Crime analysis and prediction using machine learning methods

Abstract

Relevance. As artificial intelligence continues to develop and computer power increases, there is growing interest in applying machine learning methods to tackle tasks that are challenging for humans. One such task is crime prediction, which has significant potential to enhance the effectiveness of law enforcement. Machine learning algorithms, including decision trees and random forests, can identify crime trends, uncover hidden patterns, and determine factors contributing to criminal activity.

Goal. The purpose of this article is to analyze the effectiveness of using machine learning methods, such as linear regression, decision trees, the k-nearest neighbors algorithm, and neural networks for crime analysis and prediction.

Research methods. Comparative analysis, experiment.

Results. The effectiveness of various machine learning methods (linear regression, Lasso regression, ridge regression, k-nearest neighbor regression, decision trees, and radial-basis neural network model) for crime analysis and prediction was analyzed. Among the considered machine learning methods, the k-nearest neighbor regression and radial-basis neural network model showed the best characteristics.

Conclusions. The analysis confirms the need for long-term and operational analysis of statistical information with subsequent prediction of factors and factors that affect crime rates using machine learning methods. The results obtained can help in studying the problem of analyzing the impact of social and demographic factors on crime, which will allow planning preventive measures, distributing law enforcement resources more effectively, etc.