Abstract
Dwarf mongoose optimization algorithm (DMOA) is a recently proposed meta-heuristics, it attracts widely attention due to its effectiveness in solving complex optimization. However, DMOA utilizes roulette wheel selection to evolve the individual, this method may cause the swarm evolving unevenly and the swarm diversity losing rapidly. To overcome the aforementioned weakness of DMOA, this study proposes a two-stage structured elite leader dwarf mongoose optimization algorithm (EL-DMOA). EL-DMOA employs four strategies to improve the performance of DMOA. In the leader stage, The artificial fitness is employed for selecting swarm leader according to the individual's fitness and state, then the differential operator is adopted to further evolve the selected swarm leaders. In the follower stage, the elite leaders are employed to guide the swarm moving towards the promising area. The crossover operation is employed to enhance swarm diversity and reduce the risk of falling into local optima. The experiments on CEC2017 test suite and real-life application problems show that EL-DMOA performs better than FIPS, DE/rand/1 and four recently proposed meta-heuristics. Employing differential operator to evolve the selected swarm leader can improve the quality of swarm leaders. The proposed two-stage structure can encourage the swarm evolves evenly and efficiently.