Three essays on model management

Radha Mookerjee, Purdue University

Abstract

This dissertation consists of three essays examining three distinct but related issues in the area of model management. In the first essay entitled "Model Management Systems in Organizations: An Empirical Study," we survey a variety of business organizations to study their modeling needs. The purpose of the survey is to examine the validity of several assumptions made in the academic literature concerning the perception, motivation, benefits and use of model management systems in organizations. Although much of the academic literature appears to corroborate the findings of the survey, we find that commercially available software falls short of the capabilities expected of modeling related software by modelers and model users. The major conclusion that emerges here is that model management research needs to be much more practically oriented if the area of model management is to gain widespread acceptance in business and industry. The second essay entitled "Factors Influencing Model Management System Requirements: An Empirical Study," provides insights into the interrelationships between model and data complexity, modeling intensity, and the model management functions required by modelers and model users in organizations. Using data from a national survey of practitioners, we find that users consider existing model management systems deficient in the ability to manage complex models. We also show, for example, that: (1) data complexity has an insignificant effect on modeling intensity; (2) model complexity has a positive direct effect on modeling intensity; and (3) modeling intensity has a positive direct effect on functional requirements. In the third essay, "Strategies for Model Execution in Model Management Systems," we examine a central issue that influences the effective use of model management systems. This concerns the ability to provide a convenient way to execute models with the appropriate solver and data set. A model-solution procedure-solver classification scheme is proposed that could help system designers in choosing the appropriate model execution strategy. We present and discuss criteria that are relevant in choosing among model execution strategies. Finally, we describe a detailed structured modeling based implementation carried out to demonstrate key model execution strategies and reveal practical difficulties that may arise in the implementation of these strategies.

Degree

Ph.D.

Advisors

Chaturvedi, Purdue University.

Subject Area

Management

COinS