What factors most significantly predict student satisfaction with K-MOOC? A machine learning approach

The purpose of this study is to develop a model that can predict the satisfaction of K-MOOC students as well as prioritize the factors that significantly impact this prediction. To this end, we used the satisfaction response data of a total of 78,893 K-MOOC students in 2021, and we processed the data using scikit-learn, a package in Python, to train a random forest model. As a result of developing a prediction model, the AUC value was 0.98, accuracy was 0.92, the F1 score was 0.85, specificity was 0.92, and sensitivity was 0.92, thus showing excellent performance overall. Further, the importance ranking of the variables that predicted student satisfaction was as follows: The first variable was content quality (practical content and effective learning materials), followed in order by instructor's knowledge and skills (expertise and teaching strategies), system quality (convenience, ease of access to information), and service quality (responsiveness and speed of operation team) as the variables that were identified as important. Based on the results of this study, detailed recommendations were made for the quality management of K-MOOC.