Feature clustering and feature discretization assisting gene selection for molecular classification using fuzzy c-means and expectation-maximization algorithm

摘要

In this paper, a novel gene selection benefiting from feature clustering and feature discretization is developed. In large numbers of genes, unsupervised fuzzy clustering algorithm facilitates the analysis of both similarities and dissimilarities. The supervised process, adopting information gain and statistical Chi-square test, is applied to approve the relevant gene clusters. Then, expectation-maximization algorithm discretizes the candidate genes and helps to recognize distinguishability. In our previously proposed selection criterion, we finalized gene selection and generated the gene subsets for molecular classification. For high-dimensional datasets congested with erroneous or ambiguous information, the current scheme is particularly suitable in its own right. The efficiency and effectiveness are verified by our experimental results.