Exploratory Data Analysis and Predicting Indian Crime Against Women Using Machine Learning Approaches

Crime forecasting is a critical endeavor aimed at preventing future criminal actions, specifically the rising rate of violence against women in India. This study proposes the model to predict crimes committed against women using data from the National Crime Records Bureau's (NCRB) monthly crime report. The dataset encompasses district and state-wise crime activities from 2001 to 2014. The proposed work objectives are identifying factors influencing these crimes and developing predictive models using machine learning (ML). The study employs Exploratory Data Analysis (EDA), to enhance the predictive power of the models. The suggested study uses XGBoost, Random Forest (RF), and Neural Network (NN). The neural network algorithm outperforms the Random Forest and XGBoost techniques among these three. The Random Forest algorithm has the worst performance. After comparing the accuracies of various methods, such as XGBoost (92%), and Random Forest (89.84%) classifiers using imbalanced datasets (without employing Synthetic Minority Oversampling Technique (SMOTE) technique), it was found that the neural network algorithm (95.81%) delivers the highest accuracy. This study looks at model accuracies and finds that, after applying the SMOTE algorithm, neural network algorithms performed best, with an accuracy of 96.76%, outperforming the accuracies of the remaining models, which included the Random Forest (93.08%) and XG-Boost (95%).