A Collaborative Filtering Recommendation Method Based on Item Category

Abstract

Traditional collaborative filtering algorithms are facing severe challenges of sparse user rating and real-time recommendation. To solve the problems, the category structure of merchandise is analyzed deeply and a collaborative filtering recommendation algorithm based on item category is proposed. A smooth filling technique is used for rating matrix with user preferences and all users rating on the item to solve the sparse problem. A user has different interests on different category. For every item, the nearest neighbors are searched within the category of the item. Not only is the search space of the users’ neighbors reduced greatly, but also search speed and accuracy are promoted. The experimental results show that the method can efficiently improve the recommendation scalability and accuracy of the recommender system.