The Research Of Fuzzy Clustering For Categorical And Mixed Valued Data

Clustering is an important technique of data mining. It finds the similarity among data from databases, then according to the similarity to classify the data. The classification makes the data of different cluster as different as possible and the data of the same cluster as similar as possible, then optimizes the query of large-scale databases and finds the useful hidden information or knowledge among data. At present, the clustering analysis about numeric valued data is quite mature, but the clustering analysis about categorical valued and mixed valued data is not perfect. Because there are lots of categorical valued and mixed valued data in practical use, this paper focuses on the research for a good clustering analysis algorithm suitable for categorical valued and mixed valued data. At the same time, we find that leading fuzzy theory into clustering analysis can optimize the result of clustering. So this paper also focuses on the research of fuzzy theory and the combination of fuzzy theory and the algorithm.At first, this paper interprets the clustering analysis of data mining. It introduces a few clustering algorithms suitable for categorical valued and mixed valued data and advances a new clustering method-CVAD(Categorical Value Attributes Decompose). On the base of fuzzy theory, we combine the fuzzy theory and CVAD and advance a clustering method which result is better-FCVAD(Fuzzy Categorical Value Attributes Decompose). Then this paper gives a minute analysis and evaluation for kinds of clustering analysis methods and realizes a clustering analysis system. At last, this paper introduces some practices of clustering analysis.