Application Research Of Clustering Analysis On Classification Data Mining

With the development of information technology, Data Mining has been paid attention extensively. As we know, Data Mining has many research areas and classification is one of the important research subjects in it. Researching into the subject has very important value not only on theoretic but also on application.This work is supported by Natural Science Foundation of Shaanxi Province. On the one hand, the research focuses on the pre-processing stage of classification system, which aims at the improvement of classification performance result by improving the feature extraction. On the other hand, by researching into the new model of classification-classifier ensemble, a new combination method is proposed to improve the classification performance. Clustering technology plays an important role in the course of research above.The main work in this paper is shown as follows:(1) An overview of clustering technology and analysis of the main clustering algorithm is provided for research later.(2) An improved PCA feature extraction method based on K-maxmin clustering algorithm is proposed, which overcomes the weak of PCA. Experimental result shows that the classification performance is better than the one of traditional PCA method in feature extraction.(3) A classifier ensemble method based on Agglomerative Hierarchical Clustering algorithm is proposed, which increases the diversity of classifiers and ensures that all members of classifiers have good classification performance. Experimental result shows that it is better than traditional Bagging method in classification performance.