A Neural Network Driven Fuzzy System Approach to Decision-Making
Vipul Kumar Gupta · The Journal of International Information Management · 1994
It is necessary to model and manage uncertainties efficiently and effectively in solving decision-making problems. Fuzzy reasoning and logic offers a natural means of handling un certainty. This paper discus.'ses the development of an architecture that utilizes the advantages of fuzzy logic and neural networks that can be used in decision-making. The present paper outlines the reasons that motivate development of models that integrate fuzzy logic and neural networks. This discussion is followed by a brief overview of fuzzy logic and its concepts that are essential in understanding the application presented in the paper. The paper then describes the classification procedure used by the model, which is followed by its application to the decision making problem in construction modularization. INTRODUCTION In solving decision-making problems, it is essential to model and manage uncertainties efficiently and effectively. There can be several causes of uncertainty in a decision-making situation, such as problem complexity, ill-posed questions, imprecision in computations, ambi guity in data/knowledge representation, problems in input interpretations, and noise of several types (Keller & Tahani, 1992). In the past, rule-based expert systems have been used for han dling uncertainty in such problems. Generally, the expert systems are based on classical logic and developers need to add special methods for handling uncertainty. Some of the methods used for handling uncertainty in expert systems include heuristic approaches, probability theory, pos sibility theory, and fuzzy theory. Fuzzy reasoning and logic offers a more natural means of handling uncertainty. All propo sitions can be modeled by possibility distributions over appropriate domains. Since fuzzy rea soning realizes the flexible reasoning similar to human logical reasoning, a considerable amount of research work has been performed in this (Takagi & Hayashi, 1991). Still there are some problems related to fuzzy reasoning to be solved, such as finding an easy to implement