Predicting positive and negative links in signed social networks by transfer learning
Predicting positive and negative links in signed social networks by transfer learning
复制标题
DOI：
10.1145/2488388.2488517
复制
发表时间：
2013-05
期刊：
影响因子：
--
通讯作者：
Jihang Ye;Hong Cheng;Zhe Zhu;Minghua Chen
中科院分区：
文献类型：
--
作者：
Jihang Ye;Hong Cheng;Zhe Zhu;Minghua Chen
与几乎完全关注积极关系的社交网络的大量研究不同，我们研究了具有积极和负面联系的社交网络。具体而言，我们专注于如何可靠，有效地预测新形成的签名社交网络（称为目标网络）中的链接迹象。由于通常只有很少的边缘符号信息在此类新形成的网络中可用，因此这种少量不足以训练好的分类器。为了应对这一挑战，我们需要现有的成熟签名网络（称为源网络）的帮助，该网络具有丰富的边缘符号信息。我们采用转移学习方法来利用源网络中的边缘符号信息，该信息可能具有边缘实例及其班级标签的不同但相关的联合分布。由于在签名网络中没有针对边缘实例的预定义特征向量，因此我们构建了可以将拓扑知识从源网络转移到目标的可通用特征。借助提取的功能，我们采用了一种类似Adaboost的转移学习算法，并使用实例加权来利用模型学习源网络中更有用的培训实例。三个真正的大型社交网络的实验结果表明，我们的转移学习算法可以比基线方法提高预测准确性40％。
Different from a large body of research on social networks that has focused almost exclusively on positive relationships, we study signed social networks with both positive and negative links. Specifically, we focus on how to reliably and effectively predict the signs of links in a newly formed signed social network (called a target network). Since usually only a very small amount of edge sign information is available in such newly formed networks, this small quantity is not adequate to train a good classifier. To address this challenge, we need assistance from an existing, mature signed network (called a source network) which has abundant edge sign information. We adopt the transfer learning approach to leverage the edge sign information from the source network, which may have a different yet related joint distribution of the edge instances and their class labels. As there is no predefined feature vector for the edge instances in a signed network, we construct generalizable features that can transfer the topological knowledge from the source network to the target. With the extracted features, we adopt an AdaBoost-like transfer learning algorithm with instance weighting to utilize more useful training instances in the source network for model learning. Experimental results on three real large signed social networks demonstrate that our transfer learning algorithm can improve the prediction accuracy by 40% over baseline methods.