This study evaluated three classification models: Naïve Bayes, Random Forest, and Neural Networks, to predict dating violence. Random Forest was superior, achieving an accuracy of 86% and effectively classifying various forms of violence, although with limitations in less represented categories. Naïve Bayes showed limited effectiveness for infrequent classes, and Neural Networks required a large amount of data to generalize effectively. The results indicate that implementing Random Forest in digital platforms can offer a 4% higher accuracy and a 1% higher F1-Score compared to Neural Networks, and a 60% higher accuracy, 28% higher sensitivity, and 1% higher F1-Score compared to Naïve Bayes. In conclusion, Random Forest is suitable for robust and varied datasets in the analysis of violence.