Abstract
Personalized recommender systems have been receiving more and more attention in addressing the serious problem of information overload accompanying the rapid evolution of the world-wide-web. Although traditional collaborative filtering approaches based on similarities between users have achieved remarkable success, it has been shown that the existence of popular objects may adversely influence the correct scoring of candidate objects, which lead to unreasonable recommendation results. Meanwhile, recent advances have demonstrated that approaches based on diffusion and random walk processes exhibit superior performance over collaborative filtering methods in both the recommendation accuracy and diversity. Building on these results, we adopt three strategies (power-law adjustment, nearest neighbor, and threshold filtration) to adjust a user similarity network from user similarity scores calculated on historical data, and then propose a random walk with restart model on the constructed network to achieve personalized recommendations. We perform cross-validation experiments on two real data sets (MovieLens and Netflix) and compare the performance of our method against the existing state-of-the-art methods. Results show that our method outperforms existing methods in not only recommendation accuracy and diversity, but also retrieval performance.
Highlights
The rapid growth of the word-wide-web has been exposing an enormous increasing amount of commodities and information to people, information overload accompanying such resources has been recently recognized as a great challenge in both business areas and academic fields [1]
Based on the above understandings and motivated by the fact that the existence of popular objects may adversely influence correct recommendations, we propose in this paper a random walk with restart model on a constructed user similarity network towards personalized recommendations
With the understanding that an effective recommender system should rank user preferred objects among top positions of a ranking list, we focused on MovieLens and the cosine similarity measure to perform the validation experiment as detailed in the section of methods, and we investigated the proportion of test objects that occupied exactly the k-th position of the final ranking lists (Pk)
Summary
The rapid growth of the word-wide-web has been exposing an enormous increasing amount of commodities and information to people, information overload accompanying such resources has been recently recognized as a great challenge in both business areas and academic fields [1]. To alleviate this problem, internet search engines have been widely utilized as a fundamental technique to PLOS ONE | DOI:10.1371/journal.pone.0114662. The keyword-based design can only provide passive filtration of overloading information and lack the capability of screening useful resources in an active way [3]. Various recommender systems have been proposed to offer personalized nomination of candidate resources by assisting individuals to efficiently filtering out overload information and positively identifying their potential interest [4], which have shown great successes in a variety of applications such as the online recommendation of books [4], CDs [5], movies [6, 7], news [8], and many other resources [9]
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
