Abstract:

A new online kernel fuzzy C-means (OKFCM) algorithm for large scale datasets based on kernel fuzzy C-means (KFCM) is proposed. In addition, taking into account the difficulties in selecting kernel parameters, an online multiple kernel fuzzy C-means (OMKFCM) algorithm is derived based on multiple kernel learning methods. Thus, the proposed algorithms not only ease the problem of selecting kernel parameters and inherit the superior clustering performance of the KFCM, but also are suitable for clustering data streams. Finally, the new online kernel algorithms are verified to have a better performance on artificial and real datasets compared with state-of-the-art partition clustering algorithms for large scale datasets.