Exploring Profitability in Albanian Banks through Decision Tree Analysis

This study intends to predict Return on Assets (ROA) and assess the significance of several key dependent variables, including Profit per Outlet, Profit per Employee, Natural Logarithm of Assets (ln(assets)), and Loan to Deposits Ratio. In this respect, decision tree regression is employed as the major analytical tool. Within the scope of quarterly data sourced from some leading Albanian banks, such as American Bank of Investments, Banka Kombëtare Tregtare, Credins Bank, Fibank, Intesa Sanpaolo Bank, ProCredit Bank, Raiffeisen Bank, Tirana Bank, and Union Bank, data spanning from January 2016 to December 2022 were taken and used in this analysis. Data collection was done through the website of the Albanian Association of Banks. The following section presents all the relevant information on the selected model, mean squared error (MSE) for training data at 0.5988 and for test data at 0.8912, mean absolute error (MAE) of 0.4764 for the training data and 0.6257 for the test data, and R-squared (R²) at 0.7706 for training data and 0.4244 for test data. All the findings clearly show that decision tree regression is very effective in the prediction of ROA and also helps in elucidating the relative importance of the selected dependent variables within the context of Albanian banking institutions.