Social data analysis for predicting next event

Abstract

Twitter is a user-friendly social network which deserves its real-time nature. With the help of an algorithm, the investigation can be made with regard to some of the real-time events such as earthquake. The target event is assumed and classified based on the keywords, number of words and their context. The probabilistic spatiotemporal model is provided which can find the Centre of the event location. The Twitter users are regarded as sensors and apply particle filter, mainly used for detecting the location. Because of the numerous earthquakes and the large number of Twitter users throughout the country, we can detect an earthquake with high probability merely by monitoring tweets. Our system detects earthquakes promptly and notification much faster than JMA (Japan Meteorological Agency) broadcast announcements.