GO GIRLS GO Cyber : Data Miner - Chandan Reddy
Contents
1. Department of Computer Science Wayne State University Data Miner Contents Gaye oe Ee dE 3 ENEE 3 ADOUR SEN E LCE 3 FS el ON e EE 3 ProregUl US sosesc ge navsnonceceunsatocsinaescuvesguncuianasncusnayieeiaontuseseoniectinonsnceareiactinmcomeaniunceds 3 OM ys ces cca pac E basateleatarsveeruetetpescieetessuanntoes staveseaucraiaed cre A 3 aye e 4 Chapter 3 Recommender System Renee near ere nee er Oe eo eee eee 5 ele NEC GE 5 Bil Demon PaO E a E A 7 Pamer D ele EE 10 Pan FCS SIF CO BE 12 Dd BERNO E 12 BZ NCIS CN Ee 13 Ee ne RE WE 14 Chapter 6 Association Analysis EE 17 Gl aye 01 EE 19 Ye Siar e Ee ig EE 23 Bae Dnne S LEV EE 23 ER een ne E E 24 eB ROVE SOI EE 25 ei uge E 27 Data Miner Chapter 1 Getting Started Overview As we are in the information age we believe that information leads to power and success Today we have far more information than we can handle which leads to find new ways to help us make better managerial choices Data mining a powerful computing technology plays a major role in this situation which is an analysis of data to extract useful patterns or information from a large data collection It is important to educate the youth generation about these novel technologies and directs them for higher education in these novel areas As Data Miner Educational Tool was built for middle school children its primary objective is to teach them the Data mining concepts in a simplified2. 76 Classification Advanced Level Impure Classification 16 Data Miner Chapter 6 Association Analysis This phase describes Association Analysis using Market Basket Transactions which is useful discovering interesting relationships in a set of transactions and talks about the Support Count which means the number of items present in a transaction Similar to the other data mining techniques Association Analysis phase also includes a demonstration part and interactive part CLICK THE ITEM TO ADD TO REMOVE FROM YOUR CART ZNMRe OD ge TRANSACTIONS OF WEEKDAYS YOUR CART e Amy gt A Wi DE l a gt E Sai W I _ ee Bob Emma CART Is EMPTY OR NO COUPONS AVAILABLE FOR THE ITEMS IN CART Figure 87 Association Analysis Interactive Part Demonstration part explains how to work with interactive part and the idea behind that In interactive part the system provides a set of items available in a shop and user has to add them to his cart In addition to that it provides a set of transactions occurred during five weekdays of a specific week such that each items is included at least in one of the given transactions User can add an item to the cart by clicking on it and remove the same item from the cart by clicking on it again When user adds an item to the cart all the transactions which contains the item set in the cart will be highlighted Figure 18 Depending on the associated items common items in these transactions w
3. manner and their real world application hence motivate them to follow their higher education in Computer Science About the User Manual The user manual of Data Miner covers all the details of the tool including data mining concepts implemented Installation Guide Prerequisites Adobe Flash Player Web browser Set Up Click on the link provided Follow the instructions provided in this manual for each module Data Miner Chapter 2 Usage Data Miner describes six main data mining concepts in separate modules Recommender System Cluster Analysis Classification Association Analysis PageRank Social Networks Hi I m Mynee your Data Mining assistant I m here to guide you to work with this application If you need any help count on me First move your mouse over these _ buttons to see what you can do Figure 1 Home Page Each module consists of a welcome page that includes a brief description of the concept demonstration and interactive part In demonstration it explains how to plays at the interactive part In addition to that some phases include quizzes to evaluate the user Data Miner Chapter 3 Recommender System Data Miner demonstrates Recommender System by movie preferences and explains two approaches such that recommendations based on the contents of the movies and recommendations based on the other users preferences 3 1 Interactive Part In the first approach user will
4. A HETFLIX k f SS d Chloe Miller was at Olive Garden Chloe Miller shared Netflix s photo Now streaming Teen Wolf Season 2 Groups GO GIRLS GO Universal Music Shelfari Book Club Frege gota Chloe Miller likes Who would be the Facebook suggestion for Lucy EE Figure 17 Social Networks Advanced Level Initially the Facebook Home pages are not visible but they will pop up when mouse over event occurs on each button provided as multiple choices 25 Data Miner Based on following Facebook home pages contents of given three Facebook users suggest a friend for Chloe Fi Type to search fer people places and things Lucas Home amp a Chloe MillerLikes iCarly s photo Chloe Miller was at Olive Garden i PS Chloe Miller shared Netflix s photo Now streaming Teen Wolf Season 2 Groups GO GIRLS GO Universal Music Shelfari Book Club Chess Magazine Likes Lucas Anthony likes Pepsi Lucas Anthony likes Angry Birds NETFLIX vie Traveled to Colorado Springs Colorado so Lucas Anthony likes Kickin it s photo Groups Universal Music Shelfari Book Club Chloe Miller Who would be the Facebook suggestion for Lucy Figure 29 Social Networks Advanced Level Mouse Over on Buttons 26 Contact Us Dr CHANDAN REDDY Assistant Professor Department of Computer Science Wayne State University Detroit MI 48202 Contact information Tel 313 577 9005 Fax 313 577 6868 Email reddy
5. Basic idea behind Classification is assigning given items to a known class which have items with similar attributes In Data Miner Classification is explained by building decision trees by dividing set of items by values of a specific attribute in each level and selecting the best split such that each classification is pure i e no branch contains items which belong to different classes This phase contains a demonstration of classification using decision tree as an animation clip which shows how the interactive part works Similar to the Cluster Analysis the interactive part of Classification contains three levels such as Beginner s level Intermediate level and Advanced level Each level will gradually increase the difficulty of classifying given objects by increasing the number of attributes classes and levels involve with the tree The objects given to be classified contain a attributes and attribute values which are displayed for Mouse Over event To make the tool less complicated the number of branches and nodes containers for each branch of the tree are given to the users so they can drag and drop items to these containers just like the objects behave in Clustering phase Attributes and Attribute values Class Figure 3 Attributes Attribute Values and Class label 5 1 Beginner Beginner s level consists of ten objects animals which belong to either one of two classes called Mammals and Fish IW provides two attributes Giv
6. be provided two sets of movies such that each set contains ten movies Any movie belongs to a single genre and the system considers only five genres Adventure Animation Comedy Family and Fantasy This approach recommends items that are similar to those that a user liked in the past User can LIKE movies and depending on the LIKEd movies system will recommend some different movies to the user THUMBS UP for your Favourite Movies CA d TINTIN Ds Unlike Figure 2 Recommender System Interactive Part 1 Inputs Data Miner User can hit on Like button to like the movie and it becomes an Unlike button He can click on the same Unlike button to undo the action i e Unlike the liked movie Once user finished his selection he can move to the next step to see the recommendation by clicking on Next button To make the Next button work user has to like at least a single movie Next page shows the Recommendation button and when the user clicks on it it will lead to the recommendation movies for the user s preference Figure 3 Recommender System Interactive Part 1 Output The second approach contains three quizzes and it teaches how to recommend movies based on friends preferences The Next button and Back button will not work until the correct answer is given Data Miner Let s Find the Best Matching Friend Preference Viewer Let s find the best matching friend for Nancy based on the movie preferences of other f
7. clicked this position may changed and user has to guess whether the position goes up down or does not change 21 Data Miner Google music Chopping News More Search tools Music YouTube hittpyiwww youtube com Auto generated by YouTube SubscribeSubscribed RAY DALTON OFFICIAL MUSICVIDEO NPR Music New Music Music Reviews and Music News NPR htipuiwww npe orgmusic NPR Music features streams live concerts and music news Billboard Music Charts Music News Artist Photo Gallery amp Free htipuiwww billboard com Daily music news charts music downloads and artist features for rock pap country rhythm and blues jazz world and hip hop Grooveshark Free Music Streaming Online Music hitps igrooveshark com Grooveshark provides free music streaming online radio stations and lets you connect with artists and friends Stereomood tum your mood into music free playlist for every hitpy www stereomood com Stereomood is the music streaming service that turns your mood into music free STEREOMDOD mag Go Music hittpy www go music com It s all about music Find your favorite artist new albums released esults will 90 up go down not change Figure 14 Google Search Results 22 Data Miner Chapter 8 Social Networks Data Miner explains the idea of Social Networks using Facebook as it is one of the most popular social network site in these days The main purpose of Social Networks module
8. cs wayne edu Data Miner 27
9. d Mammals and birds It provides three attributes Give Birth Live in Water and Can Fly and each attribute can have either one of two values Yes or No The Intermediate level also concerns only a single level of the tree The objective of this level is to teach building a decision tree with a single level from the given set of items and the difference between purity and impurity of the classification 13 Data Miner Following is the table which contains animals listed in this level Live in Water Platypus Mammals Human Mammals Cat Mammals Porcupine Mammals Pigeon Birds Owl Birds Eagle Birds bat Mammals Whale Mammals Dolphin Mammals Table 2 Classification Intermediate Level Objects Intermediate level provides three attributes to split the root and it can build three different trees Even though neither split will make a pure classification each classification creates one branch pure Branch NO is pure because it contains only MAMMALS Figure 13 Classification Intermediate Level 5 3 Advanced Level Advanced level also consists of ten animals which belong to either one of three classes called Mammals Fish and birds It provides three attributes Give Birth Live in Water and Can Fly and each attribute can have either one of two values Yes or No unlike other two difficulty levels the advanced level concerns 14 Data Miner two level of the tree to be split T
10. e Figure 6 Demonstration Step 2 STEP 3 Click on Step 3 button to play the animation Once it is completed hit Next button to go to the second movie set of the collaborative filtering When the last phase of collaborative filtering is completed the Recommender System reaches to its end and clicking on Next button directs to the Home Page while clicking on Back button directs to the initial page of Recommender System Data Miner Hope you enjoyed learning Recommender system Would you like to review it again To learn more about Data Mining Figure 7 Recommender System Last Page Note each data mining concept ends with a page like this with similar behavior Data Miner Chapter 4 Cluster Analysis Objective of Cluster Analysis is teaching how to group a set of objects based on their attributes such that object with similar attributes belong to the same group while dissimilar object belong to different groups It contains a demonstration as an animation clip which shows how the interactive part works The interactive part of Cluster Analysis contains three levels such as Beginner s level Intermediate level and Advanced level Each level will gradually increase the difficulty of grouping given objects The objects given to be clustered contain a list of attributes which are displayed for Mouse Over event Since this module simulates the Cluster Analysis clusters are predefined and the number of containers indicates the
11. e Birth and Live in Water to split the root node As each attribute can have either of two values Yes and No for the selection of each attribute two branches will be popped up Beginner s level concerns only a single level of the tree which means it only allows to split the tree once The objective of this level is to teach building a decision tree with a single level from the given set of items Following is the table which contains animals listed in this level Give Birth Live in Water Class Platypus Mammals Mammals Cat Mammals Porcupine Mammals Bat Mammals Eel d Leopard Shark Whale j 8 Dolphin Mammals Table 1 Classification Beginner s Level Objects Fishes Fishes Salmon Fishes Mammals 12 Data Miner Dolphin By ke 4 o y f s Bat Lk E C i Figure 4 Classification Beginner s Level For an instance when user selects the attribute Give Birth the root not is split into two branches with attribute values Yes and No User can drag and drop items from root node and if the animal is dropped into the right branch it will stay with that node Otherwise it returns back to the initial position in root node As this level provides two attributes to split the root it can build two different trees But neither split will make a pure classification 5 2 Intermediate Level Intermediate level consists of ten animals which belong to either one of two classes calle
12. he objective of this level is to teach building a decision tree from the given set of items with two levels until it build pure branches Following is the table which contains animals listed in Advanced level Give Birth Can Fly Live in Water Class Pigeon No Birds Owl No Birds Eagle No Birds bat Yes Mammals Human Yes Mammals Cat Yes Mammals Porcupine Mammals Python Komodo Reptiles Dolphin Mammals Table 3 Classification Advanced Level Objects Reptiles Intermediate level provides three attributes to build the tree into two levels hence it can build six different trees Unlike in other two levels some splits will make a pure classification Dolphin Branch YES is PURE because it contains only MAMMALS Click Next button to split the IMPURE branch Figure 54 Classification Advanced Level 15 Data Miner Advanced Level By GIVE BIRTH You did it Each non empty branch is pure which means each contains animals from only one class Try other attributes If you ve done go Back and split the root by a different attribute GC Ce Congratulations It s a perfect classification Figure 65 Classification Advanced Level Pure Classification Advanced Level By GIVE BIRTH Neither branch is pure because each non empty branch contains animals belong to different classes Try to split by a different attribute If you are done go Back Figure
13. hich are not yet added to the cart the system will display associated items about which item to be selected next in the area that associated items display with a coupon that offers some discounts depending on the items added to the cart 17 Data Miner CLICK THE ITEM TO ADD TO REMOVE FROM YOUR CART XTC OSH mw TRANSACTIONS OF WEEKDAYS YOUR CART Transaction Cart Highlighted Transactions Associated j C Items U A oupon Figure 98 Association Analysis Content of Interactive Part Once all the associated items are added to the cart the system will indicate that to the user Figure 19 CLICK THE ITEM TO ADD TO REMOVE FROM YOUR CART al les 8 CS Figure 19 Association Analysis Completion of adding associated Items 18 Data Miner Chapter 7 PageRank PageRank is a link analysis algorithm of the Google web search engine It assigns a numerical weighting to each element of a hyperlinked set of documents to measure relative importance of that element within the set Objective of PageRank phase in this tool is to describe what Google PageRank is and how it affects to a certain web page or web site The demonstration of PageRank phase consists of two parts The first part illustrates how to search over the Google search engine and what outcome will gain as results of search keywords The second part describes the incoming and outgoing link of a web page Interactive part consists of two section
14. is to describe certain facts that would help Facebook to connect people as suggestions With this tool we discussed how Facebook suggests friends to its users based on groups that they have joined pages they have liked and posts which have been published by them Social Networks module also contains a Demonstration and interactive part The Demonstration explains the idea of degrees of connection by using a network of friends in a social network Home a People you may know Lucy amp Kb Add as a Friend Ian Ki Add as a Friend Ryan a Add as a Friend Toby KI Add as a Friend Toby Ki K Add as a Friend L S E 2nd Degree Connections of Emma Figure 25 Social Network Demonstration The Interactive part consists of three levels such that Beginner s level Intermediate level and Advanced level In each level the tool provides some Facebook users with some of their information and asks to suggest a friend to the given user based on certain similarities exist between theses users 8 1 Beginner s Level The Beginner s level provides four groups of Facebook users such that each group of users are from either USA Germany India or China and a group contains three users Each user in these groups lives in different cities of their country The objective of this level is to suggest a friend to the given Facebook user by the system by picking one of these users If the answer is incorrect the tool will indicate that and if i
15. number of cluster to be created User has to click on an object drag and drop it into one of the cluster containers provided Even though cluster labels are unknown at the beginning as cluster objects of each container is predefined once user dropped an object to a container the object will stay with it if that object belongs to that cluster If it s not the right container the object will move to its initial position The assistant of the tool will provide a hint about the clusters to make the task easier for the user Cluster Containers Hint Man walks but snake crawls Objects Figure 8 Cluster Analysis Interactive Part Beginner s Level Note that objects should drop in to the middle of the container such that it does not overlap with edges of the container At the end of each level a quiz will be given to evaluate the user s understanding about attributes based on clustering The system indicates whether the submitted answer is right or wrong and does not allow moving to the next level until the right answer is submitted Next level can be reached by clicking on the Next button and Back button directs to the Difficulty Level Selection page 10 Data Miner Intermediate Level Hint Sonic is a hedgehog ez P at Figure 9 Cluster Analysis Intermediate Level Advanced Level Emu runs fast Eagle can fly fas Figure 10 Cluster Analysis Advanced Level 11 Data Miner Chapter 5 Classification
16. on a node to link it to Go Music and click again to unlink See what happens Figure 11 PageRank Interactive Part 2 As the size of the Go Music node represents its PageRank when a green node is linked to the Go Music it becomes larger and unlinking the same green node changes Go Music node size to previous state Linking Orange colored nodes do not change the size of Go Music a b c Figure 12 Page Rank Changes of Go Music Node depending on Incoming Links a Initial Status b Incoming Links fro less popular web sites c Incoming links from popular related web sites 20 Data Miner Above three figures illustrates how center node size changes In Figure 22 a shows the initial status of the Go Music node According to Figure 22 b linking Orange colored node does not affect to the GO music node size In Figure 22 c linking Green colored nodes changed the size of the center node At the end of these two sections user will be evaluated by a quiz The last interactive part of PageRank module describes the effect of incoming links from related web sites on Google search results First user has to click on Play button to see the initial Google search results page Go Music C Rhapsody inch results will g0 up go down not change Figure 13 PageRank Interactive Part 3 The initial search results show the position of Go Music web site which is the last When a node lies around the Go Music is
17. our Active User a answer and Submit DETE AE thers Alicia P KS DS S gt F gt gt mke KSSH PF gt P gt gt Windy e de S XD S PF Sver P XS Px oF Who is the best matching friend for Nancy Alicia Mike Windy Skyler Figure 4 Recommender System Quiz 1 3 2 Demonstration Recommender System includes a demonstration of Collaborative filtering which explains the underlying mechanism of finding the similar user To see how it works user has to click on Step 1 Step 2 and Step 3 buttons in each page Click on Step 1 button to play the animation Once it is completed hit Next button to see the next step How Recommender System Works Jaden Smith Lets see how Recommender System suggests you a friend First click Step 1 to see how you entered the name and preference for each movie Click Next go to Step 2 Figure 5 Demonstration Step 1 Data Miner STEP 2 Click on Step 2 button to play the animation Click on Next button in the demonstration part not the Circled Next button to calculate similarity score for each user When the step 2 demonstration is over click on Next button outside to move to the step 3 Preference Click Step 2 to see how Recommender System Suggest you a FRIEND based Gd nets goo stp 3 jeden Smith X X d gt gt x Taylor Swift P X P P X E NEE Ons McClain d d P es lelro p la l l sisl
18. s that demonstrate the same idea which is incoming links from popular relative web sites increases the PageRank more than the links coming from non relative less popular pages To explain this idea the system uses a web site called Go Music which contains news related to music such as latest music albums released popular artists songs etc User can link eight real web sites Amazon Danskin Billboard Pizza Hut Yahoo You Tube AOL and Crayola to Go Music web site and see how PageRank of Go Music varies Among these incoming links Amazon Billboard Yahoo You Tube and AOL are considered as more popular than the other three web sites and contain news related to music So links coming from these five web sites will show a higher increase in GO Music PageRank comparison to the links coming from other three web sites Danskin Pizza Hut and Crayola The first page of interactive part describes the effect of incoming links using a High Striker When a popular web page is linked to the Go Music web site lever of the High Striker will hit the bell If the incoming link is from a less popular page then the lever will rise high but will not hit the bell dd Teen Choice Awards Young people ages 13 to 19 can vote on their favourite actors fashionistas sports athletes and singers for the Teen Choice Awards This event airing on the FOX Network in the United States takes place every year at the Gibson Amphitheatre in California s Univer
19. sal City Teenagers vote online to determine the ie winners for each of the categories Winners receive genuine custom made surfboards each costing over 800 The surfboard represents the freedom teenagers experience during summer Zones me t i oh s New Music Releases Red is the fourth studio album by Taylor Swift which i was released on DIR SMIET October 22 2012 Take Me ID Home is a Fecord destined for commercia leas success Call Me Maybe is a song recorded by Ak Hy Canadian singer EN songwriter Carly Rae e Jepsen Call Me Maybe has attained commercial success worldwide reaching 1 In many countries inclusing Australia UK and US Never Say Never Related Links More Information Teen Choice Awards One Direction Demi Lovato Figure 10 PageRank Interactive Part 1 19 Data Miner In the second page of the interactive part Go Music is the center node and the user can link other nodes to it by clicking on them To unlink a linked node click on the same node Green node represents popular and relative web sites while Orange nodes represent the less popular web sites with non relative contents How Incoming Links affect to PageRank Green nodes represent very popular web sites They contain similar related information to Go Music web page Orange nodes represent web sites which are not much popular like green nodes and are not related to Go Music Click
20. t s correct then an explanation of the answer will be provided 23 Data Miner Beginner s Level Following Facebook users are from four different countries and twelve different cities that belong to those four countries LANSING FRANKFURT BERLIN MUNICH f HIROSHIMA MUMBAI NEW YORK o Who would be the Facebook suggestion for Daniel Figure 15 Social Networks Beginner s Level 8 2 Intermediate Level Intermediate level is similar to the beginner s level but it contains more details about each user than the beginner s level Intermediate Level Following Facebook users study in different institutes based on their interest and they are form different cities in USA DENVER LOS ANGELES DETROIT DEARBORN NEw YORK Los ANGELES DETROIT MUSIC SCHOOL NEw YORK DENVER DANCING ACADEMY Who would be the Facebook suggestion for Lucy Figure 16 Social Networks Intermediate Level 24 Data Miner 8 3 Advanced Level In advanced level the system provided four users with their Facebook Home pages and depending on the similarities of the content in these pages user has to select the person who can be suggested as a friend to the given Facebook user Based on following Facebook home pages contents of given three Facebook users suggest a friend for Chloe Type te search for people places and things Chloe Home oo ie a DES i S LN Mouse Over to see the Facebook page Likes Chloe Miller Likes iCar y s photo
Download Pdf Manuals
Related Search