A new feature weighted self-adaptive FCM clustering algorithm for gene expression data

When dealing with gene expression data, FCM algorithm has the following defects: sensitiveness to the initial cluster centers, need to input the cluster number preliminary and doesn´t consider the different contributions of attributes of gene expression data and so on. Accordingly, this paper presents a new feature weighted self-adaptive FCM algorithm handling gene expression data. First introduce a dataset pre-processing algorithms, then determine weights of all features of gene expression data set based on message-entropy, finally introduce the features´ weights into the objective function of FCM algorithm. Experimental results show that the new algorithm significantly improves the effectiveness of taxonomic notes for gene expression data.