基于最大最小距离和动态隧道的聚类算法

针对K-means聚类算法对初值敏感和易陷入局部最小值的缺陷,提出了一种基于最大最小距离和动态隧道的聚类算法.该算法首先利用最大最小距离法来优选初始聚类中心以避免由于聚类中心过于随机而导致其分布较为集中的情形,以提高划分初始数据集的效率.动态隧道法具有全局寻优能力,利用钻隧过程可跳出局部极小点得到更小值点,再由K-means聚类算法对其迭代优化,如此反复直至得到全局极值.实验结果表明了该算法的可行性和有效性.

Clustering algorithm based on max-min distance and dynamic tunneling

In view of the limitations that K-means clustering algorithm is sensitive to the initial points and apt to fall into local minimum point, a novel clustering algorithm based on max-min distance and dynamic tunneling is proposed. Firstly max-min distance method is used to optimally choose the starting cluster centers in case of distributedly centralized cluster centers induced by random choice, so partition efficiency of initial dataset can be improved. Then, dynamic tunneling approach, which is capable of gl...