UTILISING MACHINE LEARNING TO FORECAST AND ANALYSE CRIME RATES

Abstract

Crime poses a significant threat to the security and jurisdiction of any nation. Consequently, crime analysis has gained increasing importance as it involves discerning the when and where of criminal activities through the analysis of spatial and temporal data. Traditional methods such as paperwork, reliance on investigative judges, and statistical analysis have proven inefficient in accurately predicting the time and location of crimes. However, the integration of machine learning and data mining techniques into crime analysis has led to a substantial improvement in the accuracy of crime analysis and prediction. This study delves into various aspects of criminal analysis and prediction using a range of machine learning and data mining methods. It aims to provide a succinct overview of how these algorithms are employed in crime prediction, based on the accuracy metrics of previous research. The intention is not only to inform crime researchers about these techniques but also to support future endeavors in refining crime analysis. This review study encompasses an exploration of crime definitions, challenges in prediction systems, and classifications, accompanied by a comparative analysis. Through a comprehensive examination of the literature, it becomes evident that supervised learning approaches have been the predominant choice for crime prediction in numerous studies, surpassing other methodologies. Furthermore, Logistic Regression emerges as the most robust method for predicting crime based on existing research findings.