Improved Fruit Fly Optimization Algorithms Based On Group Cooperation And Implementations

Fruit fly optimization algorithm(FOA)is a kind of newly swarm-based evolutionary algorithm proposed by Pan in 2011 for global optimization,which is inspired by the food finding behavior of the fruit flies.Compared with other existing swarm intelligence algorithms,FOA possesses the advantages of few parameters,simple structure,easy to understand and implementation.However,as a kind of meta-heuristic approach,FOA has the same shortcomings as other evolutionary algorithms,which can easily get into the local optimal,occur premature in optimization of multi-modal problems,etc.In order to enhance the accuracy of solution of original FOA,this paper presents two improved fruit fly optimization algorithm and applied them to web service composition and multilevel image thresholding respectively:(1)As the number of services multiplying rapidly,multiple web services possess similar functional attributes.From these composite services how to select a series of candidates services that meet the best requirements of users has already begun to attract a lot of researchers' attentions.Many techniques for measuring service composition in terms of Quality of Service(QoS)have been proposed.However,most of the available methods more or less have their own drawbacks such as poor scalability,high system overhead and finding the global optimum hardly.Accordingly,we present an enhanced fruit fly optimization algorithm with novel search strategy(NSSFOA)to resolve these difficults,which improves the global search strategy on account of the current global optimum,by separating the fruit fly group into two sub-swarms with different evolutionary iterative step.Simulated experiments show that the algorithm is superior to particle swarm optimization and differential evolution algorithms in terms of feasibility,stability and solution quality.These results also demonstrate that the performance of NSSFOA is much better than other algorithms with the increment of candidate services.(2)Multilevel thresholding is widely exploited in image processing,but most of the methods are time-consuming.In this paper,we present a novel approach,multilevel thresholding with fruit fly optimization algorithm.As yet,FOA has not been applied to resolve the complex image processing problems.Thus,we introduce it into the study of multi-threshold image processing area.Moreover,we incorporate a hybrid adaptive-cooperative learning strategy with the-proposed method,called HACLFOA.The fiuit fly population is divided into two sub-populations and both of them have different iteration step range.In addition,each dimension of the solution vector will be optimized during one search,and we also make the best of the temporary global optimum information.The results of computational experiments on 24 benchmark functions demonstrate that the proposed algorithm has superior global convergence ability against other algorithms.Most significantly,extensive promising results show that the proposed algorithm is time-saving in multilevel image thresholding,and it has huge potential in the image processing field.