一种用户画像系统的设计与实现  

DESIGN AND IMPLEMENTATION OF USER PROFILE SYSTEM

作　　者：王洋 丁志刚 郑树泉[2,3,4] 齐文秀 Wang Yang;Ding Zhigang;Zheng Shuquan;Qi Wenxiu(Shanghai Institute of Computing Technology,Shanghai 200040,China;Shanghai Industrial Technology Institute,Shanghai 201206,China;Shanghai Development Center of Computer Software Technology,Shanghai 201112,China;Shanghai Embedded System Engineering Research Center,Shanghai 201112,China)

机构地区：[1]上海市计算技术研究所,上海200040 [2]上海产业技术研究院,上海201206 [3]上海计算机软件技术开发中心,上海201112 [4]上海嵌入式系统应用工程技术研究中心,上海201112

出　　处：《计算机应用与软件》2018年第3期8-14,共7页Computer Applications and Software

基　　金：上海大数据科技成果转化平台(16DZ1110101)。

摘　　要：用户画像系统通过结合用户浏览行为日志以及爬取数据作为补充,构成用户浏览行为的完整数据集。通过以Hadoop分布式集群为基础的大数据平台结合数据分析算法对该数据集进行清洗、规范化、分析与处理,分析出用户兴趣偏好,为用户标记不同权重的标签,使得企业更了解用户以及为之后针对用户精准推荐铺平道路。此外,针对K-means算法依赖初始化中心的缺陷进行了改进,从测试结果可以看出改进后的K-means准确率得到了较大提升。Abstract The user profile system constructed complete data set of users browsing behaviour by combing user browsing log and crawling data.The data set is cleaned,normalized,analysed,and processed by data analysis algorithm and a big data platform based on Hadoop distributed cluster to get interests of users and mark labels of different weights for users,which enables enterprises to understand users and paves the way of precise recommendation for users.In addition,the defect of K-means algorithm relying on the initialization centre is improved,and the improved K-means algorithm has higher accuracy.

关 键 词：用户行为分析 基于Hadoop的大数据分析平台 用户画像系统 用户价值模型 K-MEANS 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]