Authors: Saman Izadpanah; S. Kami Makki
Addresses: Department of Computer Science, Lamar University, 211 Red Bird Lane, P.O. Box 10056, Beaumont, TX 77710, USA ' Department of Computer Science, Lamar University, 211 Red Bird Lane, P.O. Box 10056, Beaumont, TX 77710, USA
Abstract: The growth of social networking generates a huge amount of user data. The interested cooperation or businesses search through this data to make real-time business decisions or generate meaningful business patterns from users' behaviour. They also build new recommender systems to predict intelligently future social trends. These tools are becoming the basis for efficient operations of e-commerce and improvement of companies' productions. For example, the recommendation systems used by many commercial companies utilise collaborative filtering techniques such as k-nearest neighbour to enhance their service delivery and determine what their users most like. Using k-nearest neighbour allows the system to compare a primary user with all others and determines how similar their interests are to primary user. This creates a neighbourhood list, consisting of every user's similarity to the primary user. From this list, it is easy to select the primary user's most similar, or nearest neighbour. This nearest neighbour will then provide the basis for the primary user's recommendations. In this paper, we present a realistic method to process large volumes of data in various formats collected from different sources for recommending bookmarks by utilising one-class collaborative filtering approach.
Keywords: bookmarks; collaborative filtering; k-nearest neighbour; k-NN; implicit; personalisation; recommender systems; social networking; recommendation systems; user similarity; user behaviour; big data.
International Journal of Big Data Intelligence, 2016 Vol.3 No.4, pp.250 - 256
Received: 13 Apr 2014
Accepted: 22 Aug 2014
Published online: 23 Oct 2016 *