FairGAD
收藏资源简介:
FairGAD数据集由乔治亚理工学院的研究团队创建,包含两个子数据集,分别基于全球知名的社交媒体平台Reddit和Twitter。这些数据集包含120万和40万条边,分别关联9000和47000个节点,利用政治倾向作为敏感属性,错误信息传播者作为异常标签。数据集的创建旨在解决现有研究中缺乏包含实际图结构、异常标签和敏感属性的数据集问题,以推动公平图异常检测的研究。通过这些数据集,研究团队评估了现有图异常检测方法的性能和公平性,揭示了它们在处理FairGAD问题时的有效性和局限性。
The FairGAD dataset was developed by a research team at the Georgia Institute of Technology. It comprises two sub-datasets based on the globally renowned social media platforms Reddit and Twitter, respectively. The two sub-datasets contain 1.2 million and 0.4 million edges respectively, which are associated with 9,000 and 47,000 nodes, taking political orientation as the sensitive attribute and misinformation spreaders as the anomaly labels. The dataset was created to address the gap in existing research where datasets with real graph structures, anomaly labels and sensitive attributes are lacking, so as to advance research on fair graph anomaly detection. Using these datasets, the research team evaluated the performance and fairness of existing graph anomaly detection methods, and revealed their effectiveness and limitations when addressing FairGAD-related problems.
数据集概述
数据集信息
- 名称: Reddit, Twitter
- 格式: PyTorch Geometric datasets
- 链接: 数据集下载链接
环境要求
- python=3.8
- pytorch
- pyg
- networkx
- scipy
实现细节
- 公平性正则化器: 在
utils.py文件中实现,包括 FairOD, HIN, correlation 和 ADCG 正则化器。 - 模型修改: 对 DOMINANT, CONAD, CoLA, DONE, AdONE 进行了修改,并新增了 VGOD 文件,包含公平性正则化器。
- 主要驱动文件:
fairGAD/test_fair_fitting.py
用户账户查询与移除
- 联系方式: 发送邮件至
nnnk [at] gatech [dot] edu,标题为 "FairGAD - Account Inquiry and Removal",包含用户名和平台(Reddit/Twitter)。
数据使用协议
- 联系方式: 发送邮件至
nnnk [at] gatech [dot] edu,标题为 "FairGAD - Data Use Agreement"。
潜在的Twitter机器人
- 节点索引: 包含可能的Twitter机器人账户的节点索引,Botometer "raw_overall" 分数大于0.9。
引用
- 论文: 《Towards Fair Graph Anomaly Detection: Problem, New Datasets, and Evaluation》
- 作者: Neo, Neng Kai Nigel 等
- 日期: 2024-02-25
- arXiv链接: arXiv链接




