Research And Application Of The Collaborative Filtering Algorithm Based On Interest Distribution

With the continuous development of information technology,information amount has greatly increased due to the information explosion,which leads to a lower information utilization rate.In this paper we first carried out a full investigation on the background and the current research situation of recommendation system.Even though collaborative filtering algorithm is the most widely used algorithm in recommendation system,in practical applications there are problems such as users are not likely to score every aspect of the target,which leads to an extremely sparse user-items rating matrix.Under such circumstances,collaborative filtering algorithm fails to distinguish some similar users,and cannot recommend items newly added to the system.In order to solve the problem of sparse scoring,this paper considered the item features and users' rating on items,and promoted an improved algorithm "User-CF",fusion of interest distribution based collaborative filtering algorithm,known as FIDCF.This algorithm studied from the aspect of item classification that users have scored,which gives a new reference standard to seek similar users and effectively calculate the neighbor user group similar to the given user.Meanwhile,this paper implements this algorithm and analyzes it from MAE,recall and precision.The result turns out that the improved algorithm has greater accuracy than the original collaborative filtering algorithm based on users and other algorithms based on user prefer models.This paper also implemented a movie recommendation system,applying this improved algorithm to real movie recommender system.