e) H a b s t r a c ta r t i c l e i n f o
Received in revised form 8 June 2015
Accepted 20 July 2015
Available online 26 July 2015
Users' attributes 1. Introduction media, a network is small with only a limited number of users with an ecision of whether to allenge, social media tion systems, such as ook and other similar
Decision Support Systems 79 (2015) 46–54
Contents lists available at ScienceDirect
Decision Supp j ourna l homepage: www.eU.S.A., have grown rapidly with innovative systems and tools in recent years. High-quality friend recommendation is crucial to the survival and growth of those social media services. At the early stage of social recommendation services from Twitter, which may assist users to make better decisions .
There is a stream of literature that focuses on the recommendingand Fritsch 2013), IS (e.g., Aral, et al., 2013), healthcare (e.g., Coustasse and Slack 2013; Lin and Vaska 2013; Yang and Yang 2013), and the public sector (Davies and Cairncross 2013; Kolb and Roberts 2013) [25–34]. Social network services, such as Facebook and Twitter in the millions of other users' homepages to make a d choose a potential friend. To meet this new ch providers began to design friend recommenda the “People You May Know” system on FacebResearch on social media has become more important, attracting research from scholars of different business disciplines, such as marketing (e.g., Kumar et al., 2013), strategy (Bharadwaj et al., 2013), human resources (Urban and Boscolo 2013), finance (e.g., Røssvoll accountable number of friends; it is easy to browse over all or many of other users' profiles to make decisions of whether to choose some users as friends. Currently, thenumber of socialmedia users has reached a very high level. In 2013, the number of users from Facebook reached 1.19 billion worldwide. It seems infeasible for a user to browse over⁎ Corresponding author.
E-mail addresses: email@example.com (Z. Zhan (Y. Liu), firstname.lastname@example.org (W. Ding), email@example.com ( firstname.lastname@example.org (Q. Su), Ping.Chen@umb.edu (P. C http://dx.doi.org/10.1016/j.dss.2015.07.008 0167-9236/© 2015 Elsevier B.V. All rights reserved.as an important emerging component of social media, may efficiently expand social media networks by proactively recommending new and potentially high-quality friends to users. Literature review has shown that prior research work on friend recommendation mainly focuses on the linking relation between users in social media but largely neglects the influence of users' attributes. In this study, we have systematically reviewed and evaluated the existing state-of-the-art friend recommendation algorithms. We introduce a new Friend
Recommendation system using a User's Total Attributes Information (FRUTAI) based on the law of total probability. The proposed method can be easily extended according to the increasing number of a user's attributes with low computation cost. Furthermore, the FRUTAI is a universal friend recommendation method and can be applied in different types of social media because it does not distinguish the structure of the network.
We have collected 7 million users' public information and their friend relationships from RenRen, commonly regarded as the Facebook of China. Using the real-world data from a dominant social media provider, we extensively evaluate the proposed method with other existing friend recommendation algorithms. Our experimental results have demonstrated the comparatively better performance of FRUTAI. In our empirical studies, we have observed that the performance of FRUTAI is related to the number of a user's friends. In particular, when a user has a small number of friends, the proposed FRUTAI algorithm performs better than other algorithms; when a user has a large number of friends, the overall performance of FRUTAI becomes less competitive but is still comparable to those of other providers, and its precision rate is quite outstanding. Our findings may provide some important practical implications to social media design and performance. © 2015 Elsevier B.V. All rights reserved.Article history:
Received 4 April 2014
Socialmedia, such as Facebook and Twitter, have grown rapidly in recent years. Friend recommendation systems,Proposing a new friend recommendation m social media providers' performance
Zhou Zhang a, Yuewen Liu a,⁎, Wei Ding b, Wei (Wayne a School of Management, Xi'an Jiaotong University, Xi'an, 710049, China b Department of Computer Science, University of Massachusetts, Boston, MA, 02125, USA c Department of MIS, College of Business, Ohio University, Athens, OH, USAg), email@example.com
W.(W.) Huang), hen).thod, FRUTAI, to enhance uang a,c, Qin Su a, Ping Chen b ort Systems l sev ie r .com/ locate /dssmodels, named link prediction models . These link prediction models are useful to predict the extent of the network by observed data and play a role as a basic question in social media structure. The possibility of connection also reflects the “quality of connection” between two users in the future. If there is a high possibility that a tie and content-based algorithms [1,19].
Adomavicius and Tuzhilin present an overview of three recommentory recommendation list. Second, it will take a long time for a system to 47Z. Zhang et al. / Decision Support Systems 79 (2015) 46–54will connect two users, this connectionwill be a strong tie, whichmeans more similarity between them. The research of link prediction has both theoretical and practical values.
Existing friend recommendation methods and algorithms are, in principle, based on two different approaches—a path-based method and a friend-of-friend method [16,18]. The path-based method uses friend linkage information by implementing the concept of the wellknown PageRank algorithm from Google. Due to its high computational cost, this type of algorithm is seldom used in commercial social media.