This work presents an innovative methodological framework that integrates Learning Analytics with advanced multivariate statistical techniques to personalize the educational experience in university experimental science courses. The study was implemented with 142 students of bioprocesses, divided into experimental and control groups. Interaction data was collected through Moodle, including access patterns, time spent on resources, activity participation, and assessment results. Principal component analysis reduced the dimensionality of 27 interaction variables, identifying 5 components that explained 78% of variance. K-means cluster analysis revealed four distinct student profiles: "deep explorers," "strategic-selective," "consistent-methodical," and "lagging-intermittent." Discriminant analysis validated this classification with 87% accuracy. Personalized pedagogical interventions were designed for each profile and implemented over 12 weeks. Results, evaluated through MANOVA, showed a significant 23% improvement in academic performance for the experimental group (p<0.01) and a 31% increase in student satisfaction. The study demonstrates how combining Learning Analytics with multivariate statistics provides educators powerful abilities to adapt content, methodologies, and assessments to students' specific needs, especially in complex scientific disciplines where conceptual and practical understanding is fundamental.

Keywords: Learning Analytics, multivariate statistics, personalized teaching, experimental sciences, bioprocesses.