Intelligent model management systems

Kyung Hoon Yang, Purdue University

Abstract

Model management systems (MMS) research has increasingly become an important concern in decision support systems (DSS). One of the most difficult and least investigated areas of MMS research deals with helping users formulate new models and interpret their results. In real life, decision-makers who are not versed in management science techniques are not generally able to map their problems directly to an appropriate model. They also cannot interpret model results because the output of a management science model contains for them an incomprehensible syntax such as vector of real numbers or tables. Helping users formulate a new model may require selecting an appropriate model from the model base, combining several models, choosing the best model from several alternative models, and selecting the most appropriate model in the absence of a perfect fit (mapping). Moreover, helping users interpret model results may require endowing semantic meaning to the results of a syntactic model by using accumulated knowledge and heuristic reasoning. This research presents methodologies to address problems in model formulation and interpretation. This study argues that the formulation of unstructured problems has much in common with processing and classifying indistinct patterns or images as in electrical engineering or in computer science. The methodology proposed draws on research in syntactic pattern recognition. Fuzzy logic theory, which has found use in the interpretation of natural language, can also be applied to model interpretation. We show how fuzzy logic can be used to map and interpret an unstructured problem for various decision models. The applicability of this approach is demonstrated by its application to a variety of examples.

Degree

Ph.D.

Advisors

Santos, Purdue University.

Subject Area

Management|Artificial intelligence

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS