Analysis of university student satisfaction using data mining

Student satisfaction is an indicator for evaluating the quality of the university experience and guiding institutional improvement efforts. This study aimed to predict university student satisfaction using data mining techniques and to identify the factors that contributed most to its classification. A total of 232 students participated, and their data were analyzed following the CRISP-DM framework. Various supervised learning algorithms were compared, and interpretation techniques such as SHAP and sensitivity analyses were employed. The support vector machine (SVM) model performed best, with an accuracy of 0.8413, a macro-F1 score of 0.7797, and a balanced accuracy of 0.7459 for distinguishing low, medium, and high levels of satisfaction. Teaching quality was the most significant contributing factor (38.3%), followed by institutional image (18.2%), skills development (13.9%), and support services (12.8%). In particular, methodological strategies, subject matter expertise, content assessment, and confidence in the training played a significant role. These results suggest that satisfaction is associated with the academic experience and the student’s connection with the institution. Data mining provides actionable evidence to improve teaching and university services and to guide decision-making toward more relevant training that better meets students’ needs.