Abstract—The k-means clustering method is a widely used clustering technique for the Web because of its simplicity and speed. However, the clustering result depends heavily on the chosen initial clustering centers, which are uniformly chosen at random from the data points. We propose a seeding method that is based on the independent component analysis for the k-means clustering method. We evaluate the performance of our proposed method and compare it with other seeding methods by using benchmark datasets. We also applied our proposed method to a Web corpus, which was provided by ODP, and the CLUTO datasets. The results from the experiments showed that the normalized mutual information of our proposed method is better than the normalized mutual information of the k-means clustering method, the KKZ method, and the k-means++ clustering method.

Index Terms—k-means clustering method, KKZ method, kmeans++ clustering method, independent components analysis, seeding

Cite: Takashi Onoda, Miho Sakai, and Seiji Yamada, "Careful Seeding Method based on Independent Components Analysis for k-means Clustering," Journal of Emerging Technologies in Web Intelligence, Vol. 4, No. 1, pp. 51-59, February 2012. doi:10.4304/jetwi.4.1.51-59