Microarray gene expression data classification using modified differential evolution based algorithm

The invention of micro-array technology helps to keep the records containing expression values of multiple genes over different experimental conditions. Clustering is an unsupervised classification technique which is considered as an important tool for inspecting micro-array data. The characteristics of such micro-array data include uncertainty, noise and imprecision. Therefore fuzzy clustering techniques have some great importance in examining such micro-array gene expression data sets. In this paper, a fuzzy modified differential evolution based clustering technique has been evolved which utilizes a newly proposed distance measure which posses the symmetry based property. The incorporation of symmetry based distance enables to recognize clusters having varieties of shapes and sizes and possessing some symmetric characteristics. Moreover introduction of search capability of differential evolution technique enables faster convergence rate as well as it is able to generate high quality results. Here symmetry based cluster validity measure, FCM-measure, is optimized by the proposed clustering technique to obtain the relevant partitioning. This proposed method shows that symmetry based distance is useful for detecting clusters from gene expression data sets. Finally, the inclusion of fuzziness properties helps the proposed method in handling overlapping clusters well. The effectiveness of the fuzzy point symmetry based modified differential evolution based clustering technique is compared with other clustering algorithms utilizing the properties of symmetry and genetic algorithms, over three publicly available benchmark micro-array datasets.