Research On Improved Biology Migration Algorithm And Its Application

Swarm intelligence optimization algorithm is an indispensable part in the field of artificial intelligence.It solves complex optimization problems by simulating the mechanism of natural ecosystem.Swarm intelligence optimization algorithm is widely used in combinatorial optimization,parameter estimation,function optimization,path planning,neural network training,graphics and image processing because of its simple principle,few adjustment parameters and easy programming.Biology migration algorithm (BMA)is a new meta-heuristic optimization algorithm inspired by the phenomenon of biological migration.The simulation results on the standard test function show that the proposed algorithm has better convergence performance than some classical intelligent algorithms.However,there are still problems of premature convergence and imbalance of exploration and exploitation.This makes it not a good solution to some complex optimization problems.At the same time,with the continuous increase in the dimension of practical optimization problems,more and more scientists use robust swarm intelligence optimization algorithms to solve optimization problems,especially the essential feature selection part in the process of machine learning.In view of the above situation,this dissertation does the following work:(1) In this thesis,an enhanced BMA,namely CLCBMA is proposed to improve the convergence performance of BMA.First of all,in order to enhance population diversity,avoid premature local optimization,and better balance exploration and development,a comprehensive learning strategy was added to the original BMA.Then,chaos theory is added to increase the ergodicity of search.Finally,an optimal alternative migration update strategy based on personal history is used to accelerate the convergence speed.48 benchmark functions and two engineering design problems validate the effectiveness of the algorithm.Experimental results and statistical analysis show that the algorithm has higher performance than the other eight meta-heuristic algorithms.(2) In order to verify the applicability of CLCBMA,a binary CLCBMA (BCLCBMA)is proposed on the basis of CLCBMA,and it is applied to feature selection problem.The validity of BCLCBMA was verified through 15 benchmark data sets in UCI database.Finally,it was compared with other 8 meta-heuristic algorithms in four performance indexes,namely the maximum value (Max),the Mean value (Mean),the standard deviation(Std) and the average feature selection size (ASS).Experimental results and statistical analysis show that the BCLCBMA can find a better subset of features,maximize the classification performance and minimize the number of selected features.