摘要
给出了一种改进的基于遗传算法的聚类方法。传统的K-means算法局部搜索能力强,但是对初始化比较敏感,容易陷入局部最优值。基本的基于遗传算法的聚类算法是一种全局优化算法,但是其局部搜索能力较差,收敛速度慢。针对这两个方法所存在的问题,提出了一种改进的聚类算法。该方法结合了两个方法的优点,引入了K-means操作,再用遗传算法进行优化,并且在该方法中改进了遗传算法中的交叉算子,大大提高了基于遗传算法的聚类算法的局部搜索能力和收敛速度。
An improved clustering approach is described in this paper.The traditional K-means algorithm is good at locally searching capability,but it is sensitive to the initialization,easy to get stuck at locally optimal values.The simple geneticalgorithm-based clustering method is a global optimize approach,but it is weak at the locally searching capability and convergence speed.Take the problems which exists in the two algorithms into account,a new clustering algorithm is put forwarded in this paper.The new approach integrates the advantages of the two algorithms ,which introduces the K-means operation and then utilizes the genetic algorithm to do some optimization and also improved the crossover operation in the genetic algorithm,improves the locally searching capability and convergence speed of the genetic algorithm-based clustering algorithm.
出处
《计算机工程与应用》
CSCD
北大核心
2007年第21期170-172,共3页
Computer Engineering and Applications
