Crime remains a persistent challenge within urban landscapes, necessitating innovative approaches for prevention and the improvement of public safety. Through the provision of advanced technologies, law enforcement organisations will be able to allocate resources, make data-driven, and informed decisions on crime prevention. The empirical study also sheds light on the effects of regional and contextual changes on crime prediction by revealing the performance differences of different models across cities. Focusing on the vibrant dynamics of two major cities, the empirical study offers an in-depth exploration of crime classification and prediction. Twelve distinct models were employed to classify crimes based on location and associated factors, which will help to analyse historical crime data and identify patterns and trends. The accuracy of each model was assessed to discern the effectiveness in predicting the occurrence of crime in different places. In the context of New York City, the models achieved remarkable accuracy with the highest reaching 99.94% meanwhile in San Francisco achieved an accuracy of 94.27%. These findings offer a valuable resource for law enforcement organisations, urban planners, and policymakers looking to improve public safety and security strategies, as well as for continuing discussions on crime analysis and predictive policing.