New hybrid flower pollination algorithm with dragonfly algorithm and jaccard index to enhance solving university course timetable problem

Abstract

University course timetable problem (UCTP) is one of the problems on which many researches have been conducted over the years because of its importance in academic institutions. A nature-inspired metaheuristic optimization algorithm, flower pollination algorithm (FPA) has been adapted, so-called adapted FPA (AFPA), to cope with UCTP in the previous work. However, AFPA suffers from the stagnation problem because of the non-diversity in the population. To improve the diversity of the population, this work introduces new Hybrid FPA with two variants: JFPA provided the Jaccard index to determine similarities among categorical data and the greedy selection mechanism to improve the selection of the random solution, and DFPA applied the navigational characteristics of the dragonfly algorithm (DA) to help in the neighborhood relationship. The results in this study indicate that the proposed algorithms have better exploration ability and fast convergence rate in comparison to previous approaches; JFPA outperforms AFPA in 3 out of 4 datasets for both small and large datasets, and DFPA outperforms AFPA and GA in all datasets while it outperforms PSO in 3 out of 4 small datasets and 2 complex large datasets.