Crime Prediction Using Machine Learning: The Case of The City of Little Rock

Crime is a severe problem that can cause physical harm, damage, or loss in society. Forecasting has always been in demand to analyze statistical criminal records to create a safe environment. In this study, we aim to develop a machine-learning model to predict criminal activities in Little Rock, Arkansas, and provide insights into crime patterns. We analyzed the publicly available crime datasets from Little Rock Police Department from January 2017 to March 2023 to identify trends and patterns in crime occurrence. We used data cleaning and exploratory data analysis techniques, such as figure-based visualizations, to prepare the data for machine learning. We applied Facebook Prophet, and NeuralProphet, a time-series machine learning model, to predict daily crime counts. The model trained data from January 2017 to November 2021 and was tested on data from November 2021 to March 2023. The model was evaluated using Root Mean Squared Error (RMSE), Mean Squared Error (MSE), and Mean Absolute Error (MAE) with results that represent a model that can accurately forecast crime occurrence in Little Rock. This study can be used to inform law enforcement agencies, policymakers, other research groups, and third parties, such as business owners, to develop effective crime prevention and detection strategies.