Social Prediction: Can we predict users action and emotions?

Transcription

1 Social Prediction: Can we predict users action and emotions? Jie Tang Knowledge Engineering Group Department of Computer Science and Technology Tsinghua University 1

2 Motivation Social behavior VS. Emotion change 2

3 Motivation: A Happy System Can we predict users activities and emotion? 3

4 Social Action Prediction SIGKDD 10 4

5 5 What can we do in SNS?

6 Social Action Twitter Flickr KDD Add favorites Tweet on Haiti Earthquake Publish in KDD Conference 6

7 User Action in Social Networks Questions: -What factors influence you to add a photo into your favorite list? - If you post a tweet on Haiti Earthquake, will your friends retweet it or reply? Challenge: - How to track and model users actions? - How to predict users actions over time? 7

8 Social Action Modeling and Prediction 1 Influence Action Prediction: Will John post a tweet on Haiti Earthquake? Time t Time t+1 3 Correlation John John 4 Action bias 8 2 Dependence Personal attributes: 1. Always watch news 2. Enjoy sports 3..

9 Statistical Study: Influence Y-axis: the likelihood that the user also performs the action at t X-axis: the percentage of one s friends who perform an action at t 1 9

10 Statistical Study: Dependence Y-axis: the likelihood that a user performs an action X-axis: different time windows 10

11 Statistical Study: Correlation Y-axis: the likelihood that two friends(random) perform an action together X-axis: different time windows 11

12 Problem formulation Nodes at time t G t =(V t, E t, X t, Y t ) Actions at time t Edges at time t Attribute matrix at time t Input: G t =(V t, E t, X t, Y t ) t = 1,2, T Output: F: f(g t ) ->Y t 12

13 NTT-FGM Model Influence Correlation Personal attributes Dependence Continuous latent action state Action Personal attributes 13

14 Model Instantiation How to estimate the parameters? 14

15 Model Learning Two-step learning Extremely time costing!! Our solution: distributed learning (MPI) 15

16 Experiment Data Set ( Action Nodes #Edges Action Stats Twitter Flickr Arnetminer Post tweets on Haiti Earthquake Add photos into favorite list Issue publications on KDD 7, , ,568 8, , ,253 2,062 34,986 2,960 Baseline SVM wvrn (Macskassy, 2003) Evaluation Measure: Precision, Recall, F1-Measure 16

17 17 Performance Analysis

18 Factor Contribution Analysis NTT-FGM: Our model NTT-FGM-I: Our model ignoring influence NTT-FGM-CI: Our model ignoring influence and correlation 18

19 19 Efficiency Performance

20 MoodCast: Emotion Prediction via Dynamic Continuous Factor Graph Model ICDM 10, IEEE Transactions on Affective Computing 11 20

21 Happy System 21 Can we predict users emotion?

22 MoodCast: Dynamic Continuous Factor Graph Model MoodCast Social correlation g(.) Jennifer Allen Mike Temporal correlation h(.) Jennifer yesterday Neutral Happy Allen Happy Jennifer today Neutral Mike Predict Jennifer tomorrow? Attributes f(.) location call sms 22

23 Dynamic Continuous Attributes We cannot use such a NTT-FGM model to do this Our solution 1. We directly define continuous feature function; 2. Use Metropolis-Hasting algorithm to learn the factor graph model. 23

24 Time t Problem Formulation G t =(V, E t, X t, Y t ) Emotion: Sad Time t-1, t-2 Learning Task: Attributes: - Location: Lab - Activity: Working 24

25 Dynamic Continuous Factor Graph Model Time t Time t : Binary function 25

26 Model Learning Attribute Social Temporal 26

27 Data Set ( Baseline SVM SVM with network features Naïve Bayes Naïve Bayes with network features Evaluation Measure: Experiment Precision, Recall, F1-Measure #Users Avg. Links #Labels Other MSN ,869 >36,000hr LiveJournal 469, ,665,166 27

28 28 Performance Result

29 Factor Contributions 29 Mobile All factors are important for predicting user emotions

30 Conclusions We propose Noise Tolerant Time-varying Factor Graphs for modeling and predicting user actions We propose MoodCast, a dynamic continuous factor graph model for modeling and predicting user emotions Next step: a joint model? 30

31 Future Work: Social Prediction To predict more 31

32 Retweet Predicting (CIKM 10) Bob Andy Dan Who will retweet it? When you post a tweet Jon 32

33 Following Prediction (submitted) time 1 time 2? y 1 =1? 1 When you follow a friend, how likely he will follow back?? 2 Can we predict more such as the following scale and following speed? 33

34 A challenging question Can social prediction really help human beings? 34

35 Representative Publications Chenhao Tan, Jie Tang, Jimeng Sun, Quan Lin, and Fengjiao Wang. Social Action Tracking via Noise Tolerant Time-varying Factor Graphs. KDD 10. Jie Tang, Yuan Zhang, Jimeng Sun, Jinghai Rao, Wenjing Yu, Yiran Chen, and ACM Fong. Quantitative Study of Individual Emotional States in Social Networks. IEEE Transactions on Affective Computing Yuan Zhang, Jie Tang, Jimeng Sun, Yiran Chen, and Jinghai Rao. MoodCast: Emotion Prediction via Dynamic Continuous Factor Graph Model. ICDM'10. 35

36 Thanks! System: HP: 36