点击次数：

论文类型：期刊论文

发表时间：2015-06-01

发表刊物：ICIC Express Letters, Part B: Applications

收录刊物：EI、Scopus

卷号：6

期号：7

页面范围：1829-1836

ISSN号：21852766

摘要：Nowadays, recommender systems have become a key part of reducing the negative impact of information overload problem in various fields, where users have the possibility of voting for their preferences on items. Collaborative filtering (CF) is one of the most popular and effective recommending techniques to provide personalized recommendations. CF-based methods usually have much better accuracy than other techniques such as content-based filtering, because they are based on the opinions of users with similar tastes or interests. However, CF-based methods suffer from the cold start problem (new user/item) which severely affected the quality of recommendation. To address new user cold start problem, we propose a novel hybrid approach, named US-SlopeOne, to improve the quality of recommendation. In US-SlopeOne, user access sequence is introduced to W-Slope model, and users   similarities are obtained by calculating the similarities of user access sequences instead of user rating similarities. Experiments on three datasets were carried out to evaluate the performance of our method. Our results show that our approach outperforms other methods and improves recommendation quality effectively ? 2015, ICIC Express Letters Office. All rights reserved.