一种基于隐式反馈的用户兴趣模型构建方法  

A Construction Method of User Interest Model Based on the Implicit Feedback

作　　者：赵文朋 丁艳辉[1,2] 金连旭 张骏 ZHAO Wenpeng;DING Yanhui;JIN Lianxu;ZHANG Jun(School of Information Science and Engineering,Shandong Normal University,Jinan 250358;Shandong Provincial Logistics Optimization and Predictive Engineering Technology Research Center,Jinan 250358)

机构地区：[1]山东师范大学信息科学与工程学院,济南250358 [2]山东省物流优化与预测工程技术研究中心,济南250358

出　　处：《计算机与数字工程》2018年第3期518-522,共5页Computer & Digital Engineering

基　　金：国家自然科学基金青年项目(编号:61303007);山东省优秀中青年科学家科研奖励基金项目(编号:BS2013DX044)资助

摘　　要：个性化推荐系统通常侧重用户显式反馈数据的收集和利用,而对于大量的隐式反馈数据,有待于进一步的充分利用。本文提出一种基于隐式反馈的用户兴趣模型构建方法,该方法首先从用户浏览行为方面分析用户兴趣度,建立用户在页面浏览时间、浏览速度、保存页面与收藏页面等经典隐式反馈行为与兴趣度之间的联系;然后,基于隐式反馈行为,提出主题兴趣度的概念,并将其与传统的基于VSM的兴趣模型相结合,构建基于主题兴趣度的用户兴趣。实验表明,本文提出的用户兴趣模型在推荐准确率上有明显提高。Personalized recommendation system usually focus on the collection and utilization of user??explicit feedback data.However,a large amount of implicit feedback data needs further utilization.In this paper,a construction method of user interest model based on the implicit feedback is proposed.Firstly,the method analyzes user interest degree from the aspect of user browsing behavior,that is to say,the method creates a relationship between interest degree and page browsing time,browsing speed,saving pages,collecting pages and other classic implicit feedback behaviors.Then,based on the implicit feedback behavior,the concept of theme interest degree is proposed and it is combined with the traditional VSM based interest model to construct user interest based on theme interestingness.Experiments show that the proposed user interest model has a significant improvement in recommendation accuracy.

关 键 词：隐式反馈 用户浏览行为 用户兴趣模型 推荐系统 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]