Assembling real networks from synthetic and unstructured subsets: the corporate reporting case

Area of Science:

• Economics
• Network Science
• Data Science

Background:

• Interfirm business transaction networks offer insights into economic structures.
• Limited availability of comprehensive network data necessitates reliance on incomplete subsets, leading to uncertain conclusions.
• Existing network research often uses partial data, similar to protein interaction studies.

Purpose of the Study:

• To construct a financial reporting network from unstructured Japanese corporate data.
• To compare this proxy network with the complete real business transaction network.
• To enhance understanding of the relationship between partial network subsets and complete networks.

Main Methods:

• Utilized unstructured financial and corporate reporting data from Japan.
• Constructed a financial reporting network as a proxy for real business transactions.
• Performed comparative analysis of the proxy and real networks at macro and local structural levels.
• Developed an agent-based pruning algorithm to reconcile network differences.

Main Results:

• Established a financial reporting network derived from Japanese corporate data.
• Demonstrated comparative analysis between the proxy and real business transaction networks.
• Identified non-trivial relationships between partial network subsets and complete networks.
• Presented a novel agent-based pruning algorithm for network reconciliation.

Conclusions:

• The comparative analysis enhances understanding of network structures derived from partial data.
• The developed algorithm offers a framework for improving network research and data extrapolation.
• Findings contribute to more robust conclusions from incomplete network datasets in economics and beyond.