Fast centroid-based clustering algorithms such as k-means usually converge to a local optimum. In this work, we propose a method for constructing a better clustering from two such suboptimal clustering solutions based on the fact that each suboptimal clustering has benefits regarding to including some of the correct clusters. We develop the new method COTCLUS to find two centroids from one clustering and replace them by two centroids from the other clustering so that the maximum decrease in the mean square error of the first clustering is achieved. After modifying centroids, k-means algorithm with few iterations is applied for fine-tuning. In an iterative algorithm, the resulting clustering is further improved using a new given clustering. The proposed method can find optimal clustering in a very small number of iterations for datasets with well-separated clusters. We demonstrate by experiments that the proposed method outperforms the selected competitive methods.

中文翻译：

---

通过使用另一个聚类中合适的质心来改进基于质心的聚类

快速基于质心的聚类算法（例如 k-means）通常会收敛到局部最优值。在这项工作中，我们提出了一种方法，用于从两个这样的次优聚类解决方案构建更好的聚类，这是基于每个次优聚类在包含一些正确的聚类方面都有好处的事实。我们开发了新方法 COTCLUS 从一个聚类中找到两个质心，并用另一个聚类中的两个质心替换它们，从而实现第一个聚类的均方误差的最大减少。修改质心后，应用很少迭代的k-means算法进行微调。在迭代算法中，使用新的给定聚类进一步改进了所得聚类。所提出的方法可以在非常少量的迭代中找到具有良好分离集群的数据集的最佳聚类。我们通过实验证明所提出的方法优于所选的竞争方法。