| 基于多源用户标签的跨域兴趣融合模型研究 |
| |
| 引用本文： | 张彬,徐建民,吴树芳.基于多源用户标签的跨域兴趣融合模型研究[J].情报科学,2020,38(4):147-152. |
| |
| 作者姓名： | 张彬 徐建民 吴树芳 |
| |
| 作者单位： | 河北大学管理学院,河北保定071002;河北大学管理学院,河北保定071002;河北大学管理学院,河北保定071002 |
| |
| 基金项目： | 保定市社科规划课题“京津冀科技服务创新的融合机制研究”(2018120);教育部产学合作协同育人项目“面向高性能计算的在线实训平台”(201802305014);国家社会科学基金项目“网络信息治理视域下社交网络不可信用户识别研究”(17BTQ068)。 |
| |
| 摘 要： | 【目的/意义】通过对大数据环境下的多源用户兴趣特征有效融合,缓解个性化推荐中用户兴趣偏好数据的稀疏性和准确性问题。【方法/过程】考虑到多域的数据权威度、内容质量及体系结构的差异化较为明显,提出了基于多源用户标签的跨域兴趣融合模型,首先把多个域中的用户兴趣进行标签化处理,然后利用跨域用户识别和标签权重归一方法得到多个域的用户实体-标签矩阵,最后使用域权重影响系数对标签进行融合,构造具有复合权重的用户兴趣标签集。【结果/结论】使用5个来源数据域进行实验与分析,融合模型能够有效提高标签用户覆盖效果,在查全率不断提高的情况,融合域能够保持较高的标签用户查准率,有效提高用户兴趣特征的描绘效果。
|
| 关 键 词： | 兴趣标签 多源用户 融合模型 跨域推荐 |
Cross-domain Interest Fusion Model Based on Multi-source User Tags |
| |
| ZHANG Bin,XU Jian-min,WU Shu-fang.Cross-domain Interest Fusion Model Based on Multi-source User Tags[J].Information Science,2020,38(4):147-152. |
| |
| Authors: | ZHANG Bin XU Jian-min WU Shu-fang |
| |
| Affiliation: | (School of Management.Hebei University,Baoding 071002,China) |
| |
| Abstract: | 【Purpose/significance】By fusing multi-source user interest features in large data environment, to alleviate the sparsity and accuracy of user interest preference data in personalized recommendation.【Method/process】Considering the obvious differences of data authority, content quality and architecture in multi-domain, a cross-domain interest fusion model based on multi-source user tags is proposed. Firstly, user interests in multiple domains are tagged, then user entity-tag matrix in multiple domains is obtained by cross-domain user identification and tag weight normalization method, and finally domain rights are used. Heavy influence coefficient is used to fuse tags and construct user interest tag set with compound weight.【Result/conclusion】Five data domains are used for experiment and analysis. Fusion model can effectively improve the coverage effect of label users. With the improvement of recall rate, fusion domain can maintain a high accuracy rate of label users and effectively improve the depiction effect of user interest features. |
| |
| Keywords: | interest tag multi-source users fusion model cross-domain recommendation |
| 本文献已被 维普 万方数据 等数据库收录！ |
| 点击此处可从《情报科学》浏览原始摘要信息 |
|
点击此处可从《情报科学》下载全文 |
|
