Fuzzy Clustering Applied Research, To Automatically Determine The Field Of Expert Knowledge

It's very important for the establishment of national expert information service platform to mine the expert information. In the paper, unstructured data, such as papers published by experts and subjects committed by experts, is mined to recognise the knowledge domain of experts automatically.The thesis analyses the pretreatment Chinese Text Mining, combining the process of Chinese Text Mining,discussing the framework's hierarchy, the system structure about judging the knowledge domain of expert is presented.And every module in the structrue is introduced particularly.After that,the thesis studys the fuzzy clustering and the application in the clustering module. To determin the optimal number of clusters, the thesis introduces some clustering validity indices, studys compactness and separation criteria, and then betterment a common validity index to assess clustering validity. Besides,becase the text data has high dimension and many zero in every dimension,the dissimilarity is not great,and the effect of clustering is not good.So in the thesis ,the kernel functions is used in fuzzy attribute C-means clustering algorithm, making clustering can be performed efficiently.It gives several simulation experiments to illustrate the reliability and stability of the new index for determining the number of clusters,and demonstrate that kernelized fuzzy attribute C-means clustering algorithm is more effective and robuster than FCM, FAMC and KFCM as well. And the new index and the improved algorithm is used in the system structure about judging the knowledge domain of expert, advancing the precision and the recall of text clustering.