A machine learning approach to improving process control

Process control mechanisms may not always succeed in producing desired outcomes. We propose an iterative approach that (a) applies data mining classification techniques in order to discover the conditions under which a controlled process produces undesired or sub-optimal outcomes and (b) uses this information to improve the control mechanism. While this approach shows promise for application in a variety of process control problem domains, in this paper we illustrate its use by applying it to a chronic disease care problem in a healthcare management organization, specifically the treatment of patients with type 2 diabetes mellitus. In particular, the proposed iterative approach is used to improve a canonical treatment strategy (based on clinical guidelines) by predicting and eliminating treatment failures, which delay or prevent patients from reaching evidence-based goals.