基于模拟退火的K-means算法研究

Research on K-means Algorithm Based on Simulated Annealing

针对K-means算法依赖于初始聚类中心、经常陷入局部最优解等缺点,利用模拟退火算法的全局优化特点,提出一种基于模拟退火的K-means算法。仿真结果表明该算法减弱了对初始聚类中心的依赖性,提高了原有算法的稳定性。该算法能够获得全局最优解,收敛于局部极小值点的可能性大大减少。

Aiming at the disadvantages such as K-means algorithm depending on the initial clustering center and often getting into local optimum solution,etc. ,this paper puts forward a kind of K- means algorithm based on simulated annealing making use of the overall optimization characteristic of simulated annealing algorithm. Simulation result indicates that the algorithm weakens the dependence on initial clustering center and improves the stability of former algorithm. The algorithm can get the overall optimum solution and weaken the possibility of getting into local minimum value greatly.