Predicting user influence under the environment of big data

A person is considered as an influential individual when his behaviors can trigger other people's reactions. Such phenomenon is called user influence in social networks. Measuring user influence provides insights into dynamics of social network interactions. This makes a fundamental step for constructing marketing strategy, recommendation systems and so on. There have been various studies focusing on this problem. However, it lacks of a satisfactory method to measure user influence in a reasonable way. It is also worth studying on user properties and past activities that contribute to the influence of each user. In this paper, we investigate the attributes of millions of social network users and the content of their messages in order to better predict user influence. These users and messages are from Sina Weibo, which is one of the most popular social networks in China. Our first contribution is to quantify the influence of individuals within a period of time by using a new approach and find the individual influence of most users changing over time, but most changes are not significant. Our second contribution is to propose a phrase merging algorithm for obtaining high-quality phrases, which are very helpful for extracting the topics that each user is interested in. Our third contribution is to predict the influence of each user with a higher precision.