Software system architecture for anthropocentric scheduling of educational workload in higher education institutions

• Дата: 2026
• Автор: Sytnik, O.O.
• Формат: Стаття
• Мова: Українська
• Опубліковано: PROBLEMS IN PROGRAMMING 2026
• Теми: software architecture; modular monolith; anthropocentric scheduling; evolutionary algorithm; chronotype; genetic operators; adaptive learning; technical higher education institution; UDC 004.4’22:378.1
• Онлайн доступ: https://pp.isofts.kiev.ua/index.php/ojs1/article/view/1024
• Назва журналу: Problems in programming

The article examines the architecture of a software system designed for anthropocentric dispatching of academic workload in technical higher education institutions. The system is implemented as a modular monolith on the .NET platform running under Windows Server. The software complex consists of four distinct functional mod ules: data collection and storage, the optimization core, adaptive learning, and result presentation to the user. Inter-module communication is carried out through an internal event processing subsystem and via asynchronous calls through dependency injection container interfaces. The optimization core implements a hybrid evolutionary dispatching algorithm with two specialized genetic operators. The first of them, the chronotype-preserving crossover operator, selects cut points proportional to the difference in circadian contributions between the parent individuals. The next step is the anthropocentric muta tion operator with weighted slot selection, which allocates the mutation budget to the lowest-quality slots. The weight coefficients of the multi-criteria objective function are adapted between semesters through a satisfaction regression mechanism. This process uses the least squares method with participant feedback applied. To obtain more accurate results, the exponential moving average method is used without manual parameter tuning. The implementation details of the software using this algorithm are provided below. The algorithm's comparative analysis was conducted on test data from a higher education institution, comprising 400 students, 60 instructors, and 42 disciplines. Across 30 independent runs, a 34% reduction in algorithm con vergence time and an 18.3% improvement in schedule quality were confirmed compared to a conventional ge netic algorithm. The satisfaction regression mechanism across three adaptation cycles increases the correlation between the weight coefficients and the actual participant priorities from 0.61 to 0.89.Problems in programming 2026; 2: 37-48