A study on multicriteria recommender system using implicit feedback and fuzzy linguistic approaches

The present Recommender systems have intrusiveness problem in its operations, provide less accuracy in recommendations and operate on uncertain nature of data. In order to make Recommender systems to provide effortless assistance along with accuracy in recommendations, a combined framework is proposed which combines the implicit relevance feedback, multicriteria ratings and fuzzy linguistic approaches. A Music Recommender System is developed as prototype model to evaluate the performance of the proposed approaches under the user-based and item-based prediction algorithms against different parameters namely data sparsity levels, training/test data ratio and neighbourhood sizes. From the experimental evaluation, it was observed that the fuzzy-implicit-multicriteria ratings based recommendation approach provides more recommendation accuracy than traditional and other recommendation approaches considered.