中文摘要：

K调和均值算法（KHM）用数据点与所有聚类中心的距离的调和平均值替代了数据点与聚类中心的最小距离，是一种对初始值不敏感、收敛速度快的有效聚类算法，但它容易陷入局部最小值。而遗传算法具有良好的全局优化能力。文中结合了KHM和遗传算法各自的优点，采用KHM计算每一代种群的聚类中心，并构造适应度函数，通过遗传算法进行一系列择优操作，成功地解决了KHM容易陷入局部最小值的问题。实验结果表明，所提出的算法不仅优化了聚类中心，而且还改善了聚类质量。

英文摘要：

In K-harmonic means clustering was an effective algorithm which was not sensitive to the initial value and converged quickly, it used harmonic means distance from the data point to all clustering centers to replace the minimum distance between the data point and all clustering centers. But it also easily converged to the local minimum, and genetic algorithm had a good global optimal capacity. Com- bined the advantages of KHM and genetic algorithm, used the KHM to calculate the clustering center of every population, and structure fitness function, through the genetic algorithm conduct a series of preferential operation, successfully solved the problem of KHM easily converged to the local minimum. The experiment showed the algorithm not only optimized the cluster centers, but also improved the clus- ter quality.