Ensemble Machine Learning-based Drivers of Profitability in Vietnamese Listed Banks

The study evaluates the influence of internal factors within banks (bank size, financial leverage, credit quality, cost management efficiency, and digital transformation) and external exogenous factors (GDP growth, inflation, and the development of Fintech companies) on the profitability of commercial banks listed on the Vietnamese stock market. The study uses the Random Forest and XGBoost regression models to estimate the contribution of bank-specific and macroeconomic variables to ROA and ROE. According to the feature importance values, the dominant factor in both models is CR, with values of about 0.55 for ROA and 0.48 for ROE, indicating that nearly half of the predictive ability is attributable to credit risk. CDS and SIZE have an importance of 0.15-0.18, and CIR and DFL have an importance of 0.05-0.08 each. FIN, INF, and GDP all register below 0.05, suggesting that there is little incremental information when the core internal factors are included. These numerical patterns support the econometric results and give a non-parametric ranking of the profitability determinants. Overall, the ML evidence indicates that the intensity of both credit risk control and digital transformation jointly dominates in predicting ROA and ROE, underscoring the need to combine technology investments with conservative risk management and efficiency improvements in Vietnamese commercial banks. The research findings provide empirical evidence to support bank managers and policymakers in improving operational efficiency amid digital transformation.