Microblog users' life time activity prediction

As the fast development of online social media, social network services have become an important research area nowadays. Particularly, microblog as new social media needs more attention. Most of current studies are usually static descriptions or explanations of what already has happened. Limited study has been conducted focusing on SNS users and analysing their behaviors dynamically. In this paper, we firstly segment microblog users based on the recency and frequency of tweet and retweet behavior, then use probability models such as Pareto/NBD and BG/NBD to predict customer lifetime vitality. Our results showed that both Pareto/NBD model and BG/NBD model showed effective ability to fit and predict SNS users' usage behavior on microblog website. Tweet behaviors of sustainably active user base are more suitable for the probability models. Managerial implications of the two models should be highlighted as well. Interaction rate and dropout rate can be considered as the vitality index of the whole user base measuring how active users are and how likely a user is active. Managerial questions such as how active the users are in this platform now and how active the users will be in the future can be answered by applying those models.