Discovering educational patterns in higher education through Data Mining Techniques.

Abstract

Educational Data Mining (EDM) applies data science methods to analyze academic information and support evidence-based decision-making. This research article, conducted at a Higher Education Institution in Mexico, identifies patterns associated with academic success or risk among undergraduate students through four phases: data collection, cleaning, modeling, and analysis. Using unsupervised techniques such as PCA and K-means, three student profiles were identified, one of which exhibited signs of emotional or academic disengagement. In the predictive stage, supervised models such as MLP, SVM, and Logistic Regression achieved accuracies of 95–96%. The findings demonstrate that machine learning is a key tool for identifying students at risk and strengthening institutional intervention.