摘要
传统的基于用户的协同过滤推荐算法在计算用户间相似性时依赖于用户-项目评分矩阵,但在实际的商业系统中,用户参与的评价往往非常少,这样计算出的相似性精确度通常很低。文中提出结合用户相似性和基于项目分类特征的相似性计算方法,计算用户间的相似性,形成目标用户的近邻集合,完成向目标用户的推荐。文中在MovieLens数据集上的实验结果表明,相对于Pearson相似性的协同过滤推荐算法,文中提出的改进算法在推荐质量方面有明显提高。
The traditional user-based collaborative filtering algorithm calculates users＇ similarity according to user-item rating matrix, but in real business system, the user-ratings data is very sparse, so the calculation accuracy is very low, The calculation method mixing user similarity and project classification features based similarity is proposed for similarity calculation between users, get target user＇ s close neighbor set,calculate recommended results. The experimental results on the MovieLens data set show that, compared with the Pearson similar collaborative filtering algorithm, the above improved algorithm raises the recommendation quality significantly.
出处
《计算机技术与发展》
2013年第9期66-69,共4页
Computer Technology and Development
基金
国家自然科学基金资助项目(61170060)
安徽省自然科学基金(11040606M135)
安徽省高等学校自然科学基金重点项目(KJ2011A083)
关键词
基于用户
协同过滤
推荐系统
项目分类
相似性计算
user-based
collaborative filtering
recommendation system
project classification
similar calculations
