Forecasting recovery rates on non-performing loans with machine learning

Abstract

We compare the performances of a wide set of regression techniques and machine learning algorithms for predicting recovery rates on non-performing loans, using a private database from a European debt collection agency. We find that rule-based algorithms such as Cubist, boosted trees and random forests perform significantly better than other approaches. In addition to loan contract specificities, the predictors referring to the bank recovery process-prior to the portfolio's sale to the debt collector-are also proven to strongly enhance forecasting performances. These variables, derived from the time-series of contacts to defaulted clients and clients' reimbursements to the bank, help all algorithms to better identify debtors with different repayment ability and/or commitment, and in general with different recovery potential.

Look Who's Talking: Interpretable Machine Learning for Assessing Italian SMEs Credit Default

“…An isolate application to the recovery rate forecasting of non-performing loans can be encountered in the credit risk field (Bellotti et al, 2021).…”

A Literature Review of Non-Performing Loan

“…Lending is the primary business of retail banking and non-performing loans (NPLs), have been the focus of attention by European regulators in recent years, as many banks still face difficulties disposing of those that materialized on their balance sheets during the financial crisis (Bellotti et al, 2020). European banks have experienced a particularly challenging period over recent years and the Great Financial Crisis (GFC) has highlighted the weakness of the European banking system and the need to further investigate banks' asset quality and transparency from both a regulatory and an accounting perspective, which pressure by different institutions for a more accurate assessment of loan portfolios led to the general need for higher provisioning in a period characterised by extremely low interest rates and low bank profitability (Bolognesi et al, 2020).…”

GDP Forecasting: Machine Learning, Linear or Autoregression?

“…We also exploit the K-Nearest Neighbour (KNN) machine learning methodology. Our goal is to achieve forecasts with high accuracy and with high degree of explainability that is a best practice for building trust between machine learning and decision-makers, as pointed out in Bellotti et al (2021). The idea is that the decision-maker should adopt the machine learning as a powerful instrument and should employ it with awareness without regarding it as a "black-box.…”