Answers are grounded in this paper and cite their sources.
Geospatial modeling study uncovers distinct crime hotspots and temporal trends in Addis Ababa, highlighting the utility of predictive machine learning for urban policing.
Urban crime remains a major challenge in rapidly expanding cities, affecting public safety, economic development, and social stability. Yet crime prediction research remains concentrated in Western cities with strong data infrastructures, leaving rapidly urbanizing non-Western contexts underexplored. This study addresses this gap through machine learning and geospatial analysis of crime in Addis Ababa, Ethiopia, using Random Forest and Kernel Density Estimation (KDE). It identifies crime hotspots, temporal patterns, and the predictive value of urban environmental features. Addis Ketema, Bole, and Nifas Silk Lafto show the highest crime concentrations, particularly for violent crime, vehicle and spare-part theft, and fraud. Crime peaks during summer, weekends, and morning and evening periods, while KDE identifies major hotspots in Merkato, Megenagna, Autobis Tera, and Atena Tera. Homicides peak in summer, whereas car thefts increase during autumn and winter. Using a stratified 70/30 train-test split and five-fold cross-validation, the Random Forest model achieved mean training accuracy of 0.76 (SD = 0.03) and test accuracy of 0.78, with macro F1-score of 0.72 and AUC-ROC of 0.81. The findings demonstrate a transferable AI-assisted workflow for targeted policing, resource allocation, and evidence-based urban safety planning in data-constrained cities.
No takes yet. Share an insight, caveat, or question.
Engda et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: