摘要
传统的协同过滤推荐算法倾向于热门项目的推荐,而对处于长尾位置的冷门项目推荐能力不足;针对这一问题,提出了一种新的缓解推荐偏好的协同过滤TopN推荐算法,该算法充分考虑用户活跃程度的影响,根据用户活跃度不同化作相应比重引入项目相似度计算以及推荐列表过程中;考虑项目种类多样的影响,对得出的项目相似度矩阵经行归一化处理,实验结果表明,改进后的算法推荐准确度有明显的提升,且降低了平均流行度提高了推荐结果的新颖度,大大提高了冷门项目的推荐力度。
The traditional collaborative filtering recommendation algorithms are inclined to recommedation of popular items,but the ability of recommendation of unpopular items in long tail position are insufficient.To solve this problem proposed a new relief recommended preferences of collaborative filtering TopN recommendation algorithm.The algorithm takes full account of influence of users active degree,according to user activity turned into the corresponding proportion of introduced item similarity calculation and recommendation list process.Considering the diversity of the impact of the project,that item similarity matrix of the normalized processing.Experimental results show that the improved algorithm recommended accuracy has improved significantly,and reduce the average epidemic degree improve the recommendation results Novelty,greatly improving the strength of recommendation unpopular project.
出处
《淮南职业技术学院学报》
2016年第1期5-9,共5页
Journal of Huainan Vocational Technical College
基金
安徽省自然科学基金(项目编号:1408085QE94)
关键词
协同过滤
冷门项目
平均流行度
collaborative filtering
unpopular projects
average popularity
