查询结果:   王三虎,王丰锦.融合用户评分和属性相似度的协同过滤推荐算法[J].计算机应用与软件,2017,34(4):305 - 308,321.
中文标题
融合用户评分和属性相似度的协同过滤推荐算法
发表栏目
算法
摘要点击数
696
英文标题
A COLLABORATIVE FILTERING RECOMMENDATION ALGORITHM BASED ON USER SCORE AND ATTRIBUTE SIMILARITY
作 者
王三虎 王丰锦 Wang Sanhu Wang Fengjin
作者单位
吕梁学院计算机科学与技术系 山西 吕梁 033000 同方股份有限公司 北京 100083    
英文单位
Department of Computer Science and Engineering,Lvliang University,Lvliang 033000,Shanxi,China Tongfang Co., Ltd, Beijing 100083,China    
关键词
推荐系统 协同过滤 相似性度量 稀疏性问题
Keywords
Recommendation system Collaborative filtering Similarity measurement Sparsity problem
基金项目
山西省教育厅教学改革项目(J2014120,J2015121)
作者资料
王三虎,副教授,主研领域:数据库应用技术,算法设计,数据挖掘。王丰锦,高工。 。
文章摘要
为了提高协同过滤推荐系统的推荐效率和准确性,更好地向用户提供个性化的推荐服务,提出一种用户评分和属性相似度的推荐算法。首先分析当前协同过滤推荐研究的现状,设计相似度、兴趣倾向相似度、置信度等指标作为评分标准,使得用户相似度的计算更加准确、有区分度。然后根据用户属性来衡量用户之间的相似度,利用MovieLens数据集和Book-Crossing数据集做对比实验,对比精度、通用性和不同稀疏度及冷启动情况下的性能。实验结果表明,本文算法不仅提高了推荐精度,而且明显优于其他协同过滤推荐算法,具有更高的实际应用价值。
Abstract
In order to improve the recommendation efficiency and accuracy of collaborative filtering recommendation system, and to provide personalized recommendation service, a recommendation algorithm based on user score and attribute similarity is proposed. Firstly, the current status of collaborative filtering recommendation research is analyzed, and the similarity, similarity of interest tendency, confidence and other indicators are used as the scoring criteria, which makes the calculation of user similarity more accurate and discriminative. Then the similarity between users is measured according to the attributes of the users. The comparison is made between the MovieLens data set and the Book-Crossing data set, and the accuracy, versatility and performance under different sparsity and cold start conditions are compared. Experimental results show that the proposed algorithm not only improves the recommendation accuracy, but also is superior to other collaborative filtering recommendation algorithms, and has higher practical application value.
下载PDF全文