Crime prediction and mapping have become crucial in modern law enforcement, necessitating advanced analytical frameworks to process large volumes of communication data efficiently. This study proposes an AI-driven crime mapping model that integrates Density-Based Spatial Clustering of Applications with Noise (DBSCAN), social network analysis, and predictive modeling to enhance suspect tracking and crime investigation. The framework leverages Call Detail Records (CDRs) and spatial data to identify frequent locations, assess social relationships, and forecast future interactions. The proposed Framework utilizes DBSCAN clustering to map and detect frequent location and movement patterns, a Relationship Strength Score (RSS) to quantify social connections, and LightGBM predictive modeling to forecast potential suspect communications. Experimental results demonstrate an 85% accuracy in identifying strong relationships within the suspect’s network and a 92% accuracy in predicting the next likely contact, with an AUC score of 94.8%. These findings show the system’s effectiveness in converting raw CDR data into actionable intelligence, supporting law enforcement agencies in resource allocation, suspect profiling, and crime prevention. The integration machine learning with traditional investigative methods, this research enhances crime mapping capabilities, facilitates proactive decision-making, and provides law enforcement with a scalable approach to analyzing complex criminal networks.