Hybrid expert system using case based reasoning and neural network for classification

Case Based Reasoning (CBR) is an analogical reasoning method, which solves problems by relating some previously solved problems to a current unsolved problem to draw analogical inferences for problem solving. But CBR faces the challenge of assigning weights to the features to measure similarity between a current unsolved case and cases stored in the case base effectively and correctly. The concept of neural network's pruning is already used to sort out feature weighting problem in CBR. But it loses generality and actual elicited knowledge in the ANN's links. This work proposes a method to extract symbolic weights from a trained neural network by observing the whole trained neural network as an AND/OR graph and then finds solution for each node that becomes the weight of a corresponding node. The proposed feature weighting mechanism is used in CBR to build a hybrid expert system for classification task and the performance of the proposed hybrid system is compared with that with other feature weighting mechanisms. The performance is validated on swine flu dataset and ionosphere, sonar and heart datasets collected from UCI repository. From the empirical results it is observed that in all the experiments the proposed feature weighting mechanism outperforms most of the earlier weighting mechanisms extracted from trained neural network. (C) 2014 Elsevier B.V. All rights reserved.