基于时间感知和用户兴趣重要度融合的文档推荐模型

Document Recommendation Based on Fusion of Time-awareness and User Interests Importance Degree

基于内容的文档推荐系统中,传统的向量空间模型VSM直接使用TF-IDF方法确定权值,不能真正体现用户兴趣重要度;同时,由于未考虑用户兴趣随时间变化而发生改变,导致推荐精度较低.因此,提出了基于时间感知和用户兴趣重要度融合的文档推荐模型.首先根据用户浏览行为和相关信息,隐式提取用户兴趣,构建新的用户兴趣向量模型.针对用户兴趣受时间因素的影响,提出使用时间延迟函数对近期浏览的文档赋予更高的时间权值;然后应用灰色关联理论,建立用户兴趣因子序列与文档兴趣度参考序列间关联模型,以确定用户兴趣重要度;进而融合时间因素和用户兴趣重要度建立文档推荐模型.最后设计了一个实验系统,对比所提出的模型与其他两个模型的推荐效果,验证了基于时间感知和用户兴趣重要度融合的模型优于其他模型,能够为用户提供更准确的文档.

In content-based document recommender system,the conventional VSM determines the w eights value according to TF-IDF,w hich cannot really embody the importance degree of user interests. And due to not considering the impact of time factor to user interests,the results are relatively low precision of document recommender system. Therefore,w e proposed a document recommendation model based on fusion of time-aw are and user interest importance degree. According to the user＇s brow sing behaviors and related information,user＇s interests are elicited. And w e put forw ard a time decay function to assign the higher time w eights to recent brow sing documents. Then the associated model betw een factor sequences of user interests and reference sequence is built to determine the user interest importance degree in line w ith Grey Incidence Theory. Further,the document recommendation model is created based on the fusion of time factor and user interest importance degree. Lastly,w e designed a trial system to compare the proposed model to other tw o recommendation models. The results show ed that our proposed model outperformed the other tw o models and can recommend more accurate documents for users.