Similarity Measures Using Fuzzified Ratings for Collaborative Filtering
Lee Soojung · Frontiers in artificial intelligence and applications · 2017
Collaborative filtering-based recommender systems have been popular for Internet users as they are helpful in searching for useful information promptly. These systems rely on similarity measures to obtain recommendations from other users based on their rating history. This paper addresses the problem of vagueness and subjectivity in user ratings. To relieve this problem, we adopt the fuzzy logic to transform ratings and then compute similarity using the fuzzified ratings. Performance of the proposed method is investigated to find that it significantly outperforms existing similarity measures using hard user ratings in terms of prediction accuracy, especially with a sparse and short-ranged dataset.