Utilizing user tag-based interests in recommender systems for social resource sharing websites

  • Authors:

  • Cheng-Lung Huang;Po-Han Yeh;Cheng-Wei Lin;Den-Cing Wu

  • Affiliations:

  • Department of Information Management, National Kaohsiung First University of Science and Technology, 2, Juoyue Rd., Nantz District, Kaohsiung 811, Taiwan;Laboratory of Business Intelligence and Data Mining, National Kaohsiung First University of Science and Technology, 2, Juoyue Rd., Nantz District, Kaohsiung 811, Taiwan;Laboratory of Business Intelligence and Data Mining, National Kaohsiung First University of Science and Technology, 2, Juoyue Rd., Nantz District, Kaohsiung 811, Taiwan;Laboratory of Business Intelligence and Data Mining, National Kaohsiung First University of Science and Technology, 2, Juoyue Rd., Nantz District, Kaohsiung 811, Taiwan

  • Venue:
  • Knowledge-Based Systems
  • Year:
  • 2014

Quantified Score

Hi-index 0.00

Visualization

Abstract

Recently collaborative tagging, also known as ''folksonomy'' in Web 2.0, allows users to collaboratively create and manage tags to classify and categorize dynamic content for searching and sharing. A user's interest in social resources usually changes with time in such a dynamic and information rich environment. Additionally, a social network is one innovative characteristic in social resource sharing websites. The information from a social network provides an inference of a certain user's interests based on the interests of this user's network neighbors. To handle the problem of personalized interests changing gradually with time, and to utilize the benefit of the social network, this study models a personalized user interest, incorporating frequency, recency, and duration of tag-based information, and performs collaborative recommendations using the user's social network in social resource sharing websites. The proposed method includes finding neighbors from the ''social friends'' network by using collaborative filtering and recommending similar resource items to the users by using content-based filtering. This study examines the proposed system's performance using an experimental dataset collected from a social bookmarking website. The experimental results show that the hybridization of user's preferences with frequency, recency, and duration plays an important role, and provides better performances than traditional collaborative recommendation systems. The experimental results also reveal that the friend network information can successfully collaborate, thus improving the collaborative recommendation process.