An international group of researchers have developed a new algorithm for recommender systems designed to address the challenge of diversity when making recommendations to users.
The researchers say that people are most interested in recommendations and information originating from users that are somewhat similar but different enough that they can introduce something new. To widen the field of user interest, the authors developed a hybrid of two algorithms. One was based on recommendations from random walks between highly connected users and material. The other mimicked the process of heat diffusion, spreading ratings at a decreasing level of potency as the recommendations came from farther away. By combining the heat diffusion approach with the more accurate random walk, the researchers found that they could establish a body of recommendations that combined novelty items with safer choices.
In addition, combining both algorithms allowed for more accurate recommendations than using either alone.
From Ars Technica
View Full Article
Abstracts Copyright © 2010 Information Inc., Bethesda, Maryland, USA
No entries found