基于遗传算法的一种改进的K-均值聚类算法

摘要

The traditional K-mean algorithm has the shortcoming that plunges into a local optimum prematurely because of sensitive selection of the initial cluster center, this paper combines the genetic algorithm and K-means algorithm and presents a genetic algorithm based on K-means clustering algorithm, the algorithm is realized using actual real number of variable length cluster center. It designs new crossover and mutation operators and uses cluster validity index DB-Index as the target function, the problem of optimizing cluster center is solved by algorithm. Compared with the previous two algorithms, this algorithm improves the clustering quality effectively, improves the global convergence rate.%传统K-均值算法对初始聚类中心敏感大,易陷入局部最优值.将遗传算法与K均值算法结合起来进行探讨并提出一种改进的基于K-均值聚类算法的遗传算法,改进后的算法是基于可变长度的聚类中心的实际数目来实现的.同时分别设计出新的交叉算子和变异算子,并且使用的聚类有效性指标DB-Index作为目标函数,该算法很好地解决了聚类中心优化问题,与之前的两种算法相比,改进后的算法改善了聚类的质量,提高了全局的收敛速度.