A comparison of feature selection approach between greedy, IG-ratio, Chi-square, and mRMR in educational mining

Abstract

Educational data mining is a widely interesting issue in data mining research field. One of the topics is feature selection method to reduce a feature set. The main purpose of this study is to compare feature selection methods for the efficiency of student performance prediction improvement. In this research, we proposed 4 feature selection methods: greedy algorithm, Information gain ratio, chi-square, and mRMR that combine with 4 classification models. The example data were 6,882 engineering students in Rajamangala University of Technology Thanyaburi, Thailand from year 2004 to 2010. The experiments demonstrate the effectiveness of the feature selection method in classification of student performance prediction. The result shows that greedy forward selection with neural network classification model presents the best efficiency couple with 91.16% accuracy.