Predictive Model Design applying Data Mining to identify causes of Dropout in University Students

Abstract

This paper presents a prediction model developed with techniques and methods Data Mining (DM) with classification rules and selection of attributes. The objective is to identify patterns related to the aspects of greater influence in the for school dropout in Higher Education Institutions (IES) in the Mexico State. The research begins with an exploratory, correlational and explanatory analysis, guided by the strategy and the life cycle of a DM project. Subsequently, the training of the model is carried out with a sample of 170 students, in which different classification algorithms are applied (JRIP, OneR, ZeroR, J48, REPTree) and selection (CfsSubsetEval and BestFirst). The best results are obtained in the identification of the causes that impact school dropout and failure by 66%, with respect to the causes reported by the IES and error margin of 47% in J48 algorithm with a confidence factor of 0.60, 0.75 and 1.0. The implemented method obtains, a satisfactory approach is achieved to address the phenomenon of dropout or failure in the Universities