A new academic study presents an AI-powered framework for early warning in credit risk, designed for China’s internet finance industry. The research introduces a deep learning system that combines convolutional neural networks (CNNs) with knowledge graph theory to improve risk identification across structured and unstructured data sources. By modeling both financial ratios and qualitative attributes, the framework enables faster and more reliable detection of enterprise-level credit default risks.
At the center of the system is a dual-branch CNN architecture that processes two distinct data streams. One stream analyzes quantitative indicators such as asset-liability ratio and return on assets, while the other evaluates qualitative features including audit opinions, market sentiment, and public disclosures. Feature extraction is enhanced through principal component analysis and grey correlation analysis. A knowledge graph further integrates and links entities across dimensions, improving semantic representation and allowing more granular pattern discovery in data-rich financial environments.
The model was trained and tested on data from 81 Chinese internet financial companies, 12 of which had documented credit defaults. After adaptive sampling and parameter tuning, the final model achieved an accuracy of 97.64 percent, with a recall rate of 98.76 percent and an F1 score of 98.55 percent. Benchmarking showed that the proposed approach outperformed support vector machines, random forest algorithms, and BP neural networks in accuracy and interpretability. The system also incorporates a credit rating mechanism and an early warning threshold of 2.8 percent, supporting practical deployment in financial risk supervision.
“This model was developed to enhance credit risk monitoring in data-intensive sectors where traditional methods often fall short,” said Yicheng Peng, lead author of the study. “By integrating CNNs with knowledge graph theory, we built a system that strengthens predictive accuracy while supporting decision transparency.”
The project reflects Yicheng Peng’s combined academic and professional expertise in AI, financial risk modeling, and data systems. Currently a Senior Consultant at West Monroe’s M&A practice in New York, Peng supports complex valuation and credit assessment projects across industries. Peng holds a Master of Science in Financial Risk Management from Pace University’s Lubin School of Business and dual bachelor’s degrees in Finance and Law from East China University of Political Science and Law. These academic foundations, combined with practical experience in enterprise analytics and regulatory compliance, continue to shape ongoing research in scalable and interpretable financial intelligence tools.
As financial ecosystems continue to digitize, this framework offers a timely approach to integrating AI into institutional credit risk evaluation, supporting early intervention and stronger safeguards against systemic disruption.