Predicting microblog users’ lifetime activities – A user-based analysis

• Pareto/NBD model and BG/NBD model showed effective ability to fit and predict SNS interaction behavior.
• We discovered that user base analysis models are more adapted to modeling intrinsic motivated behavior.
• User segmentation can improve the prediction accuracy by distinguishing currently active and inactive users.

With the rapid development of online social media, social networking services have become an important research area in recent years. In particular, microblogging as a new social media platform draws much attention from both researchers and practitioners. Although most current studies focus on the effect of social networks on the diffusion of services or information, most are descriptions or explanations of what has already happened. This study focuses on future activity by employing probability models such as the Pareto/NBD and BG/NBD models to predict user lifetime vitality. Three experiments were implemented to test the two models. Our results showed that both the Pareto/NBD model and the BG/NBD model were effective in predicting SNS user usage behavior on microblogging websites. It was found that tweeting behavior is more suitable for such probability models than retweeting behavior and user segmentation can improve prediction accuracy by distinguishing between currently active and inactive users.