Adaptive Neighbourhoods Structure Selection Mechanism in Simulated Annealing for Solving University Course Timetabling Problems

Usually, meta-heuristic approaches that use several neighbourhood structures can perform better than single neighbourhood structure. However, choosing a suitable neighbourhood structure to be applied during the search process is also a crucial decision. Therefore, this study proposes an adaptive neighbourhoods structure selection (AD-NS) mechanism, that adaptively memorised the improvement strengths for each neighbourhood structure. The neighbourhood structure with the best improvement history will be employed to generate neighbour(s) for the current iteration. Results based on the average ranked, shows that, Simulated Annealing (SA) with AD-NS approach obtained the fourth rank compared with other approaches reported in the literature. Statistical analysis on SA with AD-NS against SA with other neighbourhood structure selection mechanisms proved that, the performance of SA with AD-NS is significantly better than SA with other neighbourhood structures selection mechanisms tested in this work. This indicates that, the improvement history of neighbourhood structure can affect the performance of neighbourhood structures selection mechanism and subsequently, the performance the applied meta-heuristic.