Efficient Machine Learning Approach for Crime Detection in India
Segura-Hernández Angel, D. Narmadha Naveen, D. Naveen Sundhar, Kovuru Lourd Victoria · 2024
This study leverages Gradient Boost classifiers and hyperparameter optimization techniques, including Grid Search and Randomized Search, to develop a crime detection model focusing on offenses against women. Utilizing a dataset detailing state, year, and crime type, the research aims to create a robust detection mechanism. Hyperparameter tuning is employed to refine the model, with both Grid Search, assessing a predefined hyperparameter space, and Randomized Search, exploring hyperparameters randomly, being utilized for optimization. The models' effectiveness is evaluated using Mean Squared Error (MSE), Mean Absolute Error (MAE), and R-squared (R2) metrics. The Random Forest Classifier (RFC) is identified as the most effective, showing superior accuracy in crime detection against women, with a notably low R2error rate of 0.6. This study underscores the significant potential of machine learning in enhancing societal safety and security, particularly in detecting crimes against women.