Traffic accidents result in around 1.35 million deaths annually, and it is projected that they will rank as the fifth leading cause of mortality by the year 2030. This study compared two neural network architectures CNN and LSTM, to predict speeding violations on highways using 2022 speeding ticket data from speed cameras, obtained from the Peruvian government’s open data platform. The data were preprocessed and adapted for the models, which were then trained and evaluated with regularization techniques and hyperparameter tuning to optimize performance. The results revealed that the CNN model achieved a loss of 2.96 and an error rate of 0.57%, whereas the LSTM model had a significantly higher loss of 195.10 and an error rate of 8.97%. These findings suggest that the CNN architecture is more effective for predicting speeding violations on highways. Future research should consider incorporating additional factors and exploring hybrid architectures that combine CNN and LSTM to enhance prediction accuracy.