Jiajunzi Niu

The trend of vehicle-related crimes in Toronto, Canada is escalating rapidly and is a serious concern for residents and officials, thus the necessity to understand, predict, and prevent such crimes is paramount. This study employs GIS and geostatistical analysis, specifically using H3 hexagons as the spatial unit, in conjunction with machine learning techniques. This work identifies hotspots, investigates spatial and temporal patterns, and evaluates machine learning algorithms for hotspot prediction. Spatial autocorrelation techniques and the Hot Spot Analysis Comparison reveal significant spatial interactions and evolving hotspots. Machine learning models, specifically Support Vector Machine, Gradient Boosting and Random Forest, demonstrate accurate predictive capabilities for vehicle offences, presenting actionable insights for law enforcement. The findings of this study extend beyond Toronto, potentially offering a robust framework for understanding vehicle-related offences and contributing towards informed policy-making and strategic crime prevention in other cities.