Research On Hierarchy-based Fuzzy Clustering Algorithm

Data Mining is a superiorhot area in the information and database technology, and is commonly considered as one of the key technology with wild developing perspective. Clustering is an important area for research in Data Mining, and also an important method in data partition or data grouping. Clustering has been widely used in commerce, market analysis, biology, Web classification and so on. Partitioning methods, Hierarchical methods, Density-based methods, Grid-based methods, Model-methods are the basic methods in clustering. These methods have related to all of fields of AN science, and have got great effect in specific fields and states.First, the clustering technologies are analyzed and discussed. Main clustering algorithms are summarized and classified, and their principle and key techniques are introduced systematically. Then aiming at the sensitivity of initial value in fuzzy c-means (FCM) and the defect of unconsidered distance between classes in objective function, a hierarchical fuzzy clustering algorithm (HFCM) was put forward to overcome the limitation of fuzzy c-means (FCM). HFCM can fast discover the high concentrated data areas by the agglomerative hierarchical clustering method, analyze and consolidate the data areas. Using the evaluation function to the optimum clustering scheme is found. Last, man-made data and true data are used to verify the HFCM. The experimental results indicate that HFCM has the higher classification precision and higher ability of excluding noises.