[Paper Review] Analysis of Financial Credit Risk Using Machine Learning

Jacky Chow|arXiv (Cornell University)|Jan 1, 2017

Financial Distress and Bankruptcy Prediction3 references18 citations

TL;DR

This master's thesis evaluates machine learning models for predicting corporate bankruptcy in Korean and Polish manufacturing firms using financial ratios and expert assessments. It finds that incorporating qualitative expert judgments enables >95% accuracy, while purely financial models face limitations due to the curse of dimensionality, with logistic regression achieving 66.4% accuracy on high-dimensional Polish data.

ABSTRACT

Corporate insolvency can have a devastating effect on the economy. With an increasing number of companies making expansion overseas to capitalize on foreign resources, a multinational corporate bankruptcy can disrupt the world's financial ecosystem. Corporations do not fail instantaneously; objective measures and rigorous analysis of qualitative (e.g. brand) and quantitative (e.g. econometric factors) data can help identify a company's financial risk. Gathering and storage of data about a corporation has become less difficult with recent advancements in communication and information technologies. The remaining challenge lies in mining relevant information about a company's health hidden under the vast amounts of data, and using it to forecast insolvency so that managers and stakeholders have time to react. In recent years, machine learning has become a popular field in big data analytics because of its success in learning complicated models. Methods such as support vector machines, adaptive boosting, artificial neural networks, and Gaussian processes can be used for recognizing patterns in the data (with a high degree of accuracy) that may not be apparent to human analysts. This thesis studied corporate bankruptcy of manufacturing companies in Korea and Poland using experts' opinions and financial measures, respectively. Using publicly available datasets, several machine learning methods were applied to learn the relationship between the company's current state and its fate in the near future. Results showed that predictions with accuracy greater than 95% were achievable using any machine learning technique when informative features like experts' assessment were used. However, when using purely financial factors to predict whether or not a company will go bankrupt, the correlation is not as strong.

Motivation & Objective

• To evaluate machine learning models for predicting corporate bankruptcy in manufacturing firms.
• To compare the predictive performance of various ML techniques using financial and qualitative data.
• To investigate the impact of data dimensionality on model generalization and accuracy.
• To assess the effectiveness of dimensionality reduction and feature selection in improving model robustness.
• To provide practical insights for financial risk management using data-driven approaches.

Proposed method

• Applied multiple machine learning models: logistic regression, SVM, decision trees, AdaBoost, ANN, and Gaussian processes.
• Used publicly available datasets from the UCI Machine Learning Repository for Korean and Polish corporate bankruptcy cases.
• Performed data pre-processing, including handling missing values and feature scaling.
• Employed principal component analysis (PCA) and other dimensionality reduction techniques to address high-dimensional data.
• Selected models using cross-validation and evaluated performance via accuracy, precision, recall, and F1-score.
• Integrated expert assessments as qualitative features to enhance predictive power.

Experimental results

Research questions

• RQ1 How accurately can machine learning models predict corporate bankruptcy using financial ratios alone?
• RQ2 What is the impact of incorporating expert assessments on bankruptcy prediction accuracy?
• RQ3 How does the curse of dimensionality affect model performance in high-dimensional financial datasets?
• RQ4 Which machine learning model performs best in terms of generalization and robustness on bankruptcy prediction tasks?
• RQ5 To what extent do dimensionality reduction techniques improve model performance on sparse, high-dimensional data?

Key findings

• Incorporating expert assessments enabled machine learning models to achieve over 95% accuracy in predicting bankruptcy for Korean firms.
• Purely financial models on the Polish dataset achieved a maximum accuracy of 66.4% using logistic regression, indicating limited predictive power.
• High-dimensional financial data led to overfitting in complex models due to the curse of dimensionality, reducing generalization.
• Dimensionality reduction techniques like PCA improved model stability and performance, especially for nonlinear models.
• Simpler models such as logistic regression outperformed complex models on high-dimensional, sparse datasets.
• The study confirms that qualitative factors significantly enhance predictive accuracy, even when financial data alone show weak correlations.