A fruit fly optimization algorithm with a traction mechanism and its applications

The original fruit fly optimization algorithm, as well as some of its improved versions, may fail to find the function extremum when it falls far from the origin point or in the negative range. To address this problem, in this article, we propose a new fruit fly optimization algorithm, named as the traction fruit fly optimization algorithm, which is mainly based on the combination of “traction population” and dynamic search radius. In traction fruit fly optimization algorithm, traction population consists of the worst individual recorded in the iterative process, the individual in the center of the interval, and the best fruit flies individual through different transformations, which is used to avoid the algorithm stopping at a local optimal. Moreover, our dynamic search radius strategy will ensure a wide search range in the early stage and enhance the local search capability in the latter part of the algorithm. Extensive experiment results show that traction fruit fly optimization algorithm is superior to fruit fly optimization algorithm and its other improved versions in the optimization of extreme values of continuous functions. In addition, through solving the service composition optimization problem, we prove that traction fruit fly optimization algorithm can also obtain a better performance in the discrete environment.