In crime data analysis, the accurate prediction and identification of trends are often hindered by imbalanced datasets, where certain categories, such as specific castes, occupations, or regions, are underrepresented. This study addresses this problem by leveraging Generative AI, specifically Generative Adversarial Networks (GANs), to augment a dataset of crime complainants. The original dataset, containing information such as age, caste, occupation, and city, was limited by missing data and underrepresentation in key demographic categories, which restricted the performance of machine learning models. To solve this, synthetic data was generated using GANs to create realistic complainant profiles, enhancing the dataset’s diversity and improving the robustness of predictive models. Clustering analysis was performed to identify patterns in complainant profiles, and a Random Forest classifier was used to predict occupations based on age, caste, and city. The augmented dataset significantly improved the model’s accuracy, precision, and recall metrics, compared to the model trained solely on the original dataset. The findings suggest that Generative AI can be effectively applied to enhance predictive capabilities in crime analysis by augmenting underrepresented data categories. These insights could have significant implications for law enforcement agencies, allowing them to better allocate resources, predict crime trends, and serve underrepresented populations more effectively.