THE STUDENT PERFORMANCE PREDICTION SYSTEM USING MACHINE LEARNING

Computer Science Engineering

The major task of any learning institution is to offer the student the best learning experience and knowledge needed to develop him/her. A major move toward this objective is identifying learners in need of extra help and setting measures to enhance their performance in schools. The paper is a machine learning experiment involving four techniques aimed at developing a predictive model to reflect show the expected level of student performance in the course projects in computer science taught at the Al-Muthanna University, College of Humanities. The methods discussed are Artificial Neural Networks (ANN), Naive bayes, Decision Trees, and Logistic Regression. In this work, there is a certain emphasis on two behavioral factors namely the effect of internet use as an academic resource and the effects of time spent on social networking platforms on educational success. These points are recorded in terms of certain factors that show either the dependency of students on the internet as the time-spender and courses provider, or their excessive consumption of social media. The evaluation of the models was carried out by employing the ROC index and overall classification accuracy in comparison to other measures like the error rate, precision recall and F-measure. The data driving this study was obtained by using surveys of students along with formal grade records. The most effective of the four models was the feed-forward multilayer ANN with ROC value of 0.807 and an accuracy of 77.04. Moreover, the Decision Tree(DT) model identified five important variables used to greatly determine student performance..

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.