Return on Equity in Albanian Banks: A Data-Driven Analysis Using XGBoost

This paper estimates the determinants of bank profitability in Albania with respect to return on equity. This study will employ annual reports from eleven banks in Albania covering the period 2014 to 2023, supplemented by extra sources from the Albanian Association of Banks and the Bank of Albania. Data analyzed at a great level of detail using Python. It evaluates an integrated conceptual framework of the influence of financial metrics, institutional infrastructure, and customer service channels on RoE. An XGBoost regression model is applied to predict RoE since this algorithm has advanced ensemble learning capabilities in refining predictions by iterative error corrections. The model is fine-tuned for Mean Squared Error minimization and performs very well on the training data, as evidenced by an R² score of 0.9999. The model slightly overfits, as can be indicated by the lower performance metrics on test data: MSE of 31.8768 and an R² score of 0.6024. Concretely, number of outlets, investments in securities, and debit card issuance are the variables most influencing RoE, as can be inferred from the importance analysis. This paper contributes to the literature with useful insights into financial performance and operational dynamics in Albanian banks, evidencing variables that play a significant role in profitability and, therefore, supply scope for further research and model improvement.