Understanding the Dynamic Nature of Criminal Activity Using Grid-Based Multivariate Spatial and Temporal Clustering

Abstract

Understanding the dynamic nature of criminal activity is crucial for effective law enforcement and urban planning. Crime patterns are influenced by a multitude of factors, including socio-economic conditions, demographic shifts, and environmental changes. Traditional crime analysis methods often fail to capture these complex interactions, leading to less effective interventions. Existing algorithms, such as K-means, DBSCAN, and Hierarchical Clustering, are limited in their ability to simultaneously handle the spatial and temporal dimensions of crime data, resulting in suboptimal identification and tracking of crime hotspots. These methods also struggle with the dynamic and continuously evolving nature of crime data, often requiring extensive parameter tuning and offering limited scalability. Proposed model addresses these limitations by employing grid-based multivariate spatial and temporal clustering. This approach segments the geographic space into manageable cells, allowing for a detailed examination of crime hotspots and their evolution over time. By incorporating multiple variables such as crime type, frequency, and temporal patterns, our model captures the complexity of criminal behaviour more effectively. Performance metrics demonstrate the robustness of our model, with an Adjusted Rand Index (ARI) of 0.82, purity of 0.85, and homogeneity of 0.83, significantly outperforming existing algorithms. Additionally, our model maintains high temporal stability (0.80) and spatial coherence (0.78), providing reliable and actionable insights for real-time crime analysis and strategic planning.