Analysis of classifiers in a predictive model of academic success or failure for institutional and trace data

This Research Full Paper presents recent research on the Educational Data Mining (EDM) field. In recent years, EDM has contributed significantly to the prevention of various challenges in academia. This paper presents an analysis of classifiers for a comparative study of EDM impact, using institutional data and trace data generated by a virtual learning environment to predict academic success/failure. For this, a model of educational data mining using logistic regression, support vector machine, naive bayes and J48 as classifiers and cross validation as a test method, was elaborated and was used to compare the prediction accuracy and the execution time of each classifier. The model was applied on a public dataset with 32,593 students, distributed among seven courses. The results on accuracy and execution time of each classifier allowed us to make recommendations on the suitability of using them. The results also revealed that it is better to separate the trace data from the institutional data in the model application regardless of the classifier. (Abstract)