Research On Classification Of Genes Based On Evolutionary Algorithm And Spatiotemporal Independent Component Analysis

Classification of genes is an important issue of microarray technology and genechip, which divides vast amounts of gene expression data into a small number ofgroups with biological significance and contributes to the research on functionalrelationships and work ways of different combination of genes. This paper first studygene classification based on ant colony and genetic which are evolutionary algorithms,propose improvements according to the existing problems. Then spatiotemporalindepent component analysis is applied to gene classification.Inherent parallelism and robustness make ant colony algorithm an effectivesolution to complex optimization and clustering problems. However, the traditionalalgorithm of ant algorithm is of slow evolution and easily falling into local extremes.We take original model of ant problem as antigens, construct solutions and updatepheromone using the way of clone selection and immune memory to avoid prematureconvergence and stagnation phenomenon. At last this improved algorithm is comparedwith the traditional ant algorithm and RFE, ALMA algorithms, a better ofconvergence speed and classification results is achieved.Orderly and highly distinguishable characteristics gene collections can beobtained by combination genetic algorithm with independent component analysis. Butthe same number of genes within each class can’t be guaranteed, which affect theclassification of training, reduce the classification precision. We do a fuzzy cluster fora set of feature genes generated by independent compont analysis. We select from thesame number of feature genes from each cluster at random to construct chromosomes.Selection，crossover and mutation operations is used to obtain the best classificationeffect sets. Compared with the pre-classification and associated algorithms, a better ofclassification results is achieved.Spatiotemporal indepent component analysis can make use of independence ofsample and gene direction. In gene classification, gene direction is defined as spacedirection, sample direction is defined as time direction and direction between them isdefined as spatiotemporal direction. Separation matrix is generated randomly,different direction of source array is solved by appropriate independent componentanalysis, and three iterative processes are analyzed. By comparison experiments withstatistical with significant, the independent in time-based direction and spatiotemporal direction can achieve a better result.