Research And Application Of Fruit Fly Optimization Algorithm Based On Group Cooperation And Periodic

As a heuristic random algorithm,swarm intelligence algorithm is always the most popular algorithm for researchers.These algorithm has wide application.The target object isn't require special structure.It's only need to provide a small amount of information and the target objects.This kind of algorithm has simple implementation process.In this thesis,we study the Fruit Fly Optimization Algorithm(FOA).It is a kind of swarm intelligence algorithm and bionic intelligent algorithm.The FOA concluded from the simulation of the fruit files foraging behavior.For now,the research and application of FOA is still in infancy.The depth and breadth in the field of this theoretical research are not as good as other swarm intelligence algorithm.Even so,FOA still made some good achievement.For example,such as the enterprise performance evaluation in the economic filed,the path planning in the transportation filed,the scientific theory research and so on.There are many advantage of the FOA.For example,the FOA is easy to understand and program and it has good global search capability and fast convergence speed.However,the FOA also have some obviously limitations.First of all,in the process of iteration of the FOA,the moving distance of files in the feasible region and direction are random.It may cause poor stability of the algorithm and the optimization accuracy is not high;For another,the algorithm is not very good at using the flies group historical information.It lead to the efficiency of the algorithm is relatively low and large probability into local extreme point.Therefore,an important research content of this article is aim at the limitation of the algorithm,we design a FOA which has more superior performance,and also use this kind of improved FOA in the preliminary application research of the cloud workflow task scheduling.With the rise of cloud computing industry and flourish,as the core of cloud computing----the research on cloud workflow task scheduling is more and more important.The task scheduling problem is the classical NP-complete problems.Intelligent algorithm has always been the first selection scheme in this study area.Among them,the most studied algorithm are ant colony algorithm and particle swarm optimization algorithm.After many years of research,ant colony algorithm and particle swarm algorithm is tending toward perfection.The room for improvement in performance is more and more small.Therefore,it is badly in need of theoretical breakthrough in terms of scheduling algorithm.And to seek a new algorithm for scheduling policies provide other improvement ideas.To be specific,in this thesis,the main work content is summarized as followings:1 According to the disadvantage of the FOA,the thesis proposed fruit fly optimization algorithm based on group cooperation and periodic decay strategy.The proposed algorithm improved the fruit flies algorithm in three aspects:Firstly,this thesis uses the group cooperation way,grouped the fruit flies according to the difference fruit flies ability,it improved the diversity of the population and the stability of the algorithm.Secondly,this thesis improved the fruit flies information update policy,from the best individual offspring flies decided the search domain become the one kind optimal flies group of offspring determined the search domain.This can increase the efficiency of fruit flies group.Thirdly,this thesis use the step function which has the diminishing with periodic oscillation properties to guide flies group to flight.This function makes the fruit fly can be wider in the local search and more detailed in local domain search,it improved the accuracy of the optimization.2 In order to verify the improved the performance of the fruit flies algorithm,this thesis selects the representative BenchMark functions as test case,and compared with other swarm intelligence algorithm,comprehensive assessment properties of the improved algorithm.3 Using the improved algorithm this thesis proposed in the field of cloud workflow task scheduling,meanwhile,using the amazon Elastic Compute Cloud(EC2)valuation model,transform the multi-objective optimization problem such as cost of Quality of Service(QoS)constraints,time constraints,personalized needs from users into a new sign objective optimization problem which can multiple perspectives reflect the advantage and disadvantages of scheduling policy.And through to rethink the Directed Acyclic Graph(DAG)task graph model,it can better applies to the improved algorithm.