Attribute Weighting and Samples Sampling for Collaborative Filtering
QU Zhao-wei, Jingjing Yao, Xiaoru Wang, Sixing Yin · 2018
Item-Based Collaborative filtering is used in many fields. However, it recommends movies for user only according to the user's behavior records. Therefore, it fails to recommend effectively and solve problems such as cold start. To solve these problems, this paper proposes an attribute weighting collaborative filtering (AW - CF). The algorithm not only takes advantage of the movies' rating information from users but also the movies' attribute information to measure the similarity between movies. We can calculate the similarity between movies more accurate by using the movies' attribute information. Besides, we proposed an method of sampling the samples to represent the train dataset so that we can reduce training time (sampling AW-CF). The experiments based on the dataset of MovieLens show that the proposed AW-CF and sampling AW-CF algorithms achieved good effect in MAE, RMSE and make more efficient recommendations. Their effect are superior to the compared algorithms.