Comparison of machine learning classification models for predicting student academic performance

In this study, various models were developed using machine learning algorithms to predict the academic performance of students in the Programming Course at Kırıkkale University. The data, collected through a survey from 170 associate degree students, consists of 9 input attributes, including demographic information, high school education information, academic status, and socio-economic characteristics, and one output attribute. The data were analysed using the WEKA software. According to the results, the J48 algorithm was identified as the most successful model with an accuracy rate of 96.11%. The Random Forest and k-NN algorithms also closely followed J48 with accuracy rates of 95.83% and showed successful results. The Naive Bayes algorithm exhibited the lowest performance with an accuracy rate of 78.06%. In terms of error rates, the Random Forest algorithm showed the best performance with the lowest MAE, RMSE, RAE, and RRSE values. This study demonstrates the applicability of machine learning techniques in education and proves that they can be used as a tool for early detection of at-risk students. In future studies, it is planned to improve the performance of the models and test their generalizability by using larger datasets and different machine learning algorithms.