270 likes | 413 Views
Predicting Emerging Social Conventions in Online Social Networks. Farshad Kooti * Winter Mason † Krishna Gummadi * Meeyoung Cha ‡ MPI-SWS * Stevens Institute of Technology † KAIST ‡. Metric. Imperial. Linguistic conventions. Hello. Hey. Aloha. How’s it going.
E N D
Predicting Emerging Social Conventions in Online Social Networks Farshad Kooti* Winter Mason† Krishna Gummadi* Meeyoung Cha‡ MPI-SWS* Stevens Institute of Technology†KAIST‡ CIKM 2012
Metric Imperial Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
Linguistic conventions Hello Hey Aloha How’s it going Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
The retweeting convention RT @Bob: CIKM started RT @Bob: CIKM started CIKM started Bob Alice Prediction of Emerging Social Conventions in OSNs- Farshad Kooti Quoting another user while citing the original author
Why retweeting convention? Prediction of Emerging Social Conventions in OSNs- Farshad Kooti • Information-sharing channels are explicit in Twitter • Specific to Twitter: exposures within the community • Containedin Twitter, hence capturing all usages
Twitter dataset Prediction of Emerging Social Conventions in OSNs- Farshad Kooti • Used near-completedata from 03-2006 to 09-2009 • 54 million users • 1.9 billion tweets • 1.7 billion follow links • Follow links are a snapshot of the network in 2009
The retweeting variations • Searched for syntax token @username • “Adopter” refers to a user using the variation at least once Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
Our study of retweeting convention Prediction of Emerging Social Conventions in OSNs- Farshad Kooti Characterizing the emergence [ICWSM’12, best paper award] Predicting the adoption process [this work, CIKM 2012]
Defining prediction problem Prediction of Emerging Social Conventions in OSNs- Farshad Kooti Suppose we are given a social network with records of users, their interactions, and times of adoptions. However, information about which variation was adopted by user u at time t is hidden. How reliably we can infer that user u has adopted variation v at time t?
RT or via or ...? RT @john: tweet Bob tweet (RT @joe) via @jane: tweet 2,053 TWEETS 1,738 FOLLOWING 1,581 FOLLOWERS Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
Motivation & Problem Features impacting adoption Predictive power & results
Feature categories Personal Social Global Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
feature: # of followers Personal Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
features Social # of exposures # of adopter friends Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
feature: # of adopter friends Social Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
feature: adoption date Global Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
All the considered features Personal Social Global Prediction of Emerging Social Conventions in OSNs- Farshad Kooti • # of followers and friends, # of posted tweets and URLs, join date, geo-location • # of exposures, # of adopter friends • Time of adoption
Motivation & Problem Features impacting adoption Predictive power & results
Measuring the predictive power of features Prediction of Emerging Social Conventions in OSNs- Farshad Kooti • We calculate Information Gain (IG) of each feature, which shows the predictive power • IG: change in entropy (measure of uncertainty) because of the given feature • IG(Variation, feat.) = H(Variation) - H(Variation|feat.)
Predictive power of features: results • Findings: • # of exposures has more predictive power than# of adopter friends • Geographyisnotimportant Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
Prediction methodology Prediction of Emerging Social Conventions in OSNs- Farshad Kooti • Using different ML classifiers: Bayesian models, boosting, decision trees, etc. • Bagging yields the best result • Feature selection techniques to find best subset of features (excluded 8 features)
Prediction accuracy Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
Dealing with unbalanced classes Prediction of Emerging Social Conventions in OSNs- Farshad Kooti • Problem: • Most of the adoptions (68%) are RT • A simple classifier of always predicting the most used variation performs good • Solution: • Take the same number of cases from two groups (baseline: 50%)
Prediction accuracy from balanced data Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
Stronger definitions Prediction of Emerging Social Conventions in OSNs- Farshad Kooti
Summary Prediction of Emerging Social Conventions in OSNs- Farshad Kooti • Predicting adoption of social conventions • Investigated impact of various factors • Global feature trumps social and personal features • The number of exposures had more predictive power than number of adopter friends • Using the features from network is not enough for a prediction with high accuracy
Thank you! Prediction of Emerging Social Conventions in OSNs- Farshad Kooti