STANet: Spatio-Temporal Attention-Enhanced WaveNet for Crime Hotspot Prediction

An accurate prediction of crime hotspots is critical for optimizing law enforcement strategies and urban planning. In this paper, we introduce STANet, a Spatio-Temporal Attention-Enhanced WaveNet model developed to predict crime hotspots using spatial and temporal crime data. K-Means clustering and advanced data preprocessing techniques are combined in STANet to analyze five years of crime incidents reported in Chicago. In the model, spatial-temporal dependencies are incorporated through WaveNet architecture and enhanced through attention mechanisms in order to capture complex crime patterns more effectively. As a result of our experiments, we are able to demonstrate that STANet outperforms traditional models, such as XGBoost, DNN, and decision trees, with The model achieved an accuracy of 86%, along with precision and recall values that demonstrate a balanced performance in predicting crime hotspots. As a result of this mechanism, the model can identify and focus on the most relevant time steps dynamically, improving its accuracy in predicting the future. STANet can be used to predict hotspots for crime, offering actionable insights for resource allocation and crime prevention. To enhance the predictive capability of the model, further exploration will involve expanding the dataset and incorporating additional features.