Artificial neural system in decision-aiding for large incomplete databases

The authors propose a hybrid knowledge-based model where neural network technology is used in decision-aiding processes to handle large amounts of incomplete information. The proposed model is composed of two major subunits: a decision-making network and a knowledge acquisition module. The decision-making network, after being trained, is used as a filter. The knowledge acquisition module is responsible for training the decision-making network. It is shown that the neural network, used as a complement to conventional expert systems, has a strong adaptive learning capability in decision-making. However, what constitutes the set of training data can directly affect the quality of the decision to be made. A semi-real incomplete database has been constructed to provide an appropriate test bed for the proposed decision support system. To investigate the feasibility and performance of the proposed system, a number of simulation runs were conducted and these results are presented