Multimodal Graph Datasets for COVID-19 Forecasting using Geo-Social Media Signals

This dataset contains the preprocessed graph datasets required to reproduce the results of the study: \"Comparative Evaluation of Geo-Social Media Signals in Multimodal Graph-Based Epidemiological Forecasting\" by Dorian Arifi, Devika Jain, Mauricio Santillana, and Bernd Resch. The dataset includes multiple relational graph constructions, including spatial adjacency, mobility networks, semantic similarity, and random baseline structures. In addition, it provides node-level features such as COVID-19 case counts, geo-social media signals, and random time series used for controlled comparisons. These data support the evaluation of multimodal Graph Neural Network (GNN) models for regional COVID-19 forecasting across different relational and feature configurations.本数据集包含复现Dorian Arifi、Devika Jain、Mauricio Santillana与Bernd Resch发表的论文《多模态图基流行病学预测中的地理社交媒体信号对比评估》（"Comparative Evaluation of Geo-Social Media Signals in Multimodal Graph-Based Epidemiological Forecasting"）研究结果所需的预处理图数据集。该数据集涵盖多种关系图构建方式，包括空间邻接、移动网络、语义相似度以及随机基准结构。此外，数据集还提供节点级特征，如新冠确诊病例数、地理社交媒体信号，以及用于受控对照实验的随机时间序列。此类数据可为评估适配不同关系与特征配置的区域新冠疫情预测多模态图神经网络（Graph Neural Network, GNN）模型提供支撑。