Event Detection Based on Interactive Communication Streams in Social Network

With the rapid development of social network, users are used to discuss and plan activities with their online friends by the form of interactive communication streams in mobile social networks. The information of these activities can be applied to track the prospective behavior and following demand of users for smart service supporting systems. In this paper, we describe a prospective activity as an event that happens at scheduled location and time, and propose a method to detect the event. The data of interactive communication streams are divided into five types, including request, question, confirmation, denying and uncertainty. We employ a combined multi-classifier based on D-S evidence theory to classify interactive streams into the five types, and extract the information of event, location and time in each text. Then a reasoning model is proposed to deduce the user’s final intention of prospective activity through the series of different types of interactive streams. Based on the real data collected from social network, the experimental results show that our method could detect the information of events effectively.