A Knowledge Discovery by Fuzzy Rule Based Hopfield Network
In this paper, a new method for discovering knowledge from empirical data is proposed. This model consists of five steps. Firstly, we find the centers of fuzzy membership functions using adapted self-organizing feature map (SOFM). Secondly, we use the centers of Gaussian membership functions derived from previous step to determine the widths of Gaussian membership functions by means of the first-nearest-neighbor heuristic. Thirdly, it builds a weight network of Hopfield network so that weights reflect the importance of the network’s connections. Fourthly, Hopfield network is operated to get output values. The final step is to extract rules or knowledge via our proposed algorithm. In this algorithm, the irrelevant candidate rules are deleted so that the number of fuzzy rules and the number of antecedents can be defined. Therefore, it extracts fuzzy rules from the network. The experiments on wine recognition data show good performance concerning predictive accuracy.
KeywordsMembership Function Fuzzy Rule Knowledge Discovery Synaptic Weight Fuzzy Membership Function
Unable to display preview. Download preview PDF.
- M. Sugeno and T. Yasukawa, “A fuzzy-logic-based approach to qualitative modeling,” IEEE Trans. Fuzzy Syst., vol. 1, pp. 7–31, Feb. 1993.Google Scholar