Modelling and Predicting Online Vaccination Views using Bow-tie Decomposition: Dataset & Code

Here we document the dataset and the programming code used in our paper "Modelling and Predicting Online Vaccination Views using Bow-tie Decomposition", Royal Society Open Science (2024). In this paper, we investigate a temporal dataset that describes the Facebook vaccination views campaign in network representations, involving nearly 100 million users on Facebook from across countries, continents and languages. It was provided by Johnson et al. (2020) and Illari et al. (2022) spanning different time periods. The former contains two snapshots in February 2019 and October 2019 (before the COVID-19), which we study in our main paper. The latter contains another two snapshots in November 2019 and December 2020 (at the initial stage of the COVID-19), which we study in Supplementary Material. Both of them are openly available and documented in two papers separately of different formats (PDF & Excel), thus requiring intensive preprocessing. To make it easier for other researchers to use this dataset, here we reorganise both versions of this dataset in gpickle format (can be directly imported as attributed networks using Python) and in CSV format (for general use of the dataset).

本文档记录了我们发表于《皇家学会开放科学（Royal Society Open Science）》2024年的论文《基于蝴蝶结分解（Bow-tie Decomposition）建模与预测在线疫苗接种观点》中所使用的数据集与程序代码。本研究聚焦于一份时序数据集，该数据集以网络表征形式描述了全球范围内的Facebook疫苗接种观点宣传活动，涵盖来自不同国家、大洲与语言群体的近1亿Facebook用户。该数据集由Johnson等人（2020）与Illari等人（2022）提供，涵盖不同时间跨度：前者包含2019年2月与2019年10月（新冠疫情暴发前）的两份网络快照，为本论文核心研究对象；后者包含2019年11月与2020年12月（新冠疫情初始阶段）的另外两份网络快照，用于补充材料中的研究。两份数据集均已公开，且分别以不同格式（PDF与Excel）在两篇独立论文中进行了说明，因此需要进行大量预处理工作。为便于其他研究人员使用该数据集，本文将上述两个版本的数据集分别整理为gpickle格式（可通过Python直接导入为属性网络）与CSV格式（适配数据集的通用使用场景）。