An investigation into OR techniques for conference scheduling problems

Pylyavskyy, Yaroslav and Kheiri, Ahmed and Jacko, Peter (2025) An investigation into OR techniques for conference scheduling problems. PhD thesis, Lancaster University.

Text (2025pylyavskyyphd)
2025pylyavskyyphd.pdf - Published Version
Available under License Creative Commons Attribution-NonCommercial-NoDerivs.
Download (1MB)

Abstract

Academic conferences provide great benefits to their participants and stimulate the advancement of knowledge. In the hope of exploiting fully a conference though, an effective schedule is required. Given that many conferences have different constraints and objectives, different mathematical models and heuristic methods have been designed to address rather specific requirements of the conferences being studied per se. The aim of this thesis is the investigation of different operations research tools for the creation of a generic conference scheduler applicable to many conferences. In chapter 3, a penalty system is presented that allows organisers to set up scheduling preferences for tracks and submissions. A generic scheduling tool based on two integer programming models is presented which schedules tracks into sessions and rooms, and submissions into sessions by minimising the penalties subject to certain hard constraints. Then, in chapter 4, a decomposed two-phase matheuristic solution approach is presented as an alternative approach to mathematical models that struggle for some conference scheduling problems. The results showed that the matheuristic finds near-optimal solutions and finds solutions for instances where the mathematical model fails to provide solutions within the one hour time limit. Next, in chapter 5, we make benchmark data publicly available to facilitate the comparison and evaluation of different developed methods for conference scheduling problems. In addition, we present a selection hyper-heuristic algorithm to solve the benchmark instances and provide computational results. The aim is to encourage researchers to contribute to the benchmark dataset with new instances, constraints, and solving methods. In chapter 6, we present extended formulations of mathematical models to handle constraints that need to be resolved on time slot level. Lastly, we compare the performance of all developed methods by solving all available instances and highlight the benefits and limitations of each method.