A PSO-Optimized Model for Identifying Spatio-Temporal Hotspots of Terrorist Incidents in India

Abstract

Terrorism is a global issue that prevails throughout the world on all scales. As the distribution of terrorist activities does not follow a random pattern in space and time, its spatiotemporal analysis has drawn considerable attention in recent years. Further, timely identification of Spatio-temporal terrorist activity hotspots is vital to prioritize the security efforts put by a country's security enforcement agencies. The state-of-the-art methods for Spatiotemporal hotspot detection are based on scan statistics, which enumerates many Spatio-temporal cylinders, making it a computationally expensive approach. Therefore, this paper presents a time-efficient Particle Swarm Optimizer (PSO) based algorithm to detect the most significant Spatio-temporal hotspots. We formulated an optimization model for the problem and applied three variants of PSO viz. conventional PSO, HCL-PSO, and Ensemble PSO. Finally, these schemes have been used to detect spatio-temporal hotspots of different terrorist attacks in India. The results obtained by PSO-based methods have been compared with SaTScan over two parameters: the time required to detect the hotspot and its quality. All the PSO-based schemes significantly outperformed SaTScan in the timely identification of the hotspots. In addition, the quality of hotspots detected by HCL-PSO is at par with SaTScan, whereas the quality of hotspots detected by the other two approaches is slightly lesser than SaTScan. However, the quality of hotspots detected by the other two variants of PSO is slightly lesser than SaTScan. The results are statistically validated using Friedman's statistical test.

Description