基于项目属性与数据权重的协同过滤推荐算法  

Collaborative Filtering Recommendation Algorithm Considering Data Weight and Item Attributes

作　　者：张新猛 李松[2] 

机构地区：[1]天津市体育彩票管理中心,天津300074 [2]河北工业大学计算机科学与软件学院,天津300401

出　　处：《自动化与仪表》2016年第9期69-73,共5页Automation & Instrumentation

摘　　要：传统的基于项目的协同过滤(IBCF)算法存在相似性计算方法忽略项目属性在相似性度量中的重要参考价值和不能反映用户意向的变化的不足。基于以上不足该文提出了一种新的基于项目的协同过滤算法。此种算法分别基于评分与项目属性因素度量项目相似性,最后通过自适应平衡因子协调处理两方面的相似性结果来度量项目相似,并结合反映项目与用户意向相关程度的数据权重进行预测与推荐。试验结果表明,改进后的算法,对传统算法中存在的2个弊端进行了改善,在推荐准确度上有所提高。The traditional collaborative filtering（IBCF） algorithm based on the project of deficiencies ：firstly tradition- al collaborative filtering algorithm based on item similarity calculation methods ignore the important reference value of item property in similarity measure,secondly traditional collaborative filtering algorithm based on the project do not reflect the user＇s interest changes ;Based on the above two inadequate this paper proposes a new collaborative filter- ing algorithm based on the project. Such algorithms respectively based on the score and the project properties simi- larity to measure similarity,at last,an adaptive balance factor coordination processing results to measure similarity of two aspects of similar projects,and combined with the data weight of the project and the user＇s interest in different project weights to forecast and recommendations. The experimental results show that the improved algorithm,two dis- advantages that exist in the traditional algorithm for improved,on the recommendation accuracy improved.

关 键 词：推荐系统 协同过滤 项目属性 相似性 数据权重 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]