided with RapidMiner Studio You will find the data set named Golf in the data directory in this repository 107 5 Repository 5 3 1 Propagating Meta Data from the Repository and through the Process You have already seen that the meta data described above accompanies the actual data on its way through the RapidMiner Studio process when you create the process As previously mentioned it is however absolutely necessary for this meta data propagation and transformation that you are able to manage the data in a RapidMiner Studio repository and obtain the meta data from this For this reason we would like to remind you of and underline the necessity of using the repository for data and process management in order to pro
112. the upper output port of the normalise operator In both cases you will see the meta data of the Iris data set You will notice however that the meta data of the selected attribute have changed The range of values of a3 is now normalised to the interval 0 1 after the transformation Or to be more precise The range of values of a3 would in the case of execution be normalised 108 5 3 Data and Meta Data to the interval 0 1 Insert a further operator the operator Discretize by Frequency Connect this with the normalise operator Set the parameter range name type to short and this time select another attribute for example a2 with the same mechanism as above Now go over the output port of the new operator with the mouse and observe how the meta data has changed The selected attribute is now no longer numerical but nominal and takes the values rangel and range2 The discretization operator breaks the numerical range of values apart at a threshold value and replaces values below this value with rangel and values above this value with range2 The threshold value is automatically chosen so that there is the same number of values above as below If you wish to have the values divided up into more than two ranges of values adjust the parameter number of bins accordingly You can see the process and the indicated meta data in the following picture P
113. tion i e quick fixes Back to our example Trained analysts will recognise straight away that the data must be so called transaction data where each transaction represents a purchase We gave the following attributes for our example set e transaction_id Indicates a clear ID for the respective transactions e store_id Indicates the store where the transaction was made e customer_id Indicates the customer with whom the transaction was made e product_id Indicates the ID of the product bought e product_category Indicates the category of the product bought e date Indicates the transaction date e amount Indicates the number of objects bought e single_price Indicates the price of an individual object If we look at the last two attributes first then we see that whilst the number and the individual price of the objects are given within the transaction the associated total turnover however is not Next we therefore want to generate a new attribute with the name total_price the values of which correspond to the product from amount and single price For this we will use a further operator named Generate Attributes which is located in the group Data Transformation Attribute Set Reduction and Transformation Generation Drag the operator behind the first operator and connect the output port of the data generator with the input port of the new operator and connect the output port of the latter with
114. ton wizards to star inducing i Learn the basics of RapidMiner ess Direct marketing Seqiment analysis Preaictye mantenance 5 Leg near Nov 22 2043 1 20 17 PM SEVERE Cannot execute Validating Process Figure 2 6 Dragging the lower Log View to the middle and highlighting the new position You can combine individual views this way to create several file cards mean ing that only one is ever visible Or you can drag the Log View from below to the right hand area so that the division now runs vertically and no longer horizontally You can even undock a view completely and move it outside the RapidMiner Studio window If you would like to see a view in full for a short time then you can maximize a view and minimize it again later on This is also done if you right click on the name area of a view and select the maximize action Each view offers you the actions Close Maximize Minimize and Detach like it is displayed in Figure 2 7 25 2 First steps Welcome Figure 2 7 Actions for views Those actions are possible for all RapidMiner Studio views among others The other actions should be self explanatory 1 Close Closes the view in the current perspective You can re open the view in the current or another perspective via the menu View Show View 2 Maximize Maximizes the view in the current perspective 3 Minimize Minimizes the view in the current perspective The view is
115. u will find on the left hand side of the Process View We have never connected these before This is not even worthwhile in the basic setting at least not for the sources because the process itself then has no input Connecting the inner sinks does have an effect however All objects which arrive at a sink at the end of the process are presented in the Result Perspective as the result of the process 104 5 2 Using the Repository These input and output ports of the process have a further function however A typical process begins with a set of retrieve operators which are followed by a set of processing operators and ends with a set of store operators You can avoid having to create these operators by using the Context View which you will find in the View menu Figure 5 5 shows this Context View g Process XML wy B Gierocess gt Hted Normalize Discretize inp D dea exa dres p H oi res ore D Q Context Process input R R Name Location input 1 Normalize example set input Samplesidatailris I g Process output R R Name Location result 1 Discretize example set output result 2 disconnected amp Macros ae Macro Value Figure 5 5 The process context For Input you indicate repository entries which are to serve as an input of the process and be placed at input ports of the process For Output you indicate wh
116. uble clicking or by pressing the button in the centre with an arrow pointing to the right Select Attributes attributes a d 7 x Select Attributes attributes qQ The attribute which should be chosen Attributes Selected Attributes Le Lele amount product_category customer_id store_id date total_price product_id single_price transaction_id Figure 3 12 Individual attributes or subsets can be selected or even deleted with the operator Select Attributes Did you notice The new attribute total price which had so far only been computed within the meta data transformation was already ready for you to select here without you ever having executed the process If you examine the meta data at the output port again you will see that only the three selected attributes are left plus the transaction ID which also has a special role that of ID and was therefore not affected by the selection Since we would like to remove this ID too select the option include special attributes in the parameters of the operator Select Attributes and examine the meta data again Now only the three desired attributes are left You can find out the effects of these and all other parameters in the description of parameters in the Help View and also in the operator reference Tip It is a basic rule of RapidMiner Studio that operators from the group Data 67 3 Design of Analysis Processes
117. ubsector Number of employees Number of purchases in product group 1 e Number of purchases in product group 2 If we assume you have thousands of customers in total then you can only indicate whether 100 of these evaluated the prototype positively or not You do not yet know what the others think but you would like to The attribute prototype positively received thus adopts a special role since it identifies every one of your customers in relation to the current question We therefore also call this special attribute label since it sticks to your customers and identifies them like a brand label on a shirt or even a note on a pinboard You will also find attributes which adopt this special role in RapidMiner under the name label The goal of our efforts is to fill out this particular attribute for the total quantity of all customers We will therefore also often speak of target attribute in this book instead of the term label You will also frequently discover the term goal variable in the literature which means the same thing 1 2 2 Concepts and Examples The structuring of your customers characteristics by attributes introduced above already helps us to tackle the problem a bit more analytically In this way we ensured that every one of your customers is represented in the same way In a certain sense we defined the type or concept customer which differs consider ably from other concepts such as falling a
118. urther display configurations e g whether the text at the x axis is to be rotated so that long texts can still be read Play around a little with the settings and various possibilities and you will soon be familiar with the numerous possibilities for visualisation Tip You can change the colours used in the settings under Tools Prefer ences Changing the Chart Type The selection of the chart type significantly defines which parameters you can set In Figure 4 11 you can see an example of a Bars Stacked type chart Instead of the different axes you now set attributes according to which the data is to be grouped here store_id and which attribute is to be used for defining the stacks here product_category The height of the bars then corresponds to the sum here Aggregation is on sum of the attribute defined as value column here amount Computing Visualisations Last but not least it is to be mentioned here that there are other visualisations which are so complex that they must be computed especially These visualisa tions such as a Self Organizing Map SOM then offer a button named Calcu late with which the computation and visualisation shown in Fig 4 12 can be started 88 4 3 Display Formats amp Result Overview B ExampleSet Generate Sales Data R Movies W Books W Clothing B Sports B Health W Electronics W Home Garden Hi a B Bars Stacke
119. ve a Concept customer available which we will describe with a set of Attributes e Prototype positively received Label Nominal e Name Text e Address Text e Sector Nominal e Subsector Nominal e Number of employees Nominal e Number of purchases in product group 1 Numerical e Number of purchases in product group 2 Numerical The attribute Prototype positively received has a special Role among the attributes it is our Target Attribute here The target attribute has the Value Type Nominal which means that only relatively few characteristics in this case yes and no can be accepted Strictly speaking it is even binominal since only two different characteristics are permitted The remaining attributes all have no special role i e they are regular and have either the value type Numerical or Text The following definition is very important since it plays a crucial role in a successful professional data analysis This volume of information which describes a concept is also called meta data since it represents data via the actual data Our fictitious enterprise has a number of Examples for our concept customer i e the information which the enterprise has stored for the individual attributes in its customer database The goal is now to generate a prediction instruction from the examples for which information is available concerning the target attribute which predicts for us whether the remain
120. vide support during process design In this section we will carry out a further example for the design of a process only this time we will revert to a data set from the RapidMiner Sudio repository We will therefore now carry out the complete process for the first time from the retrieval of the data right through to the creation of the results Of course this process would typically be preceded by importing data into the repository using one of the methods presented above but in this case we shall skip this step and simply use one of the data sets already provided by RapidMiner Studio instead Load for example the provided data set Iris using a retrieve operator by simply dragging the entry concerned in the same directory as the golf data set already used above into the Process View Do not execute the process yet though Insert a normalise operator and connect its input with the output of the retrieve opera tor Set the parameter method to range transformation In this setting the operator serves for re scaling numerical values so that the minimum is currently 0 and the maximum is currently 1 Select an individual attribute which you wish to apply this transformation to e g the attribute a3 For this purpose set the filter type attribute filter type to single and select the attribute a3 at the parameter attribute Now go over the output port of retrieve first with the mouse and then over
121. visualisations such as SOMs offer a Calculate button for starting the computation The progress is indicated by a bar amp Result Overview Tree Decision Tree Zoom 2 P Mode Age 4 Description gt 31 500 21 800 Ta gt ve e Tree S Gender Payment Method Annotation Node Labels stone ssa Soult adan condib cal Edge Labels we 4 4 no yes Payment Method Gender Gender Save Image mm aes a cash cheque credit card fema male femai male p K y y FA hoe cy no yes no yes gt 264s 35 500 gt 2645 35 500 gt 72ts 72 500 E l y 4 b X no yes no yes no yes i Figure 4 13 A decision tree in a graph view 90 4 3 Display Formats Mode Two fundamental navigation methods are available in the graph which are also called modes 1 Moving The mode for moving the graph is selected by pressing the left hand button in the mode box In this case you can move the section of the graph while holding down the left mouse button in order to view different areas of the graph in detail 2 Selecting The mode for selecting individual nodes is selected by pressing the right hand button in the mode box Now you can select individual nodes by clicking on them or while holding down the mouse button define a selection box in a free area for several nodes at the same time By holding down the Shift key you can add individual nodes to the selection or ex
122. w this approach from other data analysis tools You add an operator execute it and indicate the results via the context menu or via special operators for this Even if this approach seems intuitive and easy to use for small data sets we urge you to avoid this method since it will lead to problems no later than during the analysis of large data sets In this case a copy of the data would have to be held ready at each port so that this result can be provided later on RapidMiner Studio goes a completely different way here a way which also promises greater success in the long term The meta data is transformed and propagated by the process and data is only made available where absolutely necessary This kind of RapidMiner Studio analysis thus combines the interactivity allowed by established meta data with simple process definition for the analysis of data sets including large ones Note RapidMiner Studio has a sophisticated memory management here As al ready mentioned above results are kept at the ports for a while longer These results are deleted as soon as the memory of RapidMinder Studio or other pro 78 4 2 About Data Copies and Views grams necessary for this is required This means Results can disappear from the ports and then no longer be available for visualisation This is one of the reasons for the efficiency of RapidMiner Studio and for this reason we also recommend the automatic display via connected ports as describ

RapidMiner Studio Manual - RapidMiner Documentation

Contents

Download Pdf Manuals

Related Search

Related Contents