| 基于骨干度与网络编码的链路预测模型研究 |
| |
| 引用本文： | 胡旭飞,许云峰.基于骨干度与网络编码的链路预测模型研究[J].河北工业科技,2019,36(5):310-313. |
| |
| 作者姓名： | 胡旭飞 许云峰 |
| |
| 作者单位： | 河北科技大学信息科学与工程学院,河北石家庄,050018;河北科技大学信息科学与工程学院,河北石家庄,050018 |
| |
| 基金项目： | 国家自然科学基金（61673235） |
| |
| 摘 要： | 为了研究网络表示学习在社交网络中链路预测方面的应用,提出了一种基于骨干度与网络编码的链路预测模型(BDLINE)。在网络表示学习算法LINE的基础上融入骨干度算法,通过给一阶相似度和二阶相似度中增添骨干权重,将网络编码到多维向量空间中,调试到最优参数。实验采用2个真实数据的数据集,分别在不同的算法模型上进行多次实验。实验结果表明:在链路预测方面,BDLINE均比其他网络表示学习算法的性能有所提升,AUC评测值更高,预测效果表现得更好。因此,所提出的方法可以方便地提取网络特征信息,更好地处理社交网络在链路预测中的随机性,对社交网络中预测网络节点的关联性和有效性具有一定的参考。
|
| 关 键 词： | 计算机网络 网络表示学习 链路预测 社交网络 相似性 |
| 收稿时间： | 2019/5/13 0:00:00 |
| 修稿时间： | 2019/6/30 0:00:00 |
Research on link prediction model based on backbone degree and network coding |
| |
| HU Xufei and XU Yunfeng.Research on link prediction model based on backbone degree and network coding[J].Hebei Journal of Industrial Science & Technology,2019,36(5):310-313. |
| |
| Authors: | HU Xufei and XU Yunfeng |
| |
| Abstract: | In order to study the application of network representation learning in link prediction in social networks, a link prediction model based on backbone and network coding (BDLINE) is proposed. The model integrates the backbone algorithm based on the network representation learning algorithm LINE. By adding the backbone weight to the first-order similarity and the second-order similarity, the network is encoded into the multi-dimensional vector space and debugged to the optimal parameters. The data set used in the experiment is two different real data, and multiple experiments are performed on different algorithm models. The experimental results show that in terms of link prediction, the proposed algorithm model has improved performance compared to other network representation learning algorithms; the AUC evaluation value is higher, and the prediction effect is better. It is more convenient for extracting network feature information by using the method, and can better deal with the randomness problem of social network in link prediction. The method has certain guiding significance for predicting the relevance and effectiveness of network nodes in social networks. |
| |
| Keywords: | computer network network representation learning link prediction social network similarity |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《河北工业科技》浏览原始摘要信息 |
|
点击此处可从《河北工业科技》下载全文 |
|
