Mahalanobis-Taguchi system for pattern recognition: A brief review

Abstract

Mahalanobis-Taguchi system (MTS) is a data mining method that employs Mahalanobis distance (MD) and Taguchi robust engineering philosophy, in order to explore and exploit data in a multidimensional system. Furthermore, MD calculation provides a measurement scale that is used to discriminate and interpret the relationships among data samples (abnormal vs normal). Another function of MD calculation is that it is an approach of measuring the level of dissimilarity between them (severity). One unique feature of MTS lies on its precision in assessing the variability among all levels of samples (noise). Besides, MTS possesses the ability to evaluate the significant and insignificant factors that contribute to the performance of system (optimize). This evaluation is performed through a simplistic yet effective technique (orthogonal array and signal to noise ratio). Moreover, MTS receives a tremendous amount of appreciations by various industries due to its ability to perform various functions in diverse industrial applications (among others) of optimization, inspection, diagnostics, monitoring, estimation, prediction, classification and discrimination. Despite the appreciations received by this method, some criticize the reliability of MTS on its effectiveness, while some contribute ingenious ideas (hybrid systems, intelligent search algorithms, novel mathematical computation methods, etc.) for a further improvement of this method’s implementation. Therefore, this paper provides a general review on MTS research trends and studies conducted on this method outside Japan, in order to identify future research prospects.