一种改进的基于遗传算法的聚类分析方法

C－均值聚类收敛速度快，但是它容易陷入局部最优，且对初始解很敏感。遗传算法是一种全局搜索方法，但是它收敛速度慢。为了在搜索能力和收敛速度两方面都取得较好的效果，本文提出了一种改进的基于遗传算法的聚类分析方法。实验结果表明：本文提出的算法在聚类分析中搜索到全局最优解（或近似全局最优解）的能力要优于经典遗传算法及C－均值聚类算法；且通过对变异概率的巧妙设置，提高了算法的自适应能力。

Clustering with a Modified Genetically Optimized Approach

Although C-means clustering analysis has good convergence rate, it strongly depends on initialization, and it can also easily be trapped in a local optimum. Theoretically, Global optimum can be reached with Genetic Algorithm (GA), but GA converges very slowly. Therefore, a modified genetically guided algorithm (MGGA) is proposed to optimize the C-means functions used in clustering analysis to gain both good scouting performance and fast convergence rate. Experiment results show that the proposed algorithm performs better in global optimum searching than GGA proposed in 4]. Moreover, by means of smart setting of the mutation probability, the Adaptation performance of MGGA is greatly improved.