Detecting High-Potential SMEs with Heterogeneous Graph Neural Networks

Researchers developed SME-HGT, a Heterogeneous Graph Transformer that predicts high-potential small and medium enterprises using public data from SBIR funding programs. The AI model achieved 89.6% precision in identifying promising SMEs, outperforming traditional methods by analyzing relationships between companies, research topics, and government agencies.

• →SMEs represent 99.9% of U.S. businesses and generate 44% of economic activity but identifying high-potential ones remains challenging.
• →SME-HGT uses heterogeneous graph neural networks to analyze 32,268 companies and their relationships with research topics and agencies.
• →The model achieved 0.621 AUPRC and 89.6% precision at screening depth of 100 companies, with 2.14x lift over random selection.
• →The framework relies exclusively on public data, ensuring reproducibility and practical applicability for policymakers and investors.
• →Results demonstrate that relational structure provides meaningful signals for assessing SME potential beyond traditional metrics.