Expanding tree-based classifiers using meta-algorithm approach: An application for identifying students’ cognitive level

Accurate identification of student cognitive levels is a crucial problem for a teacher in deciding the appropriate method for a teaching and learning process. Nevertheless, not much research focuses on this area. Therefore, in this paper, we investigate the problem of how to improve the classification performance to discover the more suitable students’ cognitive level. We expand tree-based classifiers using a meta-algorithm called “LogitBoost” in the mining process. Then, to support this meta-algorithm to work optimally, we introduce the multivariate normality test and the combination of the discretization method and k-NN on the pre-processing stage. These designed schemes are intended to find the student data normality and to specify the number of the students’ cognitive levels. Also, we propose a feature selection approach: correlation-and relief-based feature selection to eliminate unnecessary features. The experimental results show that our proposed method can enhance the classification performance in the identification process significantly.