Comparative Analysis of Supervised Machine Learning Models for Crime Type Classification on Chicago Data

Abstract

The increase in crime can be caused by several factors. As growth in population and density, criminal behavior, domestic crime growth, and so on. This causes difficulties for the law enforcement agencies to control and monitor on a regular crime basis since it requires forecasting and probabilities, significant progress has been made from traditional machine learning techniques (ML) to modern techniques neural networks (NNs), the advancement of these technologies have improved the operational efficiency of critical infrastructures but have also rendered these substantially more vulnerable to domestic crime. In this study, we explore how at least four machine learning techniques perform in prediction of crime under measured experimental conditions. A dataset comprising 3,000 crime records was used for the experiment, consisting of three balanced classes: narcotics, burglary, and robbery, with 1,000 samples allocated to each category. The results show that decision tree (DT) achieved accuracy of 90.33%, and random forest (RF) with an accuracy of 90.33% consistently outperformed logistic regression (LR), support Vector machine (SVM), the SVM [UY1] model reached a higher 90.16%, while the accuracy obtained by LR is peak at 90.03%.  [UY1]Add the full meaning of these similar to what you did for ML, RF, DT…...