<![CDATA[The purpose of this study was to develop and integrate a student academic performance prediction system into an e-learning platform using a mathematical modelling approach combined with machine learning algorithms. The method employed was Research and Development (R&D), encompassing stages of needs analysis, mathematical modelling, development of a machine learning-based prediction system, and implementation and evaluation. The study was conducted at Duta Bangsa University, Surakarta, involving 100 students from the Informatics Engineering study program. Data were collected through the e-learning platform, covering student activity logs such as access frequency, quiz scores, assignment completion time, and forum participation. This behavioral data was then analyzed using supervised learning algorithms, namely logistic regression and decision tree, to build a predictive model for academic performance. The resulting predictive system was integrated into the e-learning platform to deliver risk notifications and adaptive learning material recommendations automatically. To measure the improvement in academic performance, a validated academic achievement test was administered as both a pre-test and a post-test to the experimental group. This test consisted of multiple-choice and short-answer items aligned with the course learning objectives. The results showed that the decision tree model achieved a prediction accuracy of 87.4%, while logistic regression reached 81.2%. Evaluation of the system’s effectiveness using the pre-test and post-test scores revealed a significant increase in students’ academic performance. Statistical analysis with a paired t-test yielded a significance level of p < 0.001, indicating that the adaptive prediction system effectively supports more personalized and impactful learning. This study contributes to the advancement of machine learning-based prediction systems in e-learning by designing and implementing a model that leverages real student activity data. The system enables early detection of academic risks and provides automated, adaptive content recommendations, thus fostering personalized and data-driven learning in higher education. Its practical implementation helps students identify learning weaknesses promptly and receive appropriate supporting materials immediately, promoting proactive and self-regulated learning behavior. ]]>
