An imporved personalized recommendation algorithm based on fuzzy clustering
Ling Zheng, Xinyu Zhao, Shuo Cui, Dong Yue · 2010
For the collaborative filtering algorithm at present, since the uses' ratings of the item are sparse and the users' interests change over time, the similarity calculation of the items or the users is not accurate, an improved collaborative filtering algorithm is suggested. It uses the fuzzy clustering algorithm to cluster users, and transforms individual user's ratings on items into a group of similar users' ratings and thereby constructs the user fuzzy cluster - item rating matrix. In addition, when calculating the users' similarity, a weight gradually increasing as the time is given to each rating. And the weighted ratings are used to find the target user's nearest neighbors. The experiments show that the method can improve the recommendation quality of the collaborative filtering recommendation systems.