Review paper on Intelligent Crime Rate Prediction using Supervised Machine Learning

Abstract

Crime prediction has emerged as a significant research domain as cities increasingly rely on data-driven strategies to enhance public safety and optimize law enforcement resource allocation. Intelligent crime rate prediction using supervised machine learning has shown considerable promise due to its ability to learn from historical crime patterns and forecast future occurrences with high accuracy. This review paper provides a comprehensive analysis of supervised learning approaches used for crime rate and hotspot prediction, focusing on classical models such as Logistic Regression, Support Vector Machines, k-Nearest Neighbors, and Decision Trees, as well as advanced ensemble methods like Random Forests, Gradient Boosting, and XGBoost. The study also explores recent developments in deep supervised models including LSTM-based temporal predictors and hybrid spatiotemporal frameworks. Key aspects such as dataset characteristics, feature engineering, spatial–temporal aggregation, handling class imbalance, and model evaluation strategies are examined in detail. The review highlights that tree-based ensemble models generally outperform traditional algorithms due to their robustness on heterogeneous real-world crime datasets. However, challenges persist, including data quality issues, bias in police-reported crime records, limited generalizability across regions, and lack of interpretability in complex models. The paper concludes by identifying research gaps and recommending future directions such as fairness-aware modeling, explainable AI, improved spatiotemporal representation, and cross-city transfer learning to support more ethical and effective crime prediction systems.