基于用户情境和时序多样性的协同过滤优化算法  

An optimization collaborative filtering algorithm based on user context and temporal diversity

作　　者：林皓[1] 陈莉[1] 兰小艳[1] 张珏[1] 尹化荣 

机构地区：[1]西北大学信息科学与技术学院,陕西西安710127

出　　处：《西北大学学报（自然科学版）》2017年第6期822-828,共7页Journal of Northwest University（Natural Science Edition）

基　　金：国家科技支撑计划课题基金资助项目(2013BAH49F02)

摘　　要：目前针对协同过滤算法的主要研究工作集中在如何更准确地预测用户对未知商品的评分,然而衡量推荐质量的标准并不限于预测准确率,推荐列表的多样性也是诸多衡量因素之一。该文将用户情境引入到传统协同过滤推荐算法中,在评分预测阶段考虑用户所处的内部情境,对预测评分进行加权,以保证推荐结果符合用户偏好;在生成推荐阶段,基于外部情境对推荐候选项加权重排,提高推荐列表的时序多样性。实验结果表明,与其他多样性优化方法相比,基于用户情境的方法在保持推荐准确率的同时,能够有效提高系统推荐结果的时序多样性。Almost all the existing research on collaborative filtering algorithm concentrates on improving the prediction accuracy of rate,which is used to measure users' preferences for unknow items. However,accurate prediction of rating values is not the only indicator of verifying recommendation system ability,the diversity of recommendation list has the same importance to improving the recommendation quality and user experience. In this paper,users' context imformation was used to optimize the traditional item-based collaborative filtering algorithms. At the stage of rates prediction,the inner information was taken into consideration to modify the rating values,and then adjusted the recommendation list based on the external context. The goal of this effort is to improve the temporal diversity of recommendation list. Results of comparative experiment show that the proposed algorithm can effectively improve the temporal diversity of recommendation list while maintaining a certain level of prediction accuracy.

关 键 词：协同过滤 时序多样性 用户情境 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]