A Probabilistic Hierarchical Classification Model for Rating Suppliers

Abstract

We develop a probabilistic approach for a hierarchical classification problem that arises in the context of total quality management, pertaining to the rating and classification of suppliers. For discussion, we focus on one of the prototypical applications regarding the hierarchical placement of software developers. The existing approaches classify a supplier into one of several classes, with certainty. In contrast, our approach assigns a positive probability to a supplier belonging to each of the several classes. The probabilistic classification will then serve as an input to any decision making procedure that is used for the actual supplier selection. We propose a normative approach for addressing this problem and illustrate it with an example from a real scenario. Since classification with certainty is often viewed as being unrealistic, the proposed approach offers an improvement over the existing methodologies. This article is geared towards readers interested in using statistical techniques for coherent decision making with a focus on problems encountered in quality management. The proposed method does not make the decision process easier; rather, by incorporating measures of uncertainty in the underlying ingredients, it makes the process better.

• Bayesian Classification Rule
• Capability Maturity Model
• Quality Management
• Scoring

Additional information

Notes on contributors

Nozer D. Singpurwalla

Dr. Singpurwalla is a Professor of Operations Research and a Professor of Statistics. His email address is [email protected].