An Extensive Empirical Comparison of <i>k</i>-means Initialization Algorithms

The k-means clustering algorithm, whilst widely popular, is not without its drawbacks. In this paper, we focus on the sensitivity of to initial set centroids. Since cluster recovery performance can be improved by better initialisation, numerous algorithms have been proposed aiming at producing good However, it still unclear which algorithm should used in any particular scenario. With mind, compare 17 such 6,000 synthetic and 28 real-world data sets. sets were produced under different configurations, allowing us show excels each Hence, results our experiments particularly useful for those considering a non-trivial