INNS: A Logical Prediction of Crime Activities against Women Harassment by using Intensive Neural Network Strategy

Crimes against women, particularly harassment and assault, remain a significant issue globally, demanding proactive solutions to improve public safety. This paper presents INNS: A Logical Prediction of Crime Activities against Women Harassment using Intensive Neural Network Strategy, a novel hybrid model designed to predict high-risk areas for crimes against women. The model integrates machine learning and deep learning components, including Random Forest, XGBoost, Support Vector Machines (SVM), Artificial Neural Networks (ANN), Gated Recurrent Units (GRU), and K-Nearest Neighbors (KNN). The Women Safety Dataset (2023), comprising demographic, geographic, and time-based features, is used to train and evaluate the model. Preprocessing techniques like handling missing data, normalization, and Recursive Feature Elimination (RFE) for feature selection are applied to enhance the model's efficiency. Exploratory Data Analysis (EDA) is conducted to understand key patterns and trends in crime data. The INNS model achieved an accuracy of 95.47%, significantly outperforming other models like ANN (90.33%) and XGBoost (88.02%). Its high precision (93.81%) and recall (94.22%) demonstrate its capability to minimize false predictions. The model’s complex architecture, though resource-intensive, proves effective in handling the intricacies of crime data. This research highlights the potential of hybrid architectures in predicting crime-prone areas and guiding law enforcement agencies in taking proactive measures to ensure women’s safety.