Machine learning techniques for default prediction: an application to small Italian companies

Problem

Predicting loan default of small Italian companies, traditionally handled with statistical classifiers such as logistic regression.

Why it matters

Defaults by small firms create lender losses and affect credit access in the economy.

What the AI did

Various machine learning classifiers were trained on a large sample of small Italian companies, with feature selection and threshold tuning, and compared against logistic regression for default prediction.

Results

Machine learning was found slightly better than logistic regression, but the small performance gain was deemed insufficient to abandon classical classifiers.

Impact

The evidence supports only a marginal ML advantage over logistic regression in this dataset.

Record

Importance rating

More in Business & Industry