Collaborative Filtering Algorithm based on Forgetting Curve and Long Tail Theory

The traditional collaborative filtering algorithm only pays attention to the rating by users. In reality, however, user and item information is always changing with time flying. Therefore, recommendation systems need to take time-varying changes into consideration. The collaborative filtering algorithm which is based on Forgetting Curve and Long Tail theory (FCLT) is introduced for the above problems. The following two points are discussed depending on the problem: First, the user-item rating matrix can update in real time by forgetting curve; secondly, according to the Long Tail theory and item popularity, a further similarity calculation method is obtained. The experimental results demonstrated that the proposed algorithm can effectively improve the recommendation accuracy and alleviate the Long Tail effect.