The integration of machine learning (ML) and deep learning (DL) technologies has fundamentally transformed crime prediction. These advancements enable the detection of intricate spatial and temporal patterns in crime data, patterns that traditional methodologies often fail to capture. This chapter offers a synthesis of machine learning (ML) and deep learning (DL) techniques for crime hotspot prediction, tracing advances from traditional supervised and unsupervised methods to modern architectures such as convolutional neural networks, recurrent networks, graph neural networks, and Transformers. We examine how diverse data sources, including historical crime records, demographic and environmental indicators, social media feeds, and real-time IoT inputs, can be integrated to enhance predictive accuracy and situational awareness. Through detailed comparative analysis, we highlight the trade-offs between ML’s interpretability and DL’s representational power, and we introduce emerging hybrid frameworks that combine both for more robust, scalable models. We discuss the imperative of bias mitigation and the need for transparency in AI models to ensure their responsible deployment. Ethical frameworks are crucial to prevent misuse and ensure that AI tools contribute positively to societal safety. By integrating advanced technological solutions with robust ethical safeguards, this chapter elucidates how predictive modeling can significantly enhance public safety and optimize resource allocation within urban environments. This balanced approach not only leverages technological innovation but also prioritizes ethical responsibility, thereby fostering a safer, more just society.