Predicting the diffusion of brand's stories in social network

Abstract

The emergence and growing of social media allows one con-sumer to communicate with thousands or millions other consumers. The consumer-generated stories about a brand or a product can be widely propagated and as a consequence can have a big impact on the market-place and indirectly affect the success of the brand. Therefore, modeling the information diffusion in social media is crucial for business managers in order to both understand the information propagation and to better control it. Our research aims at predicting whether a tweet about a brand is going to be diffused and the level of the diffusion. We apply several machine learning classifiers using user-based, time-based and content-based features associated to tweets and developed several new features some content-based. We show that our method significantly improves F-measure by about 4% compared to the state of art. We also show that the numbers of a user’s followers, number of communities that this user belongs to, and number of likes that a user has made on his time line are the most important features for the predictive model.