Abstract:

Using KMeans iterative twice KMeans iterative to get an initial cluster center,in the text clustering process, a set of discriminative directions are chosen to construct the IMIC coordinate, each axis is constructed to rescaling function in order to improve the effectiveness of cluster policy,according to the distribution characteristics of the initial clusters. The IMIC iterative algorithm converges to the final solution.The time complexity of IMIC remains the same as KMeans by using a KMeanslike iteration strategy.The experimental results show that the IMIC algorithm has better clustering quality.

Key words: iteration convergence;text;clustering