An Application of Supervised Multivariate Control Charts on aPower Plant Data

摘要

We propose a supervised data mining approach that takes advantage of available out-of-control data. By usingclassification techniques, our proposed approach can utilize both the in-control and out-of-control data informationto construct a control chart. The probability that a data point belongs to a certain class from the classification resultis plotted and the control limits can be established. A real power plant data set is used to illustrate our proposedapproach. The comparative study shows that our proposed approach performs better than traditional multivariatecontrol chart techniques.