A generalized architecture for intelligent reactive management systems.

摘要

This dissertation extends the classification of computer-based information systems beyond those which primarily support planning to include systems which support decision-making when planning breaks down, forcing management to react to unforeseen circumstances. The need for reactive management systems has been perceived by other researchers, but very little actual design and implementation effort has been documented. The literature review thus examines the context for reactive management systems, as well as addressing specific design issues, in order to help identify the operative theoretical framework. The research method used is to first analyze computerized information systems needs for crisis response, then to describe a generalized architecture, and finally to construct a prototype using commonly available tools.; Design of the architecture assumes that the system will require artificial intelligence, and that in order to be useful, it should be generalizable, widely available, rapid in execution, modular, transportable, and capable of execution by personnel who are not computer experts. Functional requirements indicate that the system should provide broad input capability, intelligent user interface, pattern recognition facility, early reporting of unsafe or crisis conditions, and recommendation of a course of action. The resulting generalized architecture consists of six modular functions: data gathering, system monitoring, pattern recognition, status reporting/early warning, causal analysis, and recommendation. The generalized architecture is domain-free and capable of application to many specific problems with relatively minor modification. The overall design represents a use of abductive inference, attacking a problem through common sense and the use of associated details.; Although the architecture is not size-specific, the example implementation prototype is microcomputer-based. It uses standard software packages to provide database/spreadsheet functions, neural net functions, and knowledge-base/user-interface functions. The implementation is applied to a simulation of a small manufacturing firm, modeling aspects of operation relating to marketing, budget execution, cash management, and production.