This Complete Research develops a predictive model to elucidate factors affecting dropout rates in the first two years of tertiary education, using data from 1266 students at a School of Engineering in Chile. Focusing on socio-demographic variables from an institutional survey, such as family background, economic status, and employment, the study employs a quantitative, non-experimental methodology alongside Machine Learning techniques within a Knowledge Discovery in Databases (KDD) framework. Of the methods tested, including Neural Networks (NN), K-Nearest Neighbor (KNN), Naive Bayes (NB), Decision Tree (DT), and Logistic Regression (LR), the NN model proved most effective, demonstrating high Accuracy, Sensitivity, and Specificity (all above 0.7). A Weight-Based Feature Importance analysis identified economic factors, family composition, and social relationships as the top variables impacting dropout. This research enhances our understanding of factors influencing student attrition in the School of Engineering. The model acts as an early alert system, identifying potential dropouts and at-risk student groups before they commence their studies. Consequently, it allows the implementation of early interventions, such as financial support, improved study methods, and professional counseling, thereby significantly reducing dropout rates and improving student success.