Examination timetabling problem

Abstract

Examination timetabling is a major challenge in most educational institutions. Since the underlying problem is NP-hard and real-life problems are too hard to solve to optimality, heuristic approaches are adopted as solution methodologies in general. The significance of a fair exam schedule creates a need for exact solutions to the examination timetabling problem. In this thesis, we mainly focus on exact solution approaches for this problem and test their efficacy on well-known benchmark problems from the literature as well as on Bilkent University’s data. Existing formulations used for the p-hub median hub location problem and the quadratic assignment problem in the literature are adapted to the examination timetabling problem and various Bender’s decomposition and branch and cut methodologies are tailored to these formulations. A novel compact formulation based on individual student schedules with reduced model dimensions is proposed. For the literature instances in which optimal values are not known, we could find effective lower bounds. For higher dimensions, we propose matheuristic approaches based on our proposed formulations. With this study, effective lower bounds are found for unsolved problems and small-scale problems are solved to optimality in short computational times.