[Paper Review] Deep Learning vs. Gradient Boosting: Benchmarking state-of-the-art machine learning algorithms for credit scoring

Marc Schmitt|arXiv (Cornell University)|May 21, 2022

Financial Distress and Bankruptcy PredictionBusiness, Management and Accounting21 citations

TL;DR

Gradient boosting machines (GBM) generally outperform deep learning (DL) for credit scoring across multiple datasets, with faster training times; DL is not universally inferior, but GBM is typically the preferred choice for structured credit-scoring tasks.

ABSTRACT

Artificial intelligence (AI) and machine learning (ML) have become vital to remain competitive for financial services companies around the globe. The two models currently competing for the pole position in credit risk management are deep learning (DL) and gradient boosting machines (GBM). This paper benchmarked those two algorithms in the context of credit scoring using three distinct datasets with different features to account for the reality that model choice/power is often dependent on the underlying characteristics of the dataset. The experiment has shown that GBM tends to be more powerful than DL and has also the advantage of speed due to lower computational requirements. This makes GBM the winner and choice for credit scoring. However, it was also shown that the outperformance of GBM is not always guaranteed and ultimately the concrete problem scenario or dataset will determine the final model choice. Overall, based on this study both algorithms can be considered state-of-the-art for binary classification tasks on structured datasets, while GBM should be the go-to solution for most problem scenarios due to easier use, significantly faster training time, and superior accuracy.

Motivation & Objective

• Motivate the comparison between DL and GBM for credit risk scoring in industry settings.
• Assess how model choice interacts with dataset characteristics in credit scoring.
• Provide evidence to guide practitioners on state-of-the-art models for binary classification on structured data.

Proposed method

• Benchmark two state-of-the-art algorithms (DL and GBM) on credit scoring tasks.
• Use three distinct datasets with varying feature characteristics to test model robustness.
• Evaluate performance in terms of predictive accuracy and computational efficiency.

Experimental results

Research questions

• RQ1 Does GBM outperform DL for credit scoring across different structured datasets?
• RQ2 Under what dataset conditions does DL offer advantages over GBM in credit risk modeling?
• RQ3 What are the trade-offs between accuracy and training speed for DL vs GBM in this domain?

Key findings

• GBM tends to be more powerful than DL for credit scoring.
• GBM has the advantage of faster computational speed due to lower resource requirements.
• The outperformance of GBM is not guaranteed in every scenario; dataset characteristics determine the best model.
• Both algorithms can be considered state-of-the-art for binary classification on structured data.
• GBM should be the go-to solution for most problem scenarios due to easier use, significantly faster training time, and superior accuracy.