Application of machine learning algorithms for NPL prediction based on financial data. The case of Albania

Machine learning has become increasingly important in predicting nonperforming loans (NPLs), offering greater accuracy than traditional risk assessment methods. As lending activity continues to expand in Albania, accurately forecasting nonperforming loans is therefore essential to minimize potential risks to financial stability. Using data obtained from the Bank of Albania, this study aims to forecast the NPL ratio for the period 2017–2023 using seven bank-specific financial indicators. Two supervised learning algorithms: Random Forest (RF) and Support Vector Machine (SVM) are employed for this purpose and model performance was evaluated using root mean squared error (RMSE) and mean absolute error (MAE). Furthermore, this paper identifies key factors via feature importance analysis and evaluates how excluding less relevant variables impacts predictive accuracy. The results show that before feature selection, RF model outperformed the SVM, achieving the lowest RMSE and MAE values of 0.015 and 0.012, respectively. The most influential variables were identified as Loan Loss Reserves/Total Loans and Bank Loans/Total Assets, whereas several of the indicators exhibited negligible effects. Model performance for both RF and SVM improved slightly after removing less relevant variables, as indicated by lower RMSE and MAE values. While the SVM model’s accuracy increased after dimensionality reduction, RF model still outperformed it overall based on the performance metrics. Specifically, the simplified RF achieved superior results, recording an RMSE of 0.011 and an MAE of 0.009, while also attaining the highest adjusted R2 among all models. The findings confirm the importance of variable selection in enhancing model performance and demonstrate RF as the most effective tool for NPL prediction in Albania. These results provide practical implications for banks and regulators, supporting the development of early warning systems and contributing to financial stability.