情境感知的移动阅读个性化推荐算法研究  

Research on context-aware mobile reading personalized recommendation algorithm

作　　者：谢修娟[1] 莫凌飞[2] 李香菊[1] 操凤萍[1] Xie Xiujuan;Mo Lingfei;Li Xiangju;Cao Fengping(Department of Computer Engineering,Chengxian College,Southeast University,Nanjing 210088;College of Instrument Science and Engineering,Southeast University,Nanjing 210018)

机构地区：[1]东南大学成贤学院计算机工程系,南京210088 [2]东南大学仪器科学与工程学院,南京210018

出　　处：《高技术通讯》2019年第7期640-647,共8页Chinese High Technology Letters

基　　金：国家自然科学基金(61603091);江苏省教育信息化研究立项课题(20180054)资助项目

摘　　要：针对移动环境下资源的个性化推荐问题,提出一种结合情境和协同过滤的移动阅读推荐算法。该算法汇聚了两阶段的情境感知思想,首先使用朴素贝叶斯方法推算用户某情境下偏好度最高的资源类别,然后对该类别下的资源,通过情境相似度的计算,过滤出仅当前情境或者与当前情境最相似的情境的“用户-资源”二维评分模型,再运用传统的基于用户的协同过滤算法产生推荐列表。抓取新浪博文进行实验测试,结果表明相同条件下所提出算法的平均绝对偏差值比其他相关算法要低,具有更高的推荐质量。Aiming at the problem of personalized recommendation of resources in mobile environment a mobile reading recommendation algorithm combining context and collaborative filtering is proposed.Two stages of context aware ness are intergrated in the algorithm.Firstly the naive Bayesian method is used to estimate the highest preference categories in a given context.Then through the calculation of context similarity the‘user resource’two dimen sional scoring model is filtered out which only belongs to the current context or the most similar context to the cur rent context.Finally the traditional user based collaborative filtering algorithm is used to generate the recommen dation list.Experiments on Sina Bowen are carried out and the results show that the mean absolute error(MAE)of the proposed algorithm is lower than the related algorithm under the same conditions and has higher recommen dation quality.

关 键 词：情境感知 协同过滤 移动推荐 贝叶斯方法 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]