Harnessing Machine Learning and Computer Vision Strategies for Crime Forecasting
Swadha Khatod, Vedanti Shukla, Ayesha Hakim · 2025
In the face of rising crime rates, effective crime prevention has become a critical societal need. Traditional methods of crime analysis and prevention often fall short in handling the vast and complex data involved. This work leverages machine learning to enhance crime forecasting and detection, enabling more accurate and timely interventions. In the first part of the work, statistical data was utilized to predict the age and gender of offenders using Random Forest, SVM, and KNN algorithms, with Random Forest achieving an average prediction time of just 4.42 seconds. In the second part, it focused on crime detection through video datasets, employing Convolutional Neural Networks (CNN) and training five models: ResNet50, VGG16, VGG19, AlexNet, and a custom model. Among these, ResNet50 demonstrated superior performance, achieving an accuracy of 98%. The findings underscore the potential of machine learning in revolutionizing crime prevention, providing a robust and efficient tool for law enforcement agencies.