Predicting Users Behaviours in Distributed Social Networks Using Community Analysis

Abstract

Prediction of user behaviour in Social Networks is important for a lot of applications, ranging from marketing to social community management. This chapter is devoted to the analysis of the propensity of a user to stop using a social platform in a near future. This problem is called churn prediction and has been extensively studied in telecommunication networks. We first present a novel algorithm to accurately detect overlapping local communities in social graphs. This algorithm outperforms the state of the art methods and is able to deal with pathological cases which can occur in real networks. It is then shown how, using graph attributes extracted from the user’s local community, it is possible to design efficient methods to predict churn. Because the data of real large social networks is generally distributed across many servers, we show how to compute the different local social circles, using distributed data and in parallel on Hadoop HBase. Experimentations are presented on one of the largest French social blog platforms, Skyrock, where millions of teenagers interact daily.