Machine Learning Algorithm for Forecasting Banking Profitability in Southeast Asia

Abstract

This study examines the use of machine learning algorithms to forecast banking profitability in Southeast Asia using Return on Assets (ROA) as the main performance indicator. The study is motivated by the central role of banks as financial intermediaries, liquidity providers, and key institutions supporting financial stability across the ASEAN region. A quantitative predictive approach is employed using secondary data from the annual financial statements of banks in Indonesia, Malaysia, Singapore, Thailand, the Philippines, and Vietnam for the 2017–2024 period. The analysis is carried out in Google Colab using the Support Vector Regression (SVR) algorithm. Historical ROA values are used as predictors, while ROA in 2024 serves as the prediction target. Model performance is evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). The results show an MAE of 1.6299, an RMSE of 3.6195, and a MAPE of 209.33%. These results suggest that SVR can capture part of the historical pattern in bank ROA, particularly for banks with values within a more typical range. Its performance, however, becomes less stable when actual ROA is negative, close to zero, or unusually high or low. The findings therefore indicate that SVR can serve as an initial forecasting model for banking profitability, although its predictive performance should be strengthened by incorporating additional financial variables.