| 一种基于多视图学习的群组发现方法 |
| Group discovery method based on multi view learning |
| |
| DOI： |
| 中文关键词: 词群组发现；偏好获取；动态主题模型；多视图学习 |
| 英文关键词:group discovery；preference acquisition；dynamic topic model；multi view learning |
| 基金项目:国家自然科学基金(61772285，61373138)资助项目 |
| 作者 | 单位 | | 王海艳 | 南京邮电大学 计算机学院,江苏南京210023；南京邮电大学 江苏省无线传感网高技术研究重点实验室,江苏南京210023；南京邮电大学 江苏省大数据安全与智能处理重点实验室,江苏南京210023 | | 孙成成 | 南京邮电大学 计算机学院,江苏南京210023 |
|
| 摘要点击次数: 909 |
| 全文下载次数: 277 |
| 中文摘要: |
| 近年来，群组推荐因其良好的实用价值得到了广泛关注。作为群组推荐的一个重要环节，群组发现对推荐结果具有至关重要的影响。对于同一群组中的用户，用户的相似度越大，推荐的精度就越高。为了有效提升群组内用户相似度，改善推荐的精度，降低误差率，文中提出了一种基于多视图学习的群组发现方法。该方法首先提取多维度的显式信息，并用相似度矩阵表示，采用动态主题模型更新用户的偏好；然后，基于多视图学习对用户相似度矩阵分配权重，利用无监督学习训练得出隐式信息；最后，根据用户相似度矩阵和分组方法提出群组发现算法，实现用户群组划分。仿真对比实验表明本文所提的方法分组效果更好。 |
| 英文摘要: |
| In recent years,a group recommendation has attracted more and more attention from researchers due to its practical value.As an important part of group recommendation,the group discovery plays a lead role in the group recommendation.For users in the same group,the more the user similarity,the higher accuracy and,the lower error rate of the recommendation will be obtained.To effectively improve the similarity of users in the same group,reduce the error rate and increase the accuracy of recommendation,this paper proposes a group discovery method based on multi view learning to discover latent user relationships.Firstly,the method extracts multi dimensional explicit information and uses dynamic topic model to update the tendency of users.Then,it uses multi view to fuse heterogeneous data sources.To obtain more user information,the indirect information is obtained by unsupervised learning training.Finally,the groups are divided by the user similarity matrix.Simulation results show that the group discovery method based on multi view learning is better than other methods. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
|
|