A novel hybrid taguchi-grey-based method for feature subset selection

Abstract

In this paper, a novel hybrid Taguchi-Grey-based method for feature subset selection is proposed. The two-level orthogonal array is employed in the proposed method to provide a well-organized and balanced comparison of two levels of each feature (i.e., the feature is selected for partern classification or not) and interactions among all features in a specific classification problem. That is, this two-dimensional matrix is mainly used to reduce the feature subset evaluation efforts prior to the classification procedure. Accordingly, the grey-based nearest neighbor rule and the signal-to-noise ratio (SNR) are used to evaluate and optimize the features of the specific classification problem. In this manner, important and relevant features can be identified for pattern classification. Experiments performed on different application domains are reported to demonstrate the performance of the proposed hybrid Taguchi-Grey-based method. It can be easily seen that the proposed method yields superior performance and is helpful for improving the classification accuracy in pattern classification. © Springer-Verlag Berlin Heidelberg 2007.