Small businesses often struggle to get loans because banks find it hard to assess their creditworthiness. But with the rise of FinTech, new tools like machine learning (ML) are changing the game!
What’s the Study “Can we trust machine learning to predict the credit risk of small businesses?” (by Alessandro Bitetto – LIUC – with Paola Cerchiello, Stefano Filomeni, Alessandra Tanda, Barbara Tarantino) about?
This research tested whether ML can predict credit risk for small businesses better than traditional methods. Using real data from Italian SMEs (2015–2017), the study compared ML with older statistical models.
Key Findings:
- ML outperforms traditional methods when lenders have limited data (e.g., only recent financials or invoice records).
- With historical data, ML and traditional models perform similarly.
- What are the most important factors for predicting risk? Revenue (Turnover), debt levels, and payment behaviour (like late invoices).
- ML’s “black box” problem? Solved! The study used Shapley values to explain how ML makes decisions, making it transparent for lenders and borrowers.
Why It Matters?
FinTech lenders use ML to speed up loan approvals, helping small businesses access funds faster. This research shows ML is not just accurate but also explainable, building trust in digital lending.
