摘要
针对K均值算法在聚类分析中还存在对初始值异常的敏感、容易陷入局部最优等问题,本文提出了一种基于相似度计算优化和指标加权优化的改进算法,首先采用欧式距离度量方式对传统K均值算法中的相似度计算进行优化,避免其进入局部最优,然后根据实际问题对聚类的指标进行加权优化。实例仿真试验结果表明,本文提出的基于相似度计算优化和指标加权优化的改进K均值算法在烟草零售终端数据聚类分析中效果良好,有利于提高货源投放的精准化。
According to the defects such as sensitive to initial value abnormal and easy to fall into local optimum of K-means algorithm in clustering analysis, this paper puts forward a kind of improved algorithm based on the optimization of similarity calculation and index weighted. First, similarity calculation in traditional K-means algorithm is optimized by Eu-clidean distance measure, avoiding it into the local optimal. And then weighted optimize clustering indicators according to the practical problem. Instance simulation test results show that, the proposed improved K-means algorithm based on based on the optimization of similarity calculation and index weighted has good effect in clustering analysis of tobacco retail terminal data, is conducive to improve the delivery of precision.
出处
《科技通报》
北大核心
2015年第12期254-256,共3页
Bulletin of Science and Technology
关键词
K均值算法
烟草零售终端
聚类分析
货源投放
欧氏距离
K-means algorithm
the tobacco retail terminal
clustering analysis
supply delivery
Euclidean distance