| A database management system is a tool to define, store and manipulate data. A knowledge based system is composed of an inference engine for reasoning and a knowledge base for storing facts and rules. While the database system is static, the knowledge base system is dynamic. Therefore, it is advantageous of linking knowledge bases with databases. The conventional approaches to link the knowledge base and the database have several problems. For example, such a system can not deal with both fuzzy and nonfuzzy data, and requires large, rigid, unadaptable rule domains.; The objective of this research is to develop a methodology for supporting a hybrid decision system which can (1) integrate knowledge bases and databases, (2) have the benefits of different technologies, (3) process fuzzy and nonfuzzy data, (4) maintain an acceptable rule domain, (5) allow inconsistent knowledge, (6) have adapting and fault-tolerance features, and (7) operate actively. The methodology integrates object-oriented databases, fuzzy logic controllers, neural networks, and active systems. The knowledge base (the fuzzy logic controllers and neural networks) can be integrated with the database (the object-oriented database) so that the data can be organized statically and the system can be operated dynamically. Three approaches have been developed for integrating the hybrid architecture. They are: (1) using neural networks to learn the fuzzy if-then rules, (2) using the neural networks to simulate the membership functions for a fuzzy logic system, and (3) using neural networks for classification in which the input includes both fuzzy variables and nonfuzzy variables. Furthermore, the addition of triggers and constraints in the hybrid decision support system has been analyzed. An active material requirements planning model has been built for validating and verifying the research. It has also been demonstrated that the active MRP does not have conventional period-by-period limitations leading to a real-time and bucketless system. |
