基于扰动因子的相似度下的聚类算法

Clustering algorithm under the similarity based on the disturbance factor

K-均值聚类算法在计算聚类中心时对于离群点非常敏感,且容易陷入局部极小值.针对这一缺点,采用距离法消除离群点对于聚类中心的影响,并且给搜索空间增加一组随着迭代次数递减的扰动因子,建立了基于扰动因子的相似度下的聚类算法,并对改进前后的算法进行对比实验.仿真结果表明,改进后的算法更稳定,聚类效果更好.

The clustering centers are very sensitive to outliers and easy to fall into local mini- mum when they are calculated by the classical K-means clustering algorithm. Aimed at the dis advantage, the clustering algorithm under the similarity based on the disturbance factor is es- tablished by the distance method to eliminate outliers from the cluster center for influence, and to add a set of disturbance factors which decrease with the number of iterations to searching space. Finally, the improved algorithm is compared to the classical K-means clustering algo- rithm by the experiments. The results show that the improved algorithm is more stable than before, and the clustering effect is better.