An Improved Top-N Recommendation for Collaborative Filtering

Information overload has become an increasingly important problem in the Internet era. The recommendation system has attracted great attention, since it can offer different users with personalized recommendations. Traditional collaborative filtering approaches, user-based collaborative filtering approaches and item-based collaborative filtering approaches tend to recommend some popular items. In order to solve this problem, we propose an algorithm to make penalty on the influence of the highly active users and highly popular items. Experimental results from the MovieLens and real-world dataset, show that our approaches improves precision and coverage, while decreasing the average popularity.