A Case Study of Applying the Classification Task for Students Performance Prediction

Abstract

In this study the application of data mining in an academic database is presented, aiming the identification of the reasons of student dropouts by preventing failures in Programming Language class from Internet Systems course of Federal Institute of Mato Grosso do Sul (IFMS). The classification task is used, as well as decision tree technique and J4.8 algorithm, wich is run with three different options and, in each option pruned and unpruned decision trees are generated. The results show that the most realistic test is Cross-validation, with a success rate of 75.8% in classification, and that it is possible to prevent student failure in specific cases and as a consequence, to lower the number of student dropouts by failure.