您的位置:首页  > 论文页面

一个基于用户画像的商品推荐算法的设计与应用

发表时间:2018-02-28  浏览量:2419  下载量:499
全部作者: 吕超,朱郑州
作者单位: 北京大学软件与微电子学院
摘 要: 以用户画像的商品推荐为研究背景,在真实的电商环境下,使用海量数据构建用户画像和商品画像。对画像数据进行详细探讨,并针对不同的优化目标对训练目标中使用的标签数据进行区分和说明,实现了一个线上栏目的商品推荐算法。详细阐明用户画像和商品画像的构建方案,以说明关联用户和商品交互信息的标签数据的获取。定义分别以销量和点击率为优化目标,并分别给出了这两种优化目标的不同标签构建方案和实现方法。在构建用户画像、商品画像和标签数据后,详细说明特征构建的过程;在构建特征数据和优化目标后,详细阐明基于机器学习中梯度提升决策树的模型训练过程和参数调整方案得到的分数预测模型。最后利用训练好的分数预测模型、新用户列表和推荐的商品召回列表进行商品推荐。在完成商品推荐算法的设计和实现后,对模型进行应用和验证。研究结果证实,本文算法相比人工排序能获得更好的结果。
关 键 词: 计算机应用;用户画像;电子商务;梯度提升决策树;推荐算法
Title: Design and application of a commodity recommendation algorithm based on user portrait
Author: LÜ Chao, ZHU Zhengzhou
Organization: School of Software & Microelectronics, Peking University
Abstract: Based on the user portrait of the commodity recommendation as the research background, in the real electricity supplier environment, massive data is used to build user portrait and commodity portrait. With detailed discussion on the portrait data, the used label data of the training target for different optimization goals are distinguished and described, then an online column recommendation algorithm is implemented. This paper expounds the construction scheme of the user portrait and commodity portrait, and then explains the acquisition of the tag data of the interactive user and commodity interactive information. This paper defines the optimization target of sales volume and click rate respectively, and gives the different label construction schemes and realization methods of these two optimization targets. After the construction of user portrait, commodity portrait and tag data, the process of characteristics construction is described in detail. After the construction of characteristics data and the optimization goal, the score prediction model is elaborated based on the model training process and parameter adjustment method of gradient boosted decision tree in machine learning. Finally, the training score prediction model, new user list and recommended commodity recall list are used for commodity recommendation in this paper. After completing the design and implementation of the commodity recommendation algorithm, the model is applied and verified. The results show that the algorithm in this paper can get better results than the manual sorting.
Key words: computer applications; user portrait; E-commerce; gradient boosted decision tree; recommendation algorithm
发表期数: 2018年2月第4期
引用格式: 吕超,朱郑州. 一个基于用户画像的商品推荐算法的设计与应用[J]. 中国科技论文在线精品论文,2018,11(4):339-347.
 
0 评论数 0
暂无评论
友情链接