K-means Clustering Based on Attribute Weighted Matrix

Abstract

In the procedure of clustering, different attributes have different effects on clustering results. However, traditional K-means algorithm does not take this into account. So that, the clustering results cannot accurately reflect the internal similarity of data. This paper improves traditional K-means clustering algorithms based on attribute weighted matrix, so as to ensure that the more important the attribute is, the more important role it plays in clustering. The weighted K-means clustering algorithm has a better performance than the traditional, which is showed in the experiments.