Integrated model management in the data warehouse era
Introduction
The area of model management originated as a database analog for applying information technology in the service and advancement of mathematical and decision modeling [11], [26]. Contemporaneous with the evolution of model management has been the demand for integrated modeling and decision support environments to facilitate the integration of models across multiple paradigms [10], [15], [16], [19], [22]. The emergence of the Internet and Web as the hub of computing activity has changed the landscape dramatically, leading to Web-based modeling [1], [4].
Because model management has its conceptual roots in data and database management, the evolution of data management in the Web world is of particular interest. Perhaps the most significant development in this arena is the emergence within the past decade of data warehouses as analytically-oriented data stores organized for supporting business intelligence. Data warehouses in concert with the Web can provide rapid and widely distributed decision making information to the key players in an organization. Surprisingly, with few exceptions however (see [20] for example), the impact of this technology on modeling has not yet been investigated to any great extent.
We approach the relationship between modeling and data warehouses in the context of integrated modeling environments, indicating how warehouses and other Web technology can facilitate model integration in ways quite different from the initial concept of a model management system as a counterpart of a database management system. Proceeding in a phased way, beginning with an overview of data warehouse concepts, we develop the notion of decision metrics as a foundation for what information such a decision-oriented database should contain. Decision metrics provide readily understandable problem identification cues which can then be tied to decision models for subsequent problem diagnosis and sensitivity analyses. In addition to serving as possible inputs to models, decision metrics may themselves be either indirect or direct products of models as well. The relationship between metrics and models suggests extending the notion of data warehouse to encompass a model warehouse which stores model meta-information including assumptions, structure, related data requirements, and solver interfaces. Finally, we envision an integrated modeling environment as a distributed, component-based, warehouse-driven software system enabled by intelligent software agents and configurable to specific user’s and/or organization’s requirements. This “bottom up”, distributed view of an IME lies in sharp contrast to the “top down”, centralized DBMS architecture which has historically been proposed.
Section snippets
Retrospective of integrated modeling environments
The concept of an IME was born out of the model management movement which, in turn, was seen initially as a decision model counterpart to, and logical extension of, developments in database management technology [9], [26]. Thus, as Table 1 depicts, the standard pillars of database management philosophy were seen as equally applicable, if not always easily adaptable, to the management of decision and mathematical models as well, particularly in the areas of management science and operations
Data warehouses
The emergence of data warehouses has become the major database development of the 1990s. Revenues for data warehouse products and services are estimated to be anywhere from $US 5–8 billion in 1998 (depending on which source you consult) with annual growth rates of 35–40%.
Decision metrics
We motivate the discussion of decision metrics from the perspective of performance measurement.
Decision metrics and model management
Decision metrics can be either the outcomes of decision models and/or can feed the execution of other decision models in order to support the process of problem diagnosis. The support of models in the data warehouse environment can be implemented with a special kind of model warehouse as we indicate below.
Component-based IME
The close connection between metrics and models argues in favor of an integrated modeling environment that is Web-based. One distinguishing characteristic of this architecture is that it is built using configurable components rather than as a monolithic, miniature operating system as initially conceived. Fig. 9 shows a high level schematic of the components needed for this IME.
We have discussed most of the components in one form or another above. However, one area of opportunity that has not
Conclusions
We have undertaken a brief, historical evaluation of model management research, re-evaluated the need for an IME, and re-engineered the IME concept at a very high level to bring it in line with the contemporary, Web-based world. While discarding the database analogy on one hand, we have reverted to its Web counterpart, the data warehouse, on the other. We have shown ways in which the data warehouse can support the MS/OR paradigms of model management, specifically via the definition of decision
References (27)
- et al.
Decision support on demand: On emerging electronic markets for decision technologies
Decision Support Systems
(1997) - et al.
An implementation of a graph-based modeling system for structured modeling (GBMS/SM)
Decision Support Systems
(1998) An introduction to model integration and integrated modeling environments
Decision Support Systems
(1993)- et al.
Model integration and a theory of models
Decision Support Systems
(1993) - et al.
BLOOMS: A prototype modeling language with object oriented features
Decision Support Systems
(1997) Integrated modeling environments
European Journal of Operational Research
(1989)- et al.
Independence and mappings in model-based decision support systems
Decision Support Systems
(1993) - H. Bhargava, R. Krishnan, The world wide web: Opportunities for Operations Research and Management Science, INFORMS...
- H. Bhargava, R. Krishnan, S. Roehrig, D. Kaplan, M. Casey, R. Mueller, Model management in electronic markets for...
- H. Bhargava, R. Krishnan, D. Kaplan, On generalized access to a WWW network of decision support services, in:...