Collaborative Filtering Algorithm Based on Mutual Information

Recommender systems are used by E-commerce sites to suggest products to their customers and to provide consumers with information to help them determine which products to purchase. Collaborative filtering algorithm is the most extensive personalized recommendation used in recommender systems. Since not being considering the dependence between predicted item and historical item, typical collaborative filtering algorithm is not fit for multiple interests recommendation. The authors analyzed the reason and presented a new algorithm, collaborative filtering based on mutual information. By removing the historical items on which predicted item has not high dependence, the algorithms can deal with personalized recommendation for user’s multiple interests. The experiment shows new algorithms is more accurate than other algorithms, especially fitting for the environment where users have many completely different interests.