Date of Publication :28th June 2024

Abstract:Crime modeling and forecasting have become an important application of data science and machine learning. Predicting crime rates accurately can help law enforcement agencies deploy resources effectively and plan prevention strategies. Motivated by this, this project aims to analyze crime rate data in India and develop models to predict future crime rates. The dataset comprises annual crime statistics for major offenses in India from 2016-2022 collected by the National Crime Records Bureau along with socioeconomic indicators like literacy, unemployment, etc. Several supervised regression algorithms are implemented including linear regression, SVM, decision trees, and ensemble methods like random forest, stacking, etc. The models are trained on 70% of the data and tested on a 30% holdout set. Performance is evaluated using an accuracy metric. Among all models, the random forest regressor achieved the highest accuracy of 82% on test data. The key findings indicate random forest is the most accurate model for predicting crime rates. This project successfully demonstrates applying machine learning algorithms to analyze crime data and develop models that can forecast future crime rates. The results can empower law enforcement with data-driven insights to deploy resources optimally and plan effective crime prevention strategies.

Reference :