Crime hotspot detection using statistical and geospatial methods: a case study of Pune City, Maharashtra, India

The great disparities in resource availability, power, wealth, social institution and opportunities have led to rise in crime incidents especially in urban agglomerations. Therefore, addressing violence and mapping of crime hotspots is important in order to build a socially sustainable community. Current work aims to apply Space-Time Permutation Model (STPM) inbuilt in SatScan (an open source statistical tool), for crime hotspot detection in Pune city, Maharashtra, India, and compare the result with two GIS-based statistical methods namely, Kernel Density Estimation (KDE), and Getis-Ord Gi*, being utilized extensively for crime hotspot detection and mapping. The datasets of four different crime namely robbery, molestation, rape and dacoity for a period from 2012 to 2015 were used. Twenty six significant crime clusters (p < 0.1, log likelihood ratio, LLR > 1.3) obtained through SatScan show similar pattern, as obtained from KDE (92% matching) and Getis-Ord Gi* (92% matching). A total of 11 clusters were obtained as most significant clusters (with p < 0.000, LLR > 2.72, predictive accuracy index, PAI > 57.11). Locations situated in the central part of the city such as Swargate, Bibwewadi, Deccan, Bund Garden, and Farashkhana show relatively more crime clusters possibly due to inter-state road connectivity, more liquor shops, famous tourist places, large no. of auto rickshaws and tongawals, and a variety of immigrants in the search of livelihood.

Automated summary

Addressing violence and mapping crime hotspots are crucial for building a socially sustainable community. This study compares three methods for crime hotspot detection in Pune city, India, and finds that they yield similar patterns.