Developing novel meta-heuristic, hyper-heuristic and cooperative search for course timetabling problems

Henry Obit, Joe (2010) Developing novel meta-heuristic, hyper-heuristic and cooperative search for course timetabling problems. PhD thesis, University of Nottingham.

[img]
Preview
PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (14MB) | Preview

Abstract

The research presented in this PhD thesis focuses on the problem of university course timetabling, and examines the various ways in which metaheuristics, hyperheuristics and cooperative heuristic search techniques might be applied to this sort of problem. The university course timetabling problem is an NP-hard and also highly constrained combinatorial problem. Various techniques have been developed in the literature to tackle this problem. The research work presented in this thesis approaches this problem in two stages. For the first stage, the construction of initial solutions or timetables, we propose four hybrid heuristics that combine graph colouring techniques with a well-known local search method, tabu search, to generate initial feasible solutions. Then, in the second stage of the solution process, we explore different methods to improve upon the initial solutions. We investigate techniques such as single-solution metaheuristics, evolutionary algorithms, hyper-heuristics with reinforcement learning, cooperative low-level heuristics and cooperative hyper-heuristics. In the experiments throughout this thesis, we mainly use a popular set of benchmark instances of the university course timetabling problem, proposed by Socha et al. [152], to assess the performance of the methods proposed in this thesis. Then, this research work proposes algorithms for each of the two stages, construction of initial solutions and solution improvement, and analyses the proposed methods in detail.

For the first stage, we examine the performance of the hybrid heuristics on constructing feasible solutions. In our analysis of these algorithms we discovered that these hybrid approaches are capable of generating good quality feasible solutions in reasonable computation time for the 11 benchmark instances of Socha et al. [152]. Just for this first stage, we conducted a second set of experiments, testing the proposed hybrid heuristics on another set of benchmark instances corresponding to the international timetabling competition 2002 [91J. Our hybrid construction heuristics were also capable of producing feasible solutions for the 20 instances of the competition in reasonable computation time. It should be noted however, that most of the research presented here was focused on the 11 problem instances of Socha et al. [152].

For the second stage, we propose new metaheuristic algorithms and cooperative hyper-heuristics, namely a non-linear great deluge algorithm, an evolutionary nonlinear great deluge algorithm (with a number of new specialised evolutionary operators), a hyper-heuristic with a learning mechanism approach, an asynchronous cooperative low-level heuristic and an asynchronous cooperative hyper-heuristic. These two last algorithms were inspired by the particle swarm optimisation technique. Detailed analyses of the proposed algorithms are presented and their relative benefits discussed. Finally, we give our suggestions as to how our best performing algorithms might be modified in order to deal with a wide range of problem domains including more real-world constraints. We also discuss the drawbacks of our algorithms in the final section of this thesis.

Item Type: Thesis (University of Nottingham only) (PhD)
Supervisors: Landa Silva, D.
Qu, R.
Subjects: Q Science > QA Mathematics
Faculties/Schools: UK Campuses > Faculty of Science > School of Computer Science
Item ID: 13581
Depositing User: EP, Services
Date Deposited: 10 Sep 2013 13:46
Last Modified: 16 Dec 2017 10:33
URI: https://eprints.nottingham.ac.uk/id/eprint/13581

Actions (Archive Staff Only)

Edit View Edit View