Research on Crime Occurrence Prediction Using Machine Learning Methods—Considering Four Types of Crime in Chicago

The establishment of crime prediction models using big data has become a key part of current police work. This paper combines binary classification with SVM, Random Forest, XGBoost, and GBDT methods based on the data of burglary, battery, assault, and criminal damage in Chicago from 2015 to 2020, and compares the prediction results. In this experiment, GBDT and XGBoost presented relatively stable and excellent data, and the scores of F1-score were 0.69 and 0.71 respectively, with high scores. It means that the model has strong generalization ability, can better adapt to different data distributions, and can be used more in the future to provide decision support for local police work.