基于用户画像的二分图神经网络移动互联网产品推荐模型
Bipartite graph neural network product recommendation model for mobile internet based on user portrait
王浩 1范嘉俊 2江昊2
作者信息
- 1. 武汉第二船舶设计研究所,湖北武汉 430064
- 2. 武汉大学电子信息学院,湖北武汉 430074
- 折叠
摘要
产品推荐模型可帮助用户从海量的互联网数据中快速定位感兴趣的产品.针对目前的产品推荐模型无法同时做到精确化、个性化和动态化推荐的问题,提出一种基于用户画像的二分图神经网络移动互联网产品推荐模型.该模型以用户一段时间内的产品选购记录为输入,将移动互联网中的用户和产品构建为二分图,并通过图神经网络对已知的用户选购产品行为以及各用户间的选购行为联系进行优化,以得到网络嵌入,进而构建用户画像,预测用户之后可能选购的产品.在联通用户详单数据集上进行用户未来流量需求预测实验,并就用户实际增购流量包的大小进行预测,与协同过滤法进行对比,验证了所构建模型的优秀性能.
Abstract
The product recommendation model can help users quickly locate products of interest from massive in-ternet data.To solve the problem that the current product recommendation model cannot achieve accurate,person-alized and dynamic recommendation at the same time,a bipartite graph neural network product recommendation model for mobile internet based on user portrait is proposed.This model takes the product purchase records of users in a period of time as input,constructs the user and the product in the mobile internet as a bipartite graph,and optimizes the known purchase behaviour of users and the relationship in the purchase behavior among users through graph neural network to obtain network embedding,and then constructs the user portrait to predict the products that users may purchase later.The future flow demand prediction experiment is carried out on the detailed data set about user of China Unicom,and the actual additional flow packet size of users is predicted.Compared with the collaborative filtering method,the excellent performance of the model proposed in this paper is verified.
关键词
用户画像/图神经网络/产品推荐Key words
user portrait,graph neural network/product recommendation引用本文复制引用
出版年
2024
NETL
NSTL
万方数据