Abstract: Aiming at the problem that MinMax k-means algorithm was easy to generate empty solutions, slow convergence speed and low computational efficiency, we proposed an incremental MinMax k-means clustering algorithm. The algorithm started from a given initial clustering number, and generated new cluster centers by increasing a fixed step length. The fast dividing method based on data balance was used to generate incremental cluster centers, so as to avoid the large amount of calculation problem caused by traversing data and too many running times of k-means clustering algorithm in traditional incremental clustering center selection. Compared with MinMax k-means and related algorithms, the experimental results show that the algorithm is superior to the comparison algorithm in calculation efficiency and accuracy, and effectively improves the sensitivity of MinMax k-means clustering to initialization center and easy to generate empty solutions.

Key words: k-means clustering, incremental clustering, initialization, cluster center