Forecasting Bank Profitability Using Deep Learning and Macroeconomic Indicators: A Comparative Model Study

This study proposes a deep learning framework for forecasting bank profitability by integrating macroeconomic indicators with firm-level financial data. Using models such as LSTM, GRU, Transformer, and TCN—compared against traditional approaches like Linear Regression, Random Forest, and XGBoost—we evaluated predictive performance across multiple metrics. Results show the Transformer model achieved the best performance, with an R² score of 0.95 and 91% directional accuracy, outperforming all other models. The inclusion of macroeconomic variables significantly enhanced prediction accuracy. These findings highlight the effectiveness of attention-based deep learning models for financial forecasting in dynamic economic environments.