Simultaneous feature selection and semi-supervised clustering for gene-expression data

In this paper, a new multiobjective optimization based technique is developed for simultaneous feature selection and semi-supervised clustering. Thereafter the proposed technique is applied for solving the problem of classifying gene expression data. Here a modern simulated annealing based multiobjective optimization technique namely AMOSA is utilized as the background optimization methodology. Features and cluster centers are represented in the form of a string. Based on the available features and the cluster centers, genes belonging to different clusters are assigned based on point symmetry distance. Four objective functions are simultaneously optimized by AMOSA to obtain the appropriate partitioning. First two cluster validity indices are symmetry distance based Sym-index and the Euclidean distance based XB-index, which are based on some unsupervised properties. Third one is a supervised information based cluster validity index, Minkowski index and last one is a function counting the number of features. For generating the supervised information, initially Fuzzy C-mean clustering technique is applied on the given gene expression data set. Thereafter based on the highest membership values of the data points to their respective clusters, randomly 10% data points with their class labels are chosen for measuring external validity index, MS Index. The proposed technique is applied on some publicly available gene-expression data sets. Results are compared with the existing techniques of gene expression data clustering.