基于Kmeans++算法和LGBM模型的城市燃气客户分类
Classifying urban gas customers based on Kmeans++ algorithm and LGBM model
兰志轩 1王世柱 2曹译丹 3杨楠 1李宏晨1
作者信息
- 1. 昆仑数智科技有限公司,北京 102206
- 2. 聊城金捷燃气有限公司,山东 聊城 252000
- 3. 昆仑燃气有限公司天津分公司,天津 300457
- 折叠
摘要
为了方便天然气销售公司进行客户细分及客户关系管理,使销售公司能够以客户为中心进行个性化服务与专业化营销,通过收集天然气销售公司的城市燃气客户消费行为特征和消费动机特征数据,对客户进行了聚类分析研究,提出了利用Kmeans++算法聚类及训练LGBM分类模型实现对新进入客户的判别归类,形成了客户多维分类模型.研究结果表明:①针对某天然气销售公司中数量多且用气规模大的城市燃气客户进行客户细分研究,以手肘法辅助的Kmeans++聚类模型对采集的城市燃气客户消费数据进行聚类分析,聚类效果较好,各类客户群的业务特征较为明显;②构建LGBM分类模型实现对新进入客户的划分,LGBM分类模型准确率较高,对客户分类的结果与聚类得到的几大类客户特征较符合.
Abstract
In order to help natural-gas sales companies classify their urban gas customers and manage customer relation as well as offer the customer-oriented personalized services or professional marketing,some characteristics data about con-sumption behaviors and motivations of the customers were gathered from these companies to perform customer clustering analysis.The Kmeans++ clustering algorithm and the trained LGBM classification model were presented to discriminate new customers and further classify them,forming a multi-dimensional customer classification model.Results show that(ⅰ)for classifying one certain sales company's urban gas customers which are great in quantity and consumption,assisted with the elbow method,the Kmeans++ model can be used to cluster and analyze the gathered consumption data on these cus-tomers.The clustering enjoys better effect,and behaviour characteristics are more evident in various customer base;and(ⅱ)with higher accuracy,the LGBM model can sort out new customers.And the classification is in good agreement with the characteristics of major customer categories obtained from the Kmeans++ clustering.
关键词
城市燃气客户/客户分类/天然气/数字化转型/客户聚类Key words
Urban gas customer/Customer classification/Natural gas/Digital transformation/Customer clustering引用本文复制引用
出版年
2024
NETL
NSTL
万方数据