A hybrid genre-based personalized recommendation algorithm
Yajie Hu, Yi Hsuan Yang, Caihong Li, Yachen Wang, Lian Li · 2016
Because of the serious information overload problem on the internet, the recommender system as one of the most important solutions has been widely used to help users find more valuable information. However, the traditional collaborative filtering method is seriously affected by the rating sparseness and cold start to obtain the precise recommendation. In this paper, a hybrid collaborative filtering method (CGCF) based on the traditional user-based collaborative filtering, as well as, a new approach named genre-based collaborative filtering (GCF) is proposed. GCF uses term frequency-inverse document frequency(TF-IDF) to combine users' former ratings with item genres to quantize individual genre preference. Combining GCF and User-bsed collaborative filtering with dynamic weight, we proposed the CGCF. According to the experiment on Movielens dataset, when comparing with Item-based collaborative filtering, CGCF has reduced MAE by 2.2% and improved Coverage by 16.9%. When comparing with User-based collaborative filtering, CGCF has reduced MAE by 2.5% and improve Coverage by 6.2%. The results demonstrate that the proposed method improves the precision and coverage of recommendation obviously comparing with the traditional ones.