Machine Learning Methods for Crime Scene Prediction

Machine Learning (ML) has always been a useful method for scientists who are devoted to stopping crime, especially in the field of crime prediction. Challenges of crime prediction can be divided into two equally vital parts: i) the prevention of crime before they take place; ii) the timely reduction and termination of harms after the appearance of victims. This paper manages to tackle the second problem by predicting both the latitude and the longitude of crime scene with data of crime in Los Angeles from 2020 to present. This article starts from scratch by using several classical models of ML (Linear Regression, LinearSVR and DecisionTree) to perform crime scene prediction. Additionally, Mean Squared Error (MSE) is selected as the criteria to judge the efficiency and accuracy of different models during comparison on this problem. This paper hopes to provide a prospect to call attention to more research and study on ‘after-case-prediction’ rather than simply focus on forecasts before assaults. I conduct all the experiments on datasets of Los Angeles Crime Data which is available at https://www.kaggle.com/datasets/hemil26/crime-in-los-angeles. The experiment validates the effectiveness of the method.