A hybrid harmony search for university course timetabling

Combinations of metaheuristics to solve highly constrained combinatorial optimisation problems have recently become a promising research area. Some combinations emerge from integrating evolutionary algorithms and local based algorithms to strike the right balance between global exploration of the search region and local exploitation of promising regions. This study presents hybridization between harmony search algorithms and hill climbing optimisers to tackle the university course timetabling problem. The solution is essentially a harmony search algorithm. The hill climbing optimiser is responsible for improving the new harmony vector, obtained from harmony search, with a certain probability. Furthermore, inspired by the practical swarm optimisation, we modified the memory consideration operator to let it improvise rather than always working in the same way. Our algorithm converges significantly faster than does the classical harmony search algorithm. We evaluate our hybrid algorithm with a standard benchmark introduced for UCTP. Our results show that our algorithm can find higher quality solutions than can previous work.