A competitive model of popularity prediction using correlation matrix

With the rapid development of network, big data has become an important research field for many areas such as data mining, machine learning, information diffusion, the internet of things, and social networks. The multimedia information spreads faster since the rise of social network where people can create and share images, video and audio contents. More and more people share and comment on the web through the mobile terminal to express their mends. Online content captures online behavior of users who communicate or interact on a diversity of issues and topics. How to predict the popularity of the online content happened recently is a hot topic and lots of people are trying to find out the law of information diffusion hidden in it. However, many models assume that information spreads with no external interference in social networks. The research on competitive diffusion is still at the primary stage. The main contribution of this paper is to solve the problem that there are few or no work for popularity prediction based on multi-information, and propose a model based on competitive matrix with a cluster-based news classification approach. The goal of this paper is to accurately estimate the popularity of a given viral topic at final based on the observation of its historical characteristics. And this model is mainly based on the competitive matrix and gradient descent method. Also, the capability of this method provides a better performance in the popularity prediction according to an empirical study.