Spatiotemporal Crime Prediction for Women’s Safety in India Using a Hybrid CNN-LSTM Model

The safety of women is very important as crimes against women are rising daily, especially in India, and this has led to a strong need for predictive models that can help policymakers prevent crime. This study suggests a hybrid deep learning model for spatiotemporal crime prediction that combines Long Short-Term Memory (LSTM) networks with Convolutional Neural Networks (CNN). Convolutional neural networks can extract spatial features from the data and the LSTM extracts the sequential data which provides more accuracy. The dataset contains state-wise and district-level data on the crimes that happened against women from 2001 to 2014. The hybrid model outperforms the other traditional models, attaining a 97 % accuracy rate. It is shown that this hybrid architecture performs well in addressing the challenging issue of crime prediction and provides policymakers with useful information to enhance women’s safety. Future studies will focus on real-time crime prediction and the incorporation of socioeconomic factors to improve predictive capabilities.