面向企业营销的全景用户画像与模型预测

发布时间：2018-01-06 23:16

本文关键词：面向企业营销的全景用户画像与模型预测　出处：《浙江大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着网络和信息技术的不断发展,数据在业务处理基础上不断积累,我们从信息技术时代进入了数据技术时代。企业营销方式也从Product,Price,Place,Promotion 这 4P 理论转向了 Consumer,Cost,Convenience,Communication 这 4C理论,以用户为中心的精准营销是企业所需。但是现在的企业对用户的认知不清晰,用户信息不全,为了完善企业对用户的认知,本文将研究中心聚焦在全景用户画像和模型预测上,并结合KTV线上到线下的实际场景,最终实现企业的精细化运营。本文的工作主要包括以下方面:1.本文设计了一套分布式的处理框架。本文用Hadoop分布式文件系统和Hive实现数据分布式的存储和管理;用Impala系统实现用户画像的构建;用Spark集群实现模型预测;最终实现分布式的数据存储、管理和分析。2.本文实现了基于多源数据融合的用户画像构建。本文从内外部数据打通,多维度业务数据打通,多方位属性粒度等特性设计用户画像;通过Impala SQL直接获取、统计变换、自然语言处理、正则匹配、规则判定、用户事件模型等方式实现用户画像;最终企业利用用户画像实现对用户的了解,并能够满足企业营销业务。3.本文实现了由迭代决策树和线性模型融合的模型混合方式。该方法利用迭代决策树实现特征的自动发现,利用树的路径扩充特征向量,并结合线性模型提高模型的精度。本文将该方法应用到用户性别分类和用户消费额度预测模型中,并设计多种方案包括随机森林、迭代决策树等进行实验对比,验证了该方案的有效性和精确性。
[Abstract]:With the continuous development of network and information technology, data is accumulated on the basis of business processing. We have entered the data technology era from the information technology era. The 4P theory of Pricegne place Promotion turns to Convenience. Communication 4C theory, user-centered precision marketing is required by enterprises, but now the enterprise is not clear about users, user information is not complete. In order to perfect the enterprise's cognition to the users, this paper focuses on the panoramic user portrait and model prediction, and combines the actual scene of KTV online to offline. Finally realize the fine operation of the enterprise. The work of this paper mainly includes the following aspects:. 1. This paper designs a set of distributed processing framework. This paper uses Hadoop distributed file system and Hive to realize data distributed storage and management; Using Impala system to realize the construction of user portrait; Using Spark cluster to realize model prediction; Finally, distributed data storage, management and analysis. 2. This paper realizes the construction of user portrait based on multi-source data fusion. Multidirectional attribute granularity and other characteristics design user portrait; The user portrait is realized by Impala SQL, statistical transformation, natural language processing, regular matching, rule determination, user event model and so on. Finally the enterprise uses the user portrait to realize the understanding to the user. And can satisfy the enterprise marketing business. 3. This paper realizes the hybrid model by iterative decision tree and linear model fusion. This method uses iterative decision tree to realize automatic feature discovery. Using the path of tree to expand the eigenvector and combine the linear model to improve the accuracy of the model, this paper applies this method to the user gender classification and user consumption quota prediction model, and designs a variety of schemes, including random forest. The effectiveness and accuracy of the scheme are verified by comparing the iterative decision tree.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F274;TP311.13

【参考文献】