An Improved Fruit Fly Optimization Its Algorithm And Application Based On Dynamic Linear Step And Double Subgroups

Fruit Fly Optimization Algorithm is an optimization algorithm which is summed up in the food finding behavior of the fruit fly. FOA calculate the smell concentration of each fruit fly location by inputting the smell concentration judgment value into the objective function, and then find out the fruit fly with the maximum smell concentration as the current best one, through continuous iteration to find out the final best one until the end of the iteration-FOA has been extended to different areas and been used widely, such as scientific research industrial designs data mining and ANNS, etc, meanwhile, FOA has achieved good results in the practices of model coefficient adjustments optimization of initial weights of ANNs、resource allocations、 traffic road design and prediction of financial model, etc.The process of FOA is very simple, and the convergence rate is very fast. But in the meantime, the algorithm also has some shortcomings. Firstly, the direction and distance of moving in the process of food finding is randomly, the only influence factor is the step size. But the fixed step size restricts the algorithm performance, it’s difficult to balance the global and local performance of the algorithm. Secondly, it’s easy to fall into local optimum situation by the high dimension problem. In order to solve these problems, some works done were as follows:1、This paper proposes a new optimization algorithm based on dynamic linear step and double subgroups, it improved the FOA in two ways. Firstly, using a dynamic linear step to control the size of the search space, the step is determined by the weight parameter and the number of iterations. In the early stage, it searches in a greater scope to quickly locate the optimal value. In the later stage, it searches in a smaller scope to find the optimal value quickly. It could balance the global and local performance by this way. Secondly, using double subgroups to calculate alternately, these two subgroups have opposite step size, so as to solve the problem of easy to fall into local optimal situation.2> Using 6 classical Benchmark test functions to test the performance of the LD-FOA, By comparing with PSO, DE and LGMS-FOA, it’s proved that LD-FOA is superior to the other three algorithms in solving precision, convergence rate and stability.3> Using the LD-FOA to solve two NP-hard problems, include 0-1 Knapsack Problem and TSP. Firstly, using ten classic 0-1 Knapsack Problems to test the performance of the LD-FOA. The dimensions of these problems are from 10 to 100, so they could test the performance of the LD-FOA better. Secondly, it explored the application of the algorithm in TSP, and its performance is analyzed.A summary of all the work is done at the-end ofhe paper, and points out the advantages and disadvantages of the work. Then it summarized a few directions that are worth examining according to some other papers.