A Study And Implementation Of Processing Gene Expression Profile Based On Prior Information And Binary Particle Swarm Optimization

Gene expression profile data has the characteristics of high feature dimension and small number of samples.Traditional machine learning methods still have much room to improve the processing of gene expression profile data.In recent years,the method based on swarm intelligence optimization and extreme learning machine has achieved good performance in feature selection and sample classification of gene expression profile data.However,due to the lack of prior information constraints in data,swarm intelligence optimization and extreme learning machine-based methods are prone to lose key genes in feature selection process,thus affecting the accuracy of cancer recognition,and lack of interpretability in processing methods.In order to improve the performance of gene expression profile data processing and the interpretability of processing methods,this thesis uses particle swarm optimization(PSO)algorithm encoding priori information to realize feature selection of gene expression profile data on the basis of obtaining priori information of gene expression profile data combination,and applies integrated extreme learning machine to realize cancer prediction.Finally,a prototype of gene expression profile data processing is designed and implemented System.The main work of this thesis is as follows:1.In order to improve the performance of gene expression profile data processing and the interpretability of processing methods,a gene selection processing method based on improved prior information and binary particle swarm optimization is proposed.Firstly,clustering was used to select representative genes from different functional gene clusters and establish the initial candidate gene pool.Secondly,the combination priori information in gene expression profile data is obtained by combining the category priori information with Pearson coefficient.Thirdly,the improved combination priori information is coded into the binary particle swarm optimization algorithm to select the gene subset highly related to the tumor category.Finally,an integrated extreme learning machine is built to classify gene expression profiles using diversity as an integrated index.The experimental results on several datasets show that the proposed gene expression profile data processing method can not only screen out the key gene subsets related to cancer,but also improve the accuracy of cancer recognition.2.On the basis of the above work,a gene expression profile data processing system based on prior information and binary particle swarm optimization algorithm is designed and implemented.The system includes three modules: gene expression profile data import,gene selection and data classification,which can automatically predict tumors.In this system,the parameters of feature selection method and tumor classification method can be set flexibly according to the gene expression profile data set to achieve efficient processing of gene expression profile data,which truly embodies data-driven.