Enhancing Transportation Safety through AI-Driven Crime Prediction in Urban Environments

Abstract

This proposal introduces a novel AI-based crime prediction model (ST-Cokriging) focused on urban transportation nodes, such as transit stations, where criminal activities are notably concentrated. The project aims to develop a Python-powered machine learning algorithm that utilizes a multi-level geo-statistical model. This model will integrate diverse data sources, including historical crime records, transportation networks, and hub locations, to accurately forecast criminal activities within and around transportation systems. Our method uniquely combines multiple input variables, a step beyond traditional models that typically rely on either historical crime data or transportation land use data alone. The research holds significant potential to transform public policy, guide law enforcement strategies, and optimize resource allocation. By analyzing the complex interplay between crime trends, transportation infrastructure, and safety measures, we intend to equip law enforcement agencies with data-driven strategies to improve public safety. The algorithm will explore multi-source crime data, merging historical crime patterns with the geolocations of transportation hubs. This approach is particularly suited for complex urban areas like the Bay Area and focuses on transportation-related crimes, including thefts, assaults, and robberies. Utilizing a detailed crime kernel density map, the study promises to provide a comprehensive understanding of crime dynamics over time.