Using Evolutionary Computation Approach To Improve The Performance Of The Fuzzy-ART For Document Clustering

• Main Authors: Yih-Juh Liaw, 廖益助
• Other Authors: 陳大正
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/28855229853088283443

碩士 === 長榮管理學院 === 經營管理研究所 === 90 === Automatic document clustering plays an important role in the knowledge management; it is the important issue of document retrieval, too. In the literature, artificial neural networks (ANNs) have been widely applied in the document clustering applications. For overcoming the stability-plasticity dilemma that every ANN clustering system has to face, Fuzzy ART system has been proposed and widely applied. However, for obtaining better clustering, three parameters of Fuzzy ART need to be adjusted manually by means of trial and error. It is time-consuming and does not guarantee an optimum result. Evolutionary computation approaches (ECs) are optimal mathematical search technique based on the principles of natural selection and genetic recombination. Therefore, a hybrid approach (EC based Fuzzy Adaptive Resonance Theory; ECFART) incorporating an EC and a Fuzzy ART neural network has been applied to automate the Fuzzy ART parameters selection process so that the best clustering result can be obtained. A fuzzy concept network of ninety documents has been tested in order to validate the proposed approach. The result shows that the most appropriate parameters of Fuzzy ART could be obtained consistently and effectively by the proposed hybrid approach.