Modelling Academic Dropout In Computer Engineering Using Articial Neural Networks

School dropout in higher education is an academic, economic, political and social problem, which has a great impact and is difcult to resolve. In order to mitigate this problem, this paper proposes a predictive model of classication, based on articial neural networks, which allows the prediction, at the end of the rst school year, of the propensity that the informatics engineering students of a polytechnic institute in the interior of the country have for dropout. A diferentiating aspect of this study is that it considers the classications obtained in the course units of the rst academic year as potential predictors of dropout. A new approach in the process of selecting the factors that foreshadow the dropout allowed isolating 12 explanatory variables, which guaranteed a good predictive capacity of the model (AUC=78:5%). These variables reveal fundamental aspects for the adoption of management strategies that may be more assertive in the combat to academic dropout.