Predicting Crime Hotspots using Machine Learning Techniques

Abstract

This research delves into the application of machine learning algorithms for forecasting crime hotspots by leveraging historical data of public property crime in a major coastal city in southeast China. The study conducts a comparative analysis, emphasizing the predictive efficacy of various machine learning models. Results indicate that the LSTM model surpasses other methods including KNN, random forest, support vector machine, naive Bayes, and convolutional neural networks when utilizing solely historical crime data. Moreover, integrating built environment data such as points of interest (POIs) and urban road network density as covariates into the LSTM model enhances predictive accuracy. These findings bear significance for shaping policing strategies and implementing measures for crime prevention and control.