2026-10-15T04:37:23Z https://www.scidb.cn/oai

10.57760/sciencedb.43704 2026-09-14T09:20:56Z

2026-09-14 A Knowledge Graph Dataset on the Spatiotemporal Distribution of ESG Risks Faced by Chinese Digital Infrastructure Firms Overseas doi:10.57760/sciencedb.43704 en The Digital Silk Road and broader international digital infrastructure cooperation are important paths to narrow the digital divide and promote global digital connectivity. However, when Chinese digital infrastructure companies participate in overseas projects, they face environmental, social, and governance (ESG) risks related to host country governance, technology adaptation, data security, and social disputes, which may lead to project interruptions, cost overruns, trust crises, and reputation damage. The project attributes and ESG risk information are scattered across different sources, lacking a unified structured organization, which limits cross-project comparisons, risk correlation analysis, and data reuse. To this end, this study constructed an ESG risk knowledge graph dataset (ODI-KG) for overseas digital infrastructure projects involving Chinese suppliers, providing data support for related risk analysis and supplier decision-making. This study systematically collected and organized 957 overseas digital infrastructure projects participated in by Chinese suppliers from databases such as the World Bank, Asian Development Bank, AidData, and Reconnecting Asia, as well as government websites, corporate announcements, international organization reports, news media, and academic literature from 2000 to 2024. Among them, 611 projects had recorded ESG risk events, while the remaining 346 projects did not find identifiable ESG risk events in the collected evidence. By using a big language model to assist in case coding, risk factors, ESG risk events, and adverse consequences are extracted to construct a knowledge structure of "project context risk factors, risk events, and adverse consequences," covering 25 risk factors, 8 types of ESG risk events, and 5 types of adverse consequences. Using association rules to mine quantitative risk associations and identifying potential associations through support vector machine link prediction. The final graph contains 13 types of entities and 19 types of relationships, with a total of 2296 entity nodes and 21371 relationships. The released materials include a project-level case database, risk coding instructions, and 32 CSV files of knowledge graphs, including 13 physical files and 19 relational files. The project-level database covers the host country context, project characteristics, supplier background, and corresponding risk codes for the project; the graph file connects entities and relationships through unique identifiers and can be used for Neo4j graph construction. The supporting materials also include MATLAB code for association rule analysis and support vector machine link prediction, as well as source code and dependency explanations for a knowledge service system based on Streamlit.  Overseas digital infrastructure; ESG risk; Knowledge graph; Risk chain; Case data mining Hao Junlin Wang Nannan EMBARGO https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank