Abstract: Genetic algorithm has better global search capability，but has the shortcomings of premature convergence and slow end.K-Means algorithm has strong local search ability，but it’s sensitive to the initial cluster centers and easy to get stuck at locally optimal value.To solve such problems，it presents a hybrid genetic algorithm based on K-Means mutation operator.It combines the locally searching capability of the K-Means algorithm with the global optimization capability of genetic algorithm，and introduces the K-Means mutation operator into the genetic algorithm.It’s a hybrid algorithm using symbolic coding，adaptive mutation，and optimal individual retention policies.Simulation results show that the algorithm has effectively overcome the slow convergence of genetic algorithm and the locality convergence of K-Means algorithm，in order to get better clustering.

Key words: cluster analysis, K-Means algorithm, K-Means mutation operator, genetic algorithm

摘要： 遗传算法具有良好的全局搜索能力，但有过早收敛和过慢结束的缺点。K-Means算法具有很强的局部搜索能力，但算法有对初始聚类中心敏感而易陷入局部最优解。针对上述问题，提出了基于K-Means变异算子的混合遗传算法聚类，将K-Means算法的局部搜索能力与遗传算法的全局寻优搜索能力相结合，在遗传算法中引入K-Means变异算子，采用符号编码、自适应变异、最优个体保留策略的混合遗传算法。仿真实验表明，该算法有效克服了遗传算法过慢收敛和K-Means算法陷入局部收敛的问题，从而得到更好的聚类效果。

关键词: 聚类分析, K-Means算法, K-Means变异算子, 遗传算法