Abstract:

Academic scheduling requires not only effective optimization methods but also high-quality data describing the planning process. Previous research has focused primarily on scheduling algorithms, while the process of integrating, validating, and preparing operational data remains much less well-documented. In response to this identified research gap, a multi-level data engineering framework was developed to enable the automatic transformation of heterogeneous data from the dean’s office system and the FET program’s scheduling models into a coherent representation of the planning process. The framework encompasses data integration, validation, and multi-level feature engineering, resulting in an activity-level dataset that describes each teaching activity along with its organizational context. The solution was validated using real-world data spanning eight consecutive academic semesters, yielding a dataset containing 12,740 teaching activities described by 19 organizational features. The results confirm the possibility of automatically generating a multilevel data representation that can serve as a universal data preparation layer for analyses of the planning process, decision support systems, and future research utilizing data analysis and machine learning methods.