Research Of Microarry Data Based On Kernel Fuzzy Clustering

With the extensive application of microarray technology, massive amounts of gene expression data are produced. How to withdraw the useful biology or medical information from these data is the key and difficulty of microarray technology application. Cluster analysis can assemble these funcition-related genes as the co-expressive category according to the similarity of the genes, which is useful to the study of genetic function genetic control and genetic process and so on. This paper analyzes the gene expression data analysis commonly used clustering algorithm, pointing out their strengths and weaknesses, kernel fuzzy clustering algorithm is widely applied by its fuzziness division data and the processing non-linear data effect good merit in the gene expression data analysis.Firstly, in order to avoid the artificial initialization cluster parameter angle embarking, one kind of auto-adapted kernel fuzzy cluster algorithm is produced. This method fixes the smallest upper limit according to the subtractive clustering. Based on the effectivity of the different clustering results of class number in the measuring range clustering validity function, the pole is choosed as the best class number of the prediction. Using the gene expression data, the experimental result has confirmed the validity and feasibility of this method.Secondly, because kernel fuzzy clustering algorithm is easy to overlook the problem of outlier genes, one kind of outlier kernel fuzzy cluster algorithm is produced. By each gene is assigned a dynamic weight value, and continuously updated in the iteration algorithm to find outliers gene weights, clustering efficiency is improved. Deriving with the formula proves that the algorithm is convergent. Through combination of the auto-adapted kernel fuzzy cluster algorithm and outlier kernel fuzzy cluster algorithm, one kind of auto-adapted outlier kernel fuzzy C means algorithm is proposed. Using the yeast cell gene expression data test, the algorithm simulation results show that this algorithm in optimizing the precision and efficiency is better than the ever proposed kernel fuzzy method.Finally, clustering of gene expression data analysis system achieves a major genetic data loading, data preprocessing, auto-adapted outlier kernel fuzzy C means algorithm, and clustering results visualization functions. System tests the gene expression data experiments, and the final clustering results obtained reflectes the laws in biological significance.