Principal Component Analysis based Feature Selection for clustering

Feature Extraction (FE) methods have been proved to be very effective for dimension reduction, but the features attained are meaningless. In order to exploit the effectiveness of FE methods to support Feature Selection (FS), this paper proposed a new FS approach for clustering based on Principal Component Analysis (PCA) called PS. It first uses PCA to transform the data from original feature space into a new feature space whose features are linear combination of the original ones, and then evaluates the importance of the original features based on the newly generated features and the feature importance measure proposed in this paper, finally selects features incrementally according to their importance to improve the performance of the clustering algorithm. Experiment is carried out on several popular data sets and the results show the advantages of the proposed approach.