Sexual violence against minors is a social and public health problem that has devastating consequences for the victims as it causes physical and psychological damage and leaves long-term sequelae that affect emotional and social development. The objective of this study was to implement a predictive model of the recurrence of sexual violence in underage victims using Machine Learning techniques. The applied methodology was based on 5 phases: Obtaining and preparation of the dataset; Pre-processing; Feature selection (XGBoost, LightGM and Random Forest); Machine Learning models (Random Forest, Support Vector Machine, Logistic Regression, K-Nearest Neighbors, Naive Bayes, Artificial Neural Network and Decision Tree) and Evaluation of the models. The best result was obtained with the Random Forest model, combined with the feature selection technique XGBoost: The metrics Accuracy, Precision, Recall and F1-score were 0.82254, 0.83475, 0.8254 and 0.82254, respectively, reflecting its robust ability to discriminate between cases of recidivism and non-recidivism of sexual vio-lence in underage victims.