Abstract: A new elite genetic K-medoids clustering algorithm is proposed to solve the problem of local optimal solution and result instability. The algorithm uses elite strategies to control the overall evolutionary direction of genetic operations, and produces several random individuals based on the average fitness to reduce the impact of premature. At the same time, in order to improve the evolutionary efficiency, the algorithm designs a new crossover method. In order to ensure the superiority of the crossover and mutation, the algorithm introduces a competition mechanism. The simulation results of 8 data sets show that the algorithm improves the stability of clustering results and the accuracy of clustering.

Key words: genetic algorithm, elitist strategy, K-medoids algorithm, cluster analysis

摘要： 针对K-medoids算法易陷入局部最优和聚类结果不稳定的问题，提出了一种精英遗传K-medoids聚类算法。该算法使用精英策略来控制遗传操作的整体进化方向；根据种群的平均适应度引入若干随机个体来提高种群多样性，从而在一定程度上减少了遗传算法的早熟现象。为了提高进化效率，该算法设计出一种新的交叉方式；为了保证交叉变异结果的优秀性，该算法引入了一种竞争机制。8个数据集的仿真实验表明，该算法在提高聚类准确率的同时，聚类结果的稳定性也有所提高。

关键词: 遗传算法, 精英策略, K-medoids算法, 聚类分析