Abstract:

Academic course scheduling is a complex organizational problem whose solution requires not only the application of appropriate optimization methods but also the ability to assess the level of complexity of the scheduling process in advance. Machine learning methods enable the construction of predictive models to support such analysis; however, their practical application also requires the ability to interpret the resulting predictions. This paper proposes an approach that uses the Random Forest Regression model and the SHAP (SHapley Additive exPlanations) method to predict and interpret the values of a proprietary Scheduling Complexity Index (SCI). The study was conducted using actual operational data covering eight consecutive academic semesters and an activity-level dataset containing 12,740 teaching activities described by 19 organizational characteristics. The coefficient of determination, R² = 0.53, indicates that the analyzed organizational characteristics allow for the prediction of a significant portion of the variance in the SCI index. The SHAP analysis, in turn, made it possible to determine the importance of individual characteristics for the model’s predictions and to identify additional relationships related to the organizational structure of the teaching process. The results indicate that combining machine learning models with explainable artificial intelligence (EAI) methods can support the analysis of the complexity of the lesson planning process and the interpretation of the information used by the model during prediction.