An Integrated Recommendation Approach Based on Influence and Trust in Social Networks
In real human society, influence on each other is an important factor in a variety of social activities. It is obviously important for recommendation. However, the influence factor is rarely taken into account in traditional recommendation algorithms. In
- PDF / 720,542 Bytes
- 7 Pages / 439.363 x 666.131 pts Page_size
- 69 Downloads / 227 Views
School of Computer Engineering and Technology, Shanghai University, Shanghai, China Shanghai Key Laboratory of Computer Software Evaluating and Testing, Shanghai, China 3 Graduate School of Human Sciences, Waseda University, Tokorozawa, Japan {wmli,yezb}@shu.edu.cn, [email protected]
Abstract. In real human society, influence on each other is an important factor in a variety of social activities. It is obviously important for recommendation. However, the influence factor is rarely taken into account in traditional recommendation algorithms. In this study, we propose an integrated approach for recommendation by analyzing and mining social data and introducing a set of new measures for user influence and social trust. Our experimental results show that our proposed approach outperforms traditional recommendation in terms of accuracy and stability. Keywords: recommendation algorithms, similarity, influence, social trust.
1
Introduction
In recent years, recommender systems have been an important tool to help people to get information that meets their needs or interests from the mass data. Traditional recommendation algorithms generally make use of the users’ behavior data or attribute data. As one of the most successful recommendation algorithms in commercial domains, collaborative filtering is designed through computing the users’ rating data [1]. However, it still suffers from some drawbacks such as cold-start, sparsity and scalability problems. Generally speaking, people tend to accept recommendations from familiar or trusted persons [2], and their opinion on a certain thing can greatly affect their friends’ choice. With the development of online social networks, it becomes easier to collect and utilize the users’ social relationship data. Some e-commerce companies examine how to leverage social relationships to improve the customers' purchase decision making so as to increase sales [3]. Researchers introduce a variety of social information to solve sparsity problems [4] and achieve better results. Modeling trust is mostly through the users’ relationships. On the other hand, the users’ social importance is another critical feature and can be used. User influence plays an important role in product marketing. He and Chu proposed a social recommendation system using user influence as a factor, and their experimental results on a dataset collected from www.yelp.com proved its superiority over collaborative filtering [5]. But such recJames J. (Jong Hyuk) Park et al. (eds.), Future Information Technology, Lecture Notes in Electrical Engineering 309, DOI: 10.1007/978-3-642-55038-6_13, © Springer-Verlag Berlin Heidelberg 2014
83
84
W. Li, Z. Ye, and Q. Jin
ommendation systems based on user influence are difficult to provide consistent and stable answers for active user’s constantly changing query. Based on the above background, this work studies personalized recommendation on social media, such as www.yelp.com and www.dianping.com, which consist of the users’ rating data and relationships. We propose a novel recommenda
Data Loading...
