CC-K-means: a candidate centres-based K-means algorithm for text data

摘要

K-means algorithm, one of the clustering algorithms, is widely applied to solve clustering problems of various data thanks to its simplicity and efficiency. However, the randomness of selecting centre points of the traditional K-means algorithm results in some defects such as low-speed of convergence or instability of clustering results. To overcome the impact of high-dimension during text clustering, latent semantic index (LSI) model is firstly adopted to reduce the dimensions of feature vector, and then weighted adjusted cosine similarity is used to calculate the similarity between documents to obtain better clustering effects. The high-density candidate centre points are partly updated to get the final clustering centres on the basis of density in the process of finding clustering centres. Experiment results show that the proposed algorithm can accurately find representative and decentralised clustering centres, which express a better performance in clustering.