Evaluation of learning effectiveness during and after COVID-19 pandemic using decision tree and Naïve Bayes

This study aims to evaluate the effectiveness of learning during and after the COVID-19 pandemic using data mining techniques by applying the Decision Tree algorithm that can handle data with complex features and the Naive Bayes algorithm applied to small data. Both algorithms were applied to Moodle data from the 2020 and 2023 cohorts, data obtained from the Moodle platform. The data include activity logs, attendance records, assignment grades, midterm exam grades, and final exam grades, with the final grade converted into graduation status serving as the target variable. For the 2020 cohort, the Decision Tree model has 100% accuracy, while for the 2023 cohort, the accuracy of the Naive Bayes model performs better with 92% accuracy. Feature importance analysis revealed that assignments, midterm exams, and final exams were most important for predicting student outcomes in the 2020 cohort, while in 2023 it was assignments, midterm exams, final exams, and attendance. Notably, Moodle activity data did not significantly affect the prediction, indicating that in-person assessments and attendance were more important for academic success. Insights from this study guide that academic assessments such as assignments, midterms, and final exams have the greatest impact on student graduation, both in the 2020 and 2023 cohorts. The presence of attendance indicates that students’ active engagement in each lecture session plays a significant role in determining graduation. Therefore, instructors need to pay more attention and emphasize factors such as assignments, exams, and attendance in their teaching strategies to improve student learning outcomes.