A comparative machine learning analysis for forecasting university satisfaction scores in Türkiye using TUMA, YÖK, and URAP data (2016–2025)

Özet

Assessing university satisfaction levels is crucial for achieving expected performance from universities. In this regard, this study investigates the forecasting of satisfaction scores at universities in Türkiye. In the application stage, Turkiye University Satisfaction Survey (TUMA), Council of Higher Education (YÖK) Statistics, and University Ranking by Academic Performance (URAP) data were employed. This data covers the period between 2016-2025. Linear Regression, Random Forest, Gradient Boosting, Tree, Neural Network, and Support Vector Machine (SVM) machine learning algorithms were applied to predict satisfaction scores. The trials for each model were carried out using Orange version 3.39 on samples taken according to different satisfaction levels and university types. After that, the model performances were compared and the best trial results show that model performances differ according to institution and satisfaction level type. Thus, it was concluded that institutions need to use different models for their satisfaction score predictions, depending on the level of satisfaction and the type of university.