Application And Research On Flower Pollination Algorithm

Flower pollination algorithm(FPA)is a swarm intelligent optimization algorithm which is inspired by the phenomenon of flower pollination in nature.The algorithm mainly includes the self-pollination process and the cross-pollination process.These two processes correspond to the local optimization and global optimization of the optimization algorithm,and the balance of the two processes is controlled by the switching probability,which can coordinate the local mining performance and the global exploration performance of the algorithm.The algorithm has the advantages of simple frame,high search efficiency,strong robustness and easy implementation,so it has been widely used in the fields of feature selection,function optimization,engineering optimization and large integer programming,and has achieved remarkable results.However,as a new efficient element heuristic algorithm,flower pollination algorithm is still in the initial stage in the research of discrete optimization problem,so it has important practical significance and application value to develop the algorithm and expand its application scope.Based on the biological mechanism and the evolutionary characteristics of flower pollination algorithm,the following two works are carried out.(1)To seek a stable and efficient method for solving Multidimensional Knapsack Problem(MKP),we propose a solution method based on the Flower Pollination Algorithm(FPA),here named the FPA-MKP.In the FPA-MKP algorithm,each pollen corresponds to a solution of the MKP,which can be represented in two forms: the real vector form and the binary vector form.First,the population is initialized by means of chaos;then,according to the switching probability,the real vector form of the offspring is obtained by using the method of self-pollination or cross-pollination,and is mapped into a binary form.And then,combined with the value density of goods,a greedy stochastic correction strategy in the process of evolution is proposed to optimize the individual solutions,while the reverse learning mechanism is used to synchronize the two forms of the individual solutions.Finally,compared with the other three element heuristic algorithm in a large number of standard data sets,FPA-MKP algorithm shows a stronger competitiveness whether in the quality of the solution or the convergence performance.(2)In order to identify the function modules of the protein network with higher quality,a method for detecting protein interaction network function modules based on flower pollination algorithm was proposed which uses the optimization mechanism of flower pollination algorithm.Firstly,each pollen is encoded by a random walk.Then,the population is optimized by using self-pollination and cross-pollination mechanisms in flower pollination algorithm.More specially,the strategies of recombination and better-solution selection are adopted in the self-pollination while the mutation strategies based on levy mechanism and an adaptive individual-difference are employed in the cross-pollination.The four strategies together promote the evolution of the population from different angles.Finally,the simulation experiments on three public data sets show that the proposed algorithm has not only excellent overall performance but also absolute superiority in terms of two comprehensive indicators F-measure and accuracy compared with the other six classical algorithms.