A place-based big data policing approach to car theft prevention

• In Event: WG-PLACE 3: Understanding and Preventing Auto Thefts

Abstract

This presentation explores the application of a place-based big data policing approach for predicting and preventing car theft incidents. First, an overview is provided of (which and) how big data sources, including location technology data, municipal service records (e.g., parking registrations), and emerging sources like mobile phone data, can be used to quantify potential theft targets (i.e., vehicles) and their associated attributes. Furthermore, we delve into the integration of predictive big data analytics for predicting car theft occurrences. We present preliminary findings from a comparative analysis of diverse machine learning models applied to a car theft dataset spanning the period 2007-2018 for the city of Ghent, Belgium. Specifically, three machine learning algorithms are employed and compared: Ensemble Neural Network (ENN), Random Forest (RF), and K-Nearest Neighbor (KNN). The performance of these algorithms is assessed through various established performance metrics, including but not limited to, direct hit rate, near-hit rate, precision, and F1 score. Finally, the implications of our findings are discussed in light of different crime prevention strategies within the context of car theft.

Authors

• Robin Khalfa, Ghent University

• Wim Hardyns, Ghent University