Research of Correction Method in the Feature Space on Text Clustering

For the feature space of high-dimensional data on text clustering contains many redundant features, even "noise" features. The author proposed a feature space correction method, combine with a supervised feature selection methods and K-means clustering method. By analyzing the significance of the features in the clustering process and selecting the features that have more significance, to amend the initial feature space to exclude the less important features, give prominence to the main features, reduce the noise and improve the clustering effect.