Rule-based methods for diagnosing out-of-control signals

There is a rich literature on methods for detecting out-of-control signals. However, the literature on methods for diagnosing out-of-control signals is not as rich; particularly, for methods that detect the specific levels and/or values of the covariates driving the change(s) in the process. In this paper, we propose a novel approach to this problem using a combination of data mining methods. We provide a variety of simulated scenarios with different out-of control signals and show how this method is able to isolate the specific values of the covariates contributing to the out-of-control signals. In addition, we compare the performance of our method to partial dependence plots (PPDs) and suggest how the strengths of the two can be combined in order to develop an even more robust scheme for diagnosing out-of-control signals.