A Novel Event-centric Trend Detection Algorithm for Online Social Graph Analysis

Abstract

Nowadays, the identification of the most popular and important topics discussed over social networks, is became a vital societal concern. For real-time tracking the hot topics, we proposed a novel event-centric trend detection algorithm, which called Ec_TD algorithm to attempt to add event attributes into the structure of the social networks, then, mining the subgraphs induced by specific attributes which using correlation function measures the correlation of event-changing attributes based on the attribute-extended social network structure. Our experiment shows that Ec_TD algorithm is performed significantly better in real-time event detecting and mining the potential relationships between attributes and vertexes. Moreover, we used true big data to test this algorithm which has substantially reduced respond time, and to prove the feasible of the idea.