Classification method based on Taguchi’s T-method for small sample sizes

Abstract

This paper proposes a classification method using Taguchi’s T-method. Collecting sufficient number of training samples at the initial stage of production and at the beginning of the development process is difficult. Because of the small sample size, it is difficult to apply conventional classification methods. In this paper, we propose a classification method that is based on the Taguchi’s T-method for calculations with small sample sizes. The proposed method involves construction of a classification model that quantifies the Taguchi’s T-method and can be applied to small sample sizes. We confirm that the classification accuracy of the proposed method is superior to that of the Mahalanobis–Taguchi method. Through a verification experiment, we examine the change in the classification accuracy when the number of training samples is increased. It is confirmed that the conventional method outperforms the proposed method when the number of training samples is more than thrice the number of parameters.