对聚类算法K-means的研究与改进

he traditional K-means algorithm is great affected by the quality of the cluster’s number and the initial cluster selected ,in order to reduce the dependence on the initial centers and rapidly positioning the types of new data, we have the document density as parameters, divide the documents into several blocks and do clustering on every blocks, this algorithm is fit for the text data. Experiments show that this algorithm can make a higher quality for cluster, while it also do well in the new increasing data.