A method for hybrid personalized recommender based on clustering of fuzzy user profiles
Shan Xu, Junzo Watada · 2014
Personalized Recommenders can help to find potential items and then recommend them for particular users. Conventional recommender methods always work on a rating schema that items are rated from 1 to 5. However, there are several rating Schemas (ways that items are rated) in reality, which are overlooked by conventional methods. By transforming rating Schemas into fuzzy user profiles to record users' preferences, our proposed method can deal with different system rating Schemas, and improve the scalability of recommender systems. Additionally, we incorporate user-based method with item-based collaborative methods by clustering users, which can help us to gain insight into the relationship between users. The aim of this research is to provide a new method for personalized recommendation. Our proposed method is the first to normalize the user vectors using fuzzy set theory before the k-medians clustering method is adjusted, and then to apply item-based collaborative algorithm with item vectors. To evaluate the effectiveness of our approach, the proposed algorithm is compared with two conventional collaborative filtering methods, based on MovieLens data set. As expected, our proposed method outperforms the conventional collaborative filtering methods as it can improve system scalability while maintaining accuracy.