Predicting students performance in final examination using linear regression and multilayer perceptron

Currently, many educational institutions are highly oriented to improve the quality of education and students? learning achievement-examination result. To fulfil such intention, predicting students? performance by analyzing their learning behavior is one of the best way can be taken into account. Once the performance was predicted, it will be easy for teachers, school authority or other related parties to determine the appropriate policies on the issue. Relatedly, this paper aimed to provide the prediction of students? performance in final examination by applying linear regression and multilayer perceptron in WEKA- in terms of accuracy, performance and error rate- to compare their feasibility. The basis of data was derived from extraction and analysis of e-learning logged-post in discussion forum and attendance. Based on the result, it has been concluded that multilayer perceptron provides better prediction results of final examination than linear regression.