This paper uses an improved criterion based on K-means clustering algorithm applied in electric power custom?er clustering research. According to the characteristics of electricity customers to implement different marketing strategies and pro?vide differentiated services,accurate grouping of power customer need to be made. Traditional K-means clustering algorithm in data distribution uniform data of similar spherical agglomeration effect is better,once the unbalanced distribution density of data sets, class cluster size have significant difference,while the traditional K-means algorithm is easy to make thin categories carved up by high density small class clusters,resulting in electricity customer segmentation correct rate. This paper uses an improved K-means clustering algorithm based on the characteristics of the unbalanced data distribution of the actual power customers. Improved K-means algorithm puts up with a new weighting criteria,and modifies the clustering iterative process based on the criteria. The electricity customer data cluster results show that the improved K-means clustering algorithm and the cluster effect of each group of compactness can be improved effectively. The classification error conditions are improved obviously.%针对电力客户特点实行不同的营销策略和提供差异化服务,就需要对电力客户做出准确的分群.传统K-means聚类算法对数据分布均匀的类似球形的数据集聚类效果比较好,一旦数据集分布密度不均衡,类簇大小差异明显时,传统K-means算法容易使稀疏的大类簇被高密度小类簇瓜分,导致电力客户分群正确率下降.论文基于电力客户数据分布不均衡的特点,采用了一种改进的K-means聚类算法.改进的K-means算法提出一个新的加权聚类准则,并根据该准则修改了聚类迭代过程.文章最后在对电力客户数据的分群聚类结果表明,改进的K-means聚类算法的分群聚类效果中各个群类的紧凑性得到有效提高,误分情况明显改善.
展开▼