| 摘要: |
| 推荐系统能够有效地解决信息过载问题，其中，协同过滤（collaborative filtering，简称CF）是推荐系统广泛采用的技术之一.然而传统的CF技术存在可扩展性差、数据稀疏和推荐结果精度低等问题.为了提高推荐质量，将信任关系融合到推荐系统中，采用聚类（FCM）方法，对信任关系进行聚类.利用信任类预测用户间的隐式信任，最后将信任关系与用户-项目关系线性融合进行推荐.在Douban和Epinions数据集上的实验结果表明，与传统的基于CF、基于信任和用户项目聚类的推荐算法相比，该算法能够大幅度地改进推荐质量，提升算法的时间效率. |
| 关键词: 社交信任 聚类 协同过滤 数据稀疏 推荐 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金（6110048）；黑龙江省自然科学基金（F2016034） |
|
| Hybrid Recommendation Algorithm Based on Social Trust Clustering |
|
ZHU Jing-Hua1,2, WANG Chao1, MA Sheng-Chao1
|
|
1.School of Computer Science and Technology, Heilongjiang University, Harbin 150080, China;2.Heilongjiang Province Key Laboratory for Database and Parallel Computing, Harbin 150080, China
|
| Abstract: |
| Recommender system can solve the information overload problem effectively, and collaborative filtering (CF) is one of the techniques that is widely used in recommendation system. However, the traditional CF technology has problems such as poor scalability, sparse data, and low accuracy of recommendation results. In order to improve the quality of recommendations, this article integrates the trust relationship into the recommendation system in which the trust relationship is clustered by using the clustering (FCM) method. Using the trust cluster to predict implicit trust between users, the trust relationship is finally combined with the user-item relationship to give recommendations. The experimental results on the data set of Douban and Epinions show that compared with traditional CF algorithm, trust based recommendation algorithm and recommendation algorithm for user item clustering, the presented algorithm can greatly improve the recommendation quality and time efficiency. |
| Key words: trust relationship clustering collaborative filtering date sparseness recommendation |
