Research On Cancer Microarray Data Classification Algorithm Based On Improved Artificial Fish Swarm Algorithm

As a widespread disease,cancer has been seriously affecting human health and life.Since the birth of gene chip technology,a large number of cancer microarray data have been accumulated.Discovering the useful information of these data is conducive to the diagnosis and typing of cancer at the molecular level,and is of great significance to the treatment mechanism,drug development and treatment of cancer.In general microarray data,there are a lot of noise and redundant data,and the cost of direct processing is very high.Therefore,how to avoid "dimension disaster" in microarray data has become a key problem in microarray data processing.As a method of reducing data dimension in data mining,feature selection(FS)can provide an effective scheme for microarray data processing.In feature selection,wrapper method can obtain the optimal subset according to the classification accuracy of the subset.This kind of method has attracted extensive attention of scholars because of its high flexibility and accuracy.Among them,the application of meta heuristic algorithm in this field has been widely recognized.The continuous improvement of this method is of great significance to feature selection.As a classical meta heuristic algorithm,fish swarm algorithm has been used in load optimization,path optimization,image quantization,traffic control and so on.This paper will improve the fish swarm algorithm so that it can be better applied in feature selection.The main research results are as follows.(1)Aiming at the problems that artificial fish swarm algorithm(AFSA)can not perfectly balance local optimization and global optimization,and lacks the ability to jump out of local optimization,an adaptive AFSA utilizing gene exchange(AAFSA-GE)based on gene exchange is proposed.Firstly,the adaptive field of view and step size are used to improve the speed and accuracy of search,and then the chaotic behavior and gene exchange behavior are used to enhance the ability to jump out of local optimization and improve the search efficiency.In order to prove the effectiveness of the algorithm,10 classical test functions are used in the experiment.The proposed algorithm is compared with the normalized fish swarm algorithm(NFSA),the FSA optimized by PSO algorithm with extended memory(PSOEM-FSA)The comprehensive improvement of artificial fish swarm algorithm(CIAFSA)is compared.The experimental results show that AAFSAGE has better local optimization ability and global optimization ability than PSOEM-FSA and ciafsa,and higher search efficiency and better global optimization ability than NFSA.(2)Aiming at the problem of feature selection,based on the single target fish swarm algorithm,an improved binary fish swarm algorithm,fish swarm algorithm based on population evolution(FSA-PE)is proposed.Firstly,the artificial fish swarm algorithm is improved through the population evolution mechanism to obtain the ability to jump out of the local optimum while preserving the excellent characteristics of the subset as much as possible.Then,the adaptive step size and vision are used to adjust the search space and moving range of the algorithm in different environments to improve the local optimization and global optimization ability of the algorithm.Finally,adjusting the search mode of foraging behavior enables individuals to move to excellent positions all the time,speeding up the convergence speed and searching efficiency of the algorithm.The algorithm mainly strengthens the search ability of weak correlation features in the solution space through the population evolution mechanism to help the algorithm improve the search accuracy.(3)Aiming at the classification and gene selection of cancer microarray data,a feature selection model based on FSA-PE is proposed.First,we use the F-score algorithm to eliminate a large number of useless and redundant features in the data set.Then,the subset is obtained by FSA-PE algorithm,and the subset is evaluated by navie bayes(NB)classifier.Finally,through the experiments on 8 different cancer microarray data sets,the iterative graph is used to prove the improvement of FSA-PE and the effectiveness of the model.More than 90% classification accuracy is achieved in all data sets.Compared with other 15 classification models,the proposed model has significant advantages in classification accuracy.