Towards Discriminant Function Analysis based Classification

ABSTRACT The applications of data mining in the field of education are growing fast. This is due to availability of large volume of data and the need of transforming such data into valuable information and knowledge. Educational Data Mining is an area of data mining where lots of researchers are carrying out their work on various issues like enrollment management, performance evaluation, feedback system to support instructors, designing courseware, study of student's learning behavior, and dropout analysis. The proposed research covers the issue of student's dropout in computer science courses of higher education by designing a classification model. This classification model can be used to provide prior information about the dropout so that appropriate steps can be taken for reducing the dropout ratio. Machine learning is one of the disciplines in data mining that plays a key role in predictive data analysis. In this research, we have combined a machine learning method called Classification with the statistical technique known as discriminant function analysis for constructing the classification model. The accuracy of the dropout classifier has been calculated using the measure of classification success. In the last, the results of designed dropout classifier have been compared with the results of Naïve Bayesian classification method. We have used the students' data, collected from computer science courses in a university and SPSS, and WEKA as software tools for the experimental purpose.