Grouping and Heuristics for A Multi-stage Class Timetabling System

We investigate a real-world college class timetabling problem, which entails scheduling a set of class groups into a predefined timeslot zone by zone, followed by appropriate rooms. The essence of this problem is that manual timetabling by human expertise is required due to high conflict density, which means that more classes are conflicting due to the enrolment of students in overlapping semesters with different start dates. To tackle this problem, we developed a scheduler system using a multi-stage approach incorporating heuristics and grouping to produce computationally inexpensive solutions with minimal effect while reducing the involvement of human expertise. The first phase generates a feasible solution, whereas the second phase is executed for improvement. The rationale for grouping is to decompose the problem and reduce the amount of conflict. The experimental result demonstrates that our system is able to produce a promising solution that can replace the current manual timetabling.