An overview of Taguchi’ s T-Method as a prediction tool for multivariate analysis

Abstract

Analysis of prediction has attracted considerable interest in various fields. Taguchi’s T-Method is a prediction method introduced by Genichi Taguchi in mid-year 2000, among several other Mahalanobis Taguchi system tools. It was explicitly created for the prediction of multivariate data. Taguchi's T-Method has shown that even with limited sample size, making a prediction based on historical data is possible. The key elements that have been adapted in reinforcing Taguchi’s TMethod robustness are by introducing the unit-space principle and adaptation of the signal to the noise ratio (SNR) as a weighting as well as a zero-proportional theory, as proposed by Genichi Taguchi in a robust model. Taguchi’s T-Method was widely practicing in Japan and began to be practiced by non-Japanese researchers due to its simplicity and simple understanding. Up to recent, various applications of Taguchi’s T-Method been applied, which prove to be beneficial to industrial needs. This research paper outlines the T-method procedures by applying it in a few benchmark datasets and compare the accuracy with the existing multiple linear regression method for an overview. The results show that Taguchi’s T-Method is better than multiple regression in dealing with limited sample data in which the sample size is smaller than the input variables. Taguchi’s TMethod proved to have the ability to predict output with an acceptable range of prediction accuracy.