首页 | 官方网站   微博 | 高级检索  
     

基于国网商旅大数据融合背景的用户画像构建
引用本文:张长浩,余志勇,周振,石瑞杰,王新勇.基于国网商旅大数据融合背景的用户画像构建[J].电信科学,2019,35(12):148-154.
作者姓名:张长浩  余志勇  周振  石瑞杰  王新勇
作者单位:国网电子商务有限公司,北京100053;国家电网有限公司,北京100031
摘    要:基于大数据分析的商旅计划决策是掌控差旅动态、制定差旅规范的重要组成部分。基于国网商旅信息数据,针对出差过程中出行方式的优化选取、酒店住宿的个人喜好,构建一种用户画像框架技术,实现快速、准确识别敏感客户群体。首先针对用户不同特性采用双通道建模方式预测用户敏感程度;其次围绕业务审批、差旅控制、酒店评价、时间特征、数值特征等类型刻画用户,构建用户多源特征体系;最后充分利用商旅数据多源性,创建基于双层XGBoost的多视角融合模型,提升分类精确率,并通过对比实验验证方法的有效性。

关 键 词:用户画像  多视角学习  模型融合

User image construction framework based on big data fusion background for power grid business travel
Changhao ZHANG,Zhiyong YU,Zhen ZHOU,Ruijie SHI,Xinyong WANG.User image construction framework based on big data fusion background for power grid business travel[J].Telecommunications Science,2019,35(12):148-154.
Authors:Changhao ZHANG  Zhiyong YU  Zhen ZHOU  Ruijie SHI  Xinyong WANG
Affiliation:1. State Grid E-commerce Co.,Ltd.,Beijing 100053,China;2. State Grid Corporation,Beijing 100031,China
Abstract:Business travel planning decisions based on big data analysis are an important part of controlling travel dynamics and developing travel specifications.Based on the information of the national network business travel information,aiming at the optimization of the travel mode during the travel process and the personal preference of hotel accommodation,a user portrait framework technology was constructed to realize the rapid and accurate identification of sensitive customer groups.Firstly,the user's sensitivity was predicted by two-channel modeling for different characteristics of users.Secondly,users were characterized by business approval,travel control,hotel evaluation,time characteristics,numerical characteristics,etc.,and users’ multi-source feature system was constructed.Finally,business travel data was fully utilized.Multi-source,create a multi-view fusion model based on double-layer XGBoost,improve classification accuracy,and verify the effectiveness of the method by comparing experiments.
Keywords:user portrait  multi-view learning  model fusion  
本文献已被 万方数据 等数据库收录!
点击此处可从《电信科学》浏览原始摘要信息
点击此处可从《电信科学》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号