摘要
为解决烟草配送中配送区域划分问题,提出了一种改进的K-means聚类算法。计算每个点的密度并取其中最大的K个点作为初始聚类中心;通过分析比较边缘点到聚类中心距离与所有点间的平均距离,在可选范围内优先考虑边缘点,以避免边缘点对整体最优性的干扰。实例分析表明,该算法有较好的全局收敛性,有效地克服了传统K-means算法收敛于局部最优点和忽视边缘点重要性的缺点。
In order to solve the problem of distribution area segmentation of tobacco distribution, an improved K-means clustering algorithm was proposed. The density of every node was calculated and K nodes which have higher densities were chosen as the initial clustering centers. Based on comparison between the distance of verge node to clustering center and aver- age distance of all nodes, the priority were given to verge node within a reasonable scope so as to avoiding the side-effect to the global optimization caused by verge nodes. The real-time case experimental result shows that the improved clustering algo- rithm has a pretty good global convergence and overcomes shortcomings of loca/ convergence as well as ignoring the impor- tance of verge nodes in the traditional approach.
出处
《物流工程与管理》
2009年第6期84-85,共2页
Logistics Engineering and Management