An improved collaborative filtering algorithm adapting to user interest changes

In recent years, recommendation systems have been widely applied in different areas. However, typical collaborative filtering algorithms ignored the changes of user interest over time. To solve the problem, we proposed an improved item-based collaborative filtering algorithm to adapt to user interest changes. The proposed algorithm has two contributions: (1) the algorithm designed a time weight function which shows the interest changes to get a more accurate recommendation; (2) the algorithm introduced a new method to calculate the similarity between items. Essentially, this is a linear combination of time similarity and traditional rating similarity. The experiment results show that this method generates a better accuracy than the traditional collaborative filtering algorithm in recommendation predication.