A Hybrid Collaborative Filtering Algorithm Based on User-Item

Collaborative filtering is one of the most important technologies in e-commerce recommendation system. Traditional similarity measure methods work poorly when the user rating data are extremely sparse. Aiming at this issue a hybrid collaborative filtering is proposed. This method used a novel similarity measure method to predict the target item rating and it fused the advantages of the user-based algorithm and item-based algorithm with the control factor α. The experimental results show that this improved algorithm obviously enhances the recommended accuracy, and provide better recommendation quality.