Educational Data Mining (EDM) is a growing research field, concerned with applying machine learning and data mining techniques to large-scale data from educational settings. One of the main applications of EDM concerns predicting student performance, which can be used to identify at-risk students and define the necessary educational support timely. In this study, a new student performance indicator was proposed relating to the achievement of a degree that combines the three important parameters of degree grade, delay in degree enrollment, and delay in completing a student’s university career. Next, a predictive model based on a real history educational dataset collected at the University of Calabria was proposed, applying numerous Machine Learning techniques: Lasso Regressor, Support Vector Machines, Decision Tree, Random Forest, XGBoost, and an Ensemble Model. Results showed that Ensemble Model has the best overall prediction performance. Additionally, explainable AI techniques were applied to analyze the data generating student performance.