A study of some clustering techniques in view of feature reduction

This paper focuses on the study of combination of some of the existing clustering methods and utilizing them meaningfully for feature dimensionality reduction. The efficacy of different cluster validity indices is investigated in selecting good features. A fuzzy ART network is used for experimentation. Colon dataset, wine dataset and iris dataset are used in this study to see the effectiveness of combination of different techniques of cluster validity and clustering in feature reduction. The proposed method of clustering and then reducing the features seems to be very powerful to be used on huge bioinformatics datasets to reduce the number of redundant features.