One-class classification-based control charts for multivariate process monitoring

One-class classification problems have attracted a great deal of attention from various disciplines. In the present study, we attempt to extend the scope of application of the one-class classification technique to statistical process control (SPC) problems. We propose new multivariate control charts that apply the effectiveness of one-class classification to improvement of Phase I and Phase II analysis in SPC. In the proposed control charts, we use a monitoring statistic that represents the degree of being an outlier as obtained through one-class classification. The control limits of the proposed charts are established based on the empirical level of significance on the percentile, estimated by the bootstrap method. A simulation study was conducted to illustrate limitations of current one-class classification control charts and demonstrate the effectiveness of our proposed control charts.