Prediction of Student Academic Performance using ML

Abstract

It is now possible to forecast student academic success and explore the underlying links in educational data using educational data mining. This study proposes a new model that uses algorithms for machine learning, using the outcomes of their midterm exam scores as the primary data, to estimate university students' final test marks. The predictive capabilities of the supervised classification methods random forests, nearest neighbor, vector support machines, regression models, Naive Bayes, & k-nearest neighbor were calculated and compared in order to forecast the students' final exam marks. The academic performance ratings of 1854 students who registered in Turkish Language-I at a public Turkish college or university for the fall 2019–2020 academic year were included in the dataset. According to the findings, the proposed model has a system of classification that between 70 and 75 percent. Only three distinct argument types were used to make the predictions: faculty data, departmental data, and midterm test grades. These data-driven studies play a critical role in developing a framework for learning analysis in graduate school influencing the decision-making processes. The study also identifies the most efficient machine learning techniques and contributes to the early prognosis of pupils who are likely to fail.

Keywords: Machine learning, Predicting achievement, Educational data mining, Early warning systems, Learning analytics, Early warning systems.