Time-Based K-nearest Neighbor Collaborative Filtering

Abstract

Personalized recommendation technology has been developed rapidly and used more widely. LehuBT is one of the most popular Ipv6 website developed by Shanghai University. With the rapid development of users and torrents, an order to help the users to select their favors in numerous torrents, personalized recommendation technology should be added into the current LehuBT. The influence of time factor is more and more important as times go on in LehuBT, a torrent was downloaded by a user nearer to now, it's more important to indicate interesting of the user. However, time factor was not taken into account in traditional Collaborative Filtering. In this paper, a novel algorithm named TimeKNNCF (Time-based K-Nearest Neighbor Collaborative Filtering) is proposed, in which a new similarity function is proposed based on the concept of Importance Degree of Item named IDOI. For each torrents in the download list, the value of IDOI would be calculated based on the download time, the nearer the download time, the higher value of IDOI. The download list would be transformed to IDOI vector, and the similarity would be calculated based on IDOI vector. The experiments based on the real data of LehuBT show that TimeKNNCF provides better recommendations than the traditional collaborative filtering methods.