期刊论文

RESEARCH ON PERSONALIZED RECOMMENDATION MODEL ORIENTED ZHIHU

中文

计算机应用与软件

1000-386X

2018

35

5

37-41

10.3969/j.issn.1000-386x.2018.05.007

国家自然科学基金项目（61402220）；

本校为第一机构

个性化推荐能向用户提供个性化的信息服务和决策支持。针对知乎提出一种基于混合算法的个性化推荐模型。首先根据用户信息提出Person Rank算法,结合用户兴趣标签构建用户模型。其次根据问答信息提出Problem Rank算法,将问答信息标签化处理后构建问答模型。然后采用基于标签推荐、相似用户推荐、多维滑动窗口分阶段结合的方法来构建推荐算法模型。利用梯度下降算法训练模型后,实验结果表明推荐正确率稳定在0.78左右,模型所推荐的信息能够满足知乎用户需求,个性化推荐效果得以提高。

Personalized recommendation can provide users with personalized information service and decision support.A personalized recommendation model based on hybrid algorithm is proposed for ZHIHU. Person Rank algorithm is put forward based on user information,and user model is constructed according to its interest tag; Problem Rank algorithm is brought forward according to the question and answer（ Q A） information,and Q A model is built after tagging the question and answer information. The recommended algorithm model is constructed based on the com...