真实世界超图节点分类公开数据集

本数据集主要面向超图节点分类研究，由公开的高质量数据集Cora、Citeseer、Cora-CA、Zoo、NTU2012组合在一起，这些数据集均经过领域专家严格筛选与整理，确保数据来源的权威性与准确性。其中Cora数据集由2078篇科学出版物构成，具有1433维的词袋属性。该数据集将文章分为7个类，并通过引用关系构建了1579条超边。Citeseer数据集由3312篇论文构成，具有3703维的词袋属性。该数据集将文章分为6个类，并通过引用关系构建了1079条超边。Cora-CA数据集与Cora具有相同数量的文章，但使用了合著关系来定义超边，包含1433维的词袋信息和7个类别。Zoo数据集包含42条超边和101个具有17个属性的动物节点，共有7个类。NTU2012数据集包含来自67个类别的2012个3D形状和2012条超边。该数据集于2023年6月6日，在windows10系统上运行python程序处理得到。在数据加工阶段，所有原始数据通过脚本解析成统一的输入格式，对Cora、Citeseer、Cora-CA数据集分别进行数据拆分，将Zoo数据集以及NTU2012数据集分别分成20%用于训练，80%用于测试，以确保数据符合文献中所述的基本特征。数据量为2MB。This dataset is primarily targeted at hypergraph node classification research, and is assembled from five high-quality public datasets: Cora, Citeseer, Cora-CA, Zoo, and NTU2012. All these datasets have been rigorously screened and curated by domain experts to ensure the authority and accuracy of their data sources. The Cora dataset consists of 2078 scientific publications, with 1433-dimensional bag-of-words features. It categorizes the publications into 7 classes, and constructs 1579 hyperedges through citation relationships. The Citeseer dataset contains 3312 academic papers, with 3703-dimensional bag-of-words features, and divides the papers into 6 classes, building 1079 hyperedges via citation relationships. The Cora-CA dataset has the same number of publications as Cora, but uses co-authorship relationships to define hyperedges, and includes 1433-dimensional bag-of-words information and 7 classes. The Zoo dataset includes 42 hyperedges and 101 animal nodes with 17 attributes each, spanning a total of 7 classes. The NTU2012 dataset contains 2012 3D shapes belonging to 67 categories, along with 2012 hyperedges. This dataset was processed using a Python program on a Windows 10 system on June 6, 2023. During the data processing phase, all raw data was parsed into a unified input format via scripts. For the Cora, Citeseer, and Cora-CA datasets, separate data splitting was performed; for the Zoo and NTU2012 datasets, 20% of each dataset was allocated to the training set and the remaining 80% to the test set, to ensure the data conforms to the basic characteristics described in relevant literature. The total size of this dataset is 2 MB.

搜集汇总

背景与挑战

背景概述

该数据集整合了Cora、Citeseer等多个公开高质量子数据集，专用于超图节点分类研究。这些数据经过统一格式处理，包含节点属性、类别和超边信息，总数据量约为2MB。