IBSS Pioneers ML-Finance Model: Enhancing Bankruptcy Prediction

Recently, a paper titled "Bankruptcy forecasting — Market information with ensemble model" co-authored by Dr Yi Luo, Dr Jia Zhai and Dr Shimeng Shi from the International Business School Suzhou (IBSS) of Xi'an Jiaotong-Liverpool University (XJTLU), Dr Yi Cao from the School of Mathematics and Physics of XJTLU, and Dr Peng Wei from the Business School of University of Edinburgh, has been accepted and published in the The British Accounting Reivew, an ABDC A* journal. This paper introduces an innovative framework for bankruptcy prediction, combining machine learning advances with rigorous financial theory. The research not only sets a new benchmark for predictive accuracy but also extends the theoretical foundation of financial risk modeling, offering invaluable insights for academia and industry.

Researchers propose a two-layer stacking ensemble model, which incorporates four machine learning techniques—Boosted Tree, Random Forest, k-Nearest Neighbor, and Neural Networks—within a "blended meta-learning" framework. This approach ensures algorithmic diversity, robust performance, avoiding overfitting and enhancing predictive accuracy.

Theoretical contributions are equally significant. By grounding the model in Merton’s credit risk framework, the research integrates 326 asset-pricing factors, expanding beyond traditional financial ratios. This innovation demonstrates that market-based variables, including stock returns, volatility, dividends, and liquidity, are stronger predictors of bankruptcy, offering a practical validation of Merton’s framework and its applicability in explaining bankruptcy decision.

The model’s empirical performance is exceptional:

• Superior Predictive Power: It achieves high recall, specificity, F1 scores, and AUC across one-, two-, and three-year horizons, outperforming benchmark models.
• Market Variables’ Dominance: Among all factors tested, market-based indicators consistently outperform traditional financial ratios in distinguishing bankrupt firms.
• Balanced Risk Management: The model minimizes Type 1 (false negatives) and Type 2 (false positives) errors, ensuring accurate predictions of both at-risk and healthy firms.

For academia, this research strengthens the theoretical foundations of financial distress modeling by validating Merton’s framework with empirical evidence. It also broadens the scope of bankruptcy prediction by demonstrating the predictive superiority of market-based variables, paving the way for future theoretical advancements. For industry, the practical implications are transformative:

• Proactive Risk Management: Regulators and financial institutions can leverage the model for early identification of distressed firms, enabling timely interventions.
• Comprehensive Insights: By incorporating stock-market data, the model provides a holistic view of financial risks, overcoming the limitations of traditional financial ratios.
• Scalable Applications: The framework’s adaptability ensures its utility across diverse financial environments and time horizons.

This study exemplifies how cutting-edge research can bridge the gap between theoretical innovation and real-world application. By marrying advanced machine learning with financial theory, the study offers a robust solution to a long-standing challenge in financial risk management. This achievement also underscores IBSS’s commitment to conducting impactful research at the intersection of accounting, finance and advanced technology.

Dr Yi (Sherry) Luo is currently an Assistant Professor of Accounting. Her main research interests lie in financial econometrics, high-frequency volatility modeling, tail risk modeling, machine learning in accounting and finance, and financial accounting (banking specific). She has published in leading international journals, such as the Journal of Financial Econometrics, The British Accounting Review, etc. Her works have been presented at several prestigious conferences, such as the Annual Meeting of SoFiE, IAAE, BAR, FMA Europe, Asia FA, etc. Prior to her work at XJTLU, she was a PhD teaching assistant at Lancaster University. She has rich teaching experience in Econometrics, Financial Accounting, Quantitative Methods in Economics and Statistics. She also contributed to several professional training programmes like Machine Learning in Economics initiated by Timberlake. She holds the certificate of Certified Management Accountant and the license of Certified Public Accountant (Australia).

Dr Jia Zhai is currently a Senior Associate Professor of Finance. Her primary research areas include FinTech and applied financial econometrics, such as machine learning in finance, option pricing, sentiment analysis, and volatility forecasting. Other research interests may include green finance, open banking, and investment decisions.

She has published numerous articles in internationally leading journals, including the Journal of Expert Systems with Applications, Decision Support System, Quantitative Finance, European Journal of Operational Research, and European Journal of Finance. She also serves as an Associate Editor for Finance Research Letters, an ABDC-A-ranked journal. Her research has been presented at world-renowned conferences, such as INFORMS, EURO, Asia FA, and EFMA.

In recent years, Dr Jia Zhai has served as the Principal Investigator of the National Natural Science Foundation of China (NSFC) Young Scientist Fund and has received funding from the National Social Science Foundation of China (NSSFC), the Chinese Ministry of Education, and the UK Economic and Social Research Council (ESRC) for several research projects.

Prior to joining XJTLU, Dr Shimeng Shi worked at Curtin University (Malaysia campus) and Moody’s Analytics (London office). Dr Shimeng Shi's research is mainly in the areas of financial technology, green finance, and credit risk. She has published a number of articles in leading international journals, including Journal of Futures Markets, Quantitative Finance, Review of Quantitative Finance and Accounting, International Review of Financial Analysis, and European Journal of Finance. She has been the reviewer for several peer-review journals.

The British Accounting Review is a highly reputable international journal published by Elsevier. It focuses on high-quality research in accounting and finance. As an ABDC-A* ranked journal, it holds significant academic influence and serves as a key platform for advancing both theoretical and practical knowledge in the field. Its 2023 Cite Score (8.6) and Impact Factor (5.5) respectively rank the journal top 12 in Business, Management and Accounting, and 17 in Business and Finance.