Academic performance prediction using Machine Learning algorithms

Abstract

An excellent secondary school education becomes evident in students’ performance after they graduate or further their education. No matter their career choice, they can genuinely excel if they can identify areas that require them to put in more effort to have an overall excellent performance in school. In Nigeria, several solutions address students' learning needs or the administrative needs of the schools. Still, no systems cater to analysing and monitoring students’ performance, causing failures that can be averted. This dissertation reviews five different machine learning algorithms using data from students in public and private schools in rural and urban Nigeria to identify which algorithm performs best in predicting students’ performance using the Waikato Environment for Knowledge Analysis tool for modelling and the cross-industry standard process for data mining (CRISP-DM) research methodology. The result shows the Decision Tree as the algorithm with the best performance for the dataset. It is recommended that the findings be used to build a system embedded into a school management or learning management software to enable students, parents, and teachers to channel the right resources into areas where it has been predicted that the student will underperform to change the narrative.