Crime Forecasting using Interpretable Regression Techniques

Abstract

Over the past years there has been an increase in crimes like theft and burglary, however, this is not evenly distributed because most criminals repeatedly commit crime in the same area until they are arrested. The Pareto principle or the principle of factor sparsity, helps to explain why so many crimes happen in specific places. Fortunately, the field of electronics and informatics has seen a significant advancement in recent times, particularly in the area of artificial intelligence (AI) and its applications. One such application is the use of AI in crime forecasting and analysis, which has the potential to facilitate smart living and smart city initiatives, as well as economic development. This paper presents a study on the use of regression techniques and predictive models, Linear regression, LASSO regression and ridge regression for crime prediction and magnitude estimation, with an average predictive accuracy of 94.0%. The most important features contributing to crime prediction were identified and analyzed using heatmaps thus providing insights for proactive action by crime prevention authorities. Additionally, Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive explanations (SHAP) were provided to enhance the interpretability and accountability of the developed models. The results of this study have very positive potential implications for smart living and smart city initiatives for developing economies.