Automated Knowledge Acquisition in a Neural Network-Based Decision Support System for Incomplete Database Systems

Abstract

: A neural network–based decision support system has been designed and simulated to be used as a filter to improve the system performance of large incomplete databases enhanced with maybe algebra. To train the network, a knowledge-acquisition module equipped with a fuzzy logic technique was used to automatically generate a set of training pairs according to the semantics of the underlying database, the specific characteristics of the user query, and user requirements. Based on the notion of relative graded membership, a fuzzy logic–based controller was used to monitor and measure the quality of each training pattern as a means to generate a set of “good” training pairs. Finally, the proposed scheme has been simulated and analyzed to determine the effectiveness of the automatic training pairs generation process.