A bioinformatic variant fruit fly optimizer for tackling optimization problems

The fruit fly optimization algorithm (FOA) is a swarm-based algorithm inspired by fruit flies’ food search behaviors in nature. The conventional FOA is attracting widespread interests due to its briefness, and simplicity in structure. However, FOA still has some disadvantages presented in the exploration and exploitation abilities when it is used to solve different optimization problems. To optimize these drawbacks, the performance FOA can be improved by employing different operators that can help it explore more promising areas in the search space and in finding better solutions in the local area of the candidate optimal solutions. In this paper, an improved FOA approach (called BSSFOA) that employs (1) bat sonar strategy to strengthen the exploration, (2) hybrid distribution that combined Gaussian distribution with student distribution to enhance the exploitation is proposed. In BSSFOA, the FOA uses the bat sonar strategy to search for the global optima, while the hybrid distribution mechanism is used to search the local area of the global optima in the hope of finding better solutions. To assess the performance of the proposed approach, a comprehensive set of 30 benchmark functions was used with the continuous version of the BSSFOA. Moreover, a discrete version of BSSFOA was proposed as a searching mechanism in the feature selection process, where 17 well-known datasets were used to assess the ability of the BSSFOA to search the best performing features among these datasets. The obtained results reveal the superiority of the BSSFOA in solving both continuous and discrete optimization problems. Therefore, it can be concluded that the employed mechanisms have constructive impacts in mitigating the core problems of FOA.