Title: An unsupervised classification scheme for multi-class problems including feature selection based on MTS philosophy
Authors: Prasun Das, Sandip Mukherjee
Addresses: SQC & OR Unit, Indian Statistical Institute, 203, B.T. Road, Kolkata – 700 108, India. ' SQC & OR Unit, Indian Statistical Institute, 203, B.T. Road, Kolkata – 700 108, India
Abstract: In this paper, the proposed Unsupervised Mahalanobis Distance Classifier (UNMDC) scheme is a multi-class unsupervised classifier with the basic philosophy of supervised Mahalanobis–Taguchi System (MTS) based monitoring procedure. A comparative study between the MTS and the proposed UNMDC is performed with various simulated experiments for different types of correlation structure and location parameters, published data and real-life data sets of different sizes and dimensions. The advantages of domain knowledge independent thresholds, multi-class separation, identifying process shifts during multivariate process-monitoring and feature selection in case of detection of abnormals are the special merits of this algorithm.
Keywords: unsupervised learning; Mahalanobis distance; MTS philosophy; feature selection; threshold; severity level; Mahalanobis–Taguchi system; process monitoring.
International Journal of Industrial and Systems Engineering, 2009 Vol.4 No.6, pp.665 - 682
Available online: 26 Jun 2009 *Full-text access for editors Access for subscribers Purchase this article Comment on this article