A Trust-Based Prediction Approach for Recommendation System

摘要

The recommendation system has been widely used in e-commerce, but still suffers from data sparsity and cold-start problems. This paper combines the user trust relationship with the collaborative filtering recommendation system and puts forward the recommendation approach based on trust delivery (TDR), in order to solve the above two problems. Through calculating the quantifying trust values between users, the prediction score of an unrated item can be figured out to achieve effective recommendation. Compared with other recommendation algorithms, TDR achieves better performance on standard Mean Absolute Error (MAE) and Coverage.