Feature Constrained Parallel Data Processing Approach for Spatiotemporal Event Detection

The enormous usage of Online Social Networks (OSN) leads to unleashing usage of the smart technologies in social life which becomes intertwined. Generating meaningful patterns out of various location-based streaming social big data fascinated the techies around the globe to develop innovative approaches to enhance emergency support on-demand. In the specific situation like disasters, nature setbacks down at unpredictable instances. During such circumstances, a scalable and low latency analytics approach is highly required to identify the event and its location at very high speed for recovery. In this paper, the spatiotemporal event detection approach is proposed based on dynamic features emerged from the location of interest in OSN using a map-reduce framework. The experimental results address the advantage of MapReduce paradigm is indeed suitable for scalable and high-speed data streams with minimal latency. In addition, an information theoretic emergency logistic mapper is discussed as a part of disaster recovery phase which can be accomplished via social big data analysis in near future.