个性化信息推荐方法研究

PDF下载HTML

Title:

Research on the recommendation method of personalized information

作者:

姜信景, 齐小刚, 刘立芳

西安电子科技大学 数学与统计学院, 陕西 西安 710071

Author(s):

JIANG Xinjing, QI Xiaogang, LIU Lifang

School of mathematics and statistics, Xi’an University, Xi’an 710071, China

关键词:

网络信息; 截取因子; 信息推送; 基于内容的推荐; 基于内容相似的协同过滤; 基于行为相似的协同过滤; 混合推荐; 组合推荐

Keywords:

network information; interception factor; information push; content-based recommendation; behavior-based similarity collaborative filtering; content-based similarity collaborative filtering; mixed recommendation; combined recommendation

分类号:

TP18;O29

DOI:

10.11992/tis.201701002

摘要:

随着信息技术和互联网的发展，人们进入了信息过量且愈发碎片化的时代。当前，个性化信息推送是用户获取网络信息的有效渠道。由于信息的更新速度快和用户兴趣更新等问题，传统的推荐算法很少关注甚至忽略上述因素，造成最终的推荐结果欠佳。为了给用户更好的个性化推荐服务，论文首次引入截取因子，提出了组合推荐算法（CR算法）。该算法的实质是将截取因子引入到基于内容的推荐算法与基于用户的协同过滤算法中，进而生成混合推荐算法。在推荐列表中，CR算法产生的推荐结果由两部分组成：一部分由混合推荐算法生成，另一部分由基于用户的协同过滤算法生成。根据信息的发布时间，决定该信息由哪类算法产生推荐：当浏览时间与当前时间的间隔不大于某个值时，采用混合推荐算法；否则，直接采用基于用户的协同过滤算法。基于真实数据的实验结果表明，CR算法优于同类算法。

Abstract:

It’s an excessively informational and more fragmented era that is contributed to the development of information technology and the Internet. At present, personalized recommendation is a relatively effective way to help users gain various network information. Recommendations may not be ideal as the traditional algorithms rarely focus on the fast speed of information updating and change of users interests. We propose a combined recommendation algorithm by introducing an interception factor and calls it the CR algorithm. The core idea of it is to introduce the interception factor to the content-based recommendation algorithm and user-based collaborative filtering algorithm. The mixed recommendation consists of the content-based recommendation algorithm and user-based collaborative filtering algorithm. Recommending results of CR algorithm are divided into the outcomes produced by mixed recommendation algorithm and the user-based collaborative filtering algorithm. It is the publishing time of information that decides which algorithm should be chosen to produce recommendations: the mixed recommendation algorithm is selected when the difference between browsing time and message publishing time does not exceed some threshold, or directly chooses the user-based collaborative filtering. Simulation results based on real data show the algorithm we proposed is superior to other existing algorithms.