Efficient Emergency Event Tracking Using Features of Web Data

Huge amount of information has become now available in web services due to their popularity. This web data contains user-contributed information for a variety of emergency events. However, tracking these emergency events is often limited by the lack of efficient tools to analyze the potential events or topics over time, since these events are inherently difficult to predict due to the interference of other unpredictable evolutions. In this paper we propose a two-phase approach, in which we first introduce a novel extraction algorithm to acquire relevant web data and then we utilize a limit theory to determine the periodical convergence time of a specific event, and an event tracking model is constructed using the extracted web data. Based on the significance of multiple features weights and clustering solutions, the interplay between the ordinary events and latent events is discovered to efficiently track the emergency events. Finally, we conduct extensive experiments to verify the effectiveness and efficiency of our approach.