Comparasion of Feature Selection Techniques in different Classifiers for predicting Students’ Academic Performance

Abstract

Teaching evaluation is one of the most vital needs that have to be carried out to determine the quality of education in every academic institution. Higher education is mostly assessed by using the students’ grades achieved in the examination. A new emerging research area which provides educational organizations to predict the performance of their students is Educational Data Mining (EDM). In EDM, Feature Selection (FS) is the important task applied in getting better the quality of prediction model for educational data sets. FS algorithm removes the irrelevant or redundant features from the educational repositories and influences the accuracy of classifiers used in various EDM practices. In order to implement the prediction model with better accuracy, it is necessary to apply the good quality of educational data set with the most relevant features. Therefore, the decision depends on suchlike good quality data set can be able to enhance the educational quality by predicting the students’ performance. So, it is indispensable to choose carefully which feature selection algorithm is more appropriate with the classifier. For this purpose, this system is implemented to make comparisons between five feature selection techniques and their impact on the fifteen classifiers on the Students’ Performance Data set from University of Jordan using WEKA tool. The results of present comparative study effectively support the educational area to recommend the best combination of feature selection method and classification algorithm.