Analysing the Bankers’ Ratings Worldwide Using Machine Learning Techniques

This study applies machine learning techniques to analyse The Banker’s Top 1000 bank rankings, identifying financial metrics associated with ranking positions. The model achieves 94.16% prediction accuracy in out-of-sample tests, with decision tree models performing better than featureless baselines in regression and classification tasks. Through feature selection methods, we identify six key metrics linked to ranking performance: total assets, total liabilities, gross total loans, gross total deposits, total operating income, and loan-to-asset ratio. These metrics provide insights into ranking outcomes. This work is the first to apply neural networks and feature selection to a decade-long dataset of The Banker’s rankings, offering empirical insights for evaluating global financial performance.