Two-way Clustering Problems Based On Ant Colony Algorithm Research

Bioinformatics is developing rapidly in recent years, which DNA gene expressionchips come out, and it can measure thousands of gene expression levels, and generate adata set. The gene expression data matrix contains biological information, through theanalysis of gene expression data matrix, we can further explore the genetic dataintrinsically linked to the discovery and cure some genetic diseases.Traditional clustering methods cluster only from a single row or columndirection.However, due to some genetic samples will not show certain regularity in allthe columns or rows. Therefore, we must cluster matrix from rows and columns in bothdirections. This lead to a two-way clustering algorithm proposed, and two-wayclustering algorithm has played a huge role in exploring the genetic mysteries.In this thesis, we use the idea of ant colony algorithm to solve the two-wayclustering problem, proposing a two-way clustering algorithm based on ant colony. Inthis algorithm we establish two-way clustering model based on ant colony algorithm tofind two-way clusters.Search by sending a considerable number of ants, the solution setof constantly updated in order to obtain the global optimum. The thesis includes thefollowing:1. The algorithm uses the mean square residuals concept proposed by Cheng andChurch to judge the quality of two-way clusters, establish two-way clustering modelbased on ant colony algorithm. The algorithm is searching the biclusters with smallmean square residuals and big volume under the help of this bicluster model.2. The search strategy in the algorithm is adaptive, by random searching of antsand complement each other, adjust the size of the array of pheromone, we can usepositive feedback mechanisms to continually optimize the quality of the bicluster, whilenegative feedback mechanism to ensureglobal search.3. Two-way clustering algorithm based on ant colony is achieved.Algorithmparameters are selected in the experiment. We analysis and processing of yeast geneexpression data sets using two-way clustering algorithm based on ant colony algorithmto find the gene expression data set of two-way clusters,and the experimental results were compared with other two-way clustering algorithm.Two-way clustering algorithm based on ant colony algorithm applied to geneexpression data can quickly find the two-way clusters.It use negative feedbackmechanism, to avoid falling into local optimum, and use positive feedbackstrategy toimprove the quality of the results. Meanwhile, the distributed nature of the ant colonyalgorithm can make the task easy decomposition to facilitate multiple hostssimultaneously.