Cross-industry risk contagion early warning system: a machine learning framework

An effective early warning system at the company level is crucial for preventing financial distress. Traditional warning models mainly rely on internal financial indicators of enterprises, ignoring the transmission of systemic risks from external sources. This limitation frequently leads to poor performance in predicting corporate financial distress, particularly during complex cross-industry risk propagation. This study constructs a novel intelligent risk warning model that integrates a newly developed ASR-FDP (Accounting-Systematic Risk Financial Distress Prediction Model) hybrid model with advanced machine learning techniques to warn of corporate financial crises. Utilizing a dataset spanning from 2011 to 2021, we developed the ASR-FDP predictive model, incorporating both cross-industry systemic risk exposure indicators and company-level financial indicators. The results indicate that the ASR-FDP model, based on random forests, exhibits exceptional performance (accuracy of 93%, F1 score of 0.96), significantly outperforming traditional single-indicator models. Our research highlights two key findings: (1) systemic risk indicators possess substantial predictive power independent of financial information indicators, contributing 19% to the predictive ability in financial distress prediction. This addresses the limitation of traditional models, which focus solely on internal risks while neglecting external risks. The ASR-FDP model more accurately predicts financial distress compared to models relying solely on financial indicators. (2) The intelligent early warning system based on machine learning identifies early warning thresholds capable of achieving a 11-month advance prediction, with prediction lead time and accuracy far exceeding traditional methods. This methodological advancement offers regulators a dynamic monitoring tool for preventive risk mitigation, while also providing theoretical insights into nonlinear risk propagation paths in interconnected markets.