Research and application on KNN method based on cluster before classification

In order to mine the hidden knowledge, and solve the problem of ldquodata is superfluous but knowledge is sparerdquo, data mining is used widely. Data classification is an important task of data mining, which has been at the center of research interest in recent years. The research of this paper is on classification. Starting from fuzzy KNN classification method, based on the idea of cluster before classification, a method - KNN method based on cluster before classification - is proposed. Paper executes particular and extensive experiments, which include two parts: validating the influence the parameters have on new method and comparing the expansibility of the new classification method and fuzzy KNN method with the increasing of the dataset size and the number of attributes. The experimental results show the benefits of new method when dealing with larger datasets. Finally, the proposed method is applied to data preprocessing. Executing the classification on vector map datasets about vegetation, the vegetation style can be obtained, so data preprocessing is completed; we also display the classification results using the software of geography information system such as ArcGIS.