Cyber Risk and Banks Profitability: A Machine Learning Approach

Abstract

Researchers assessed the cyber risk effect on banking profitability through financial metrics obtained from multiple banks enduring several years. This research investigates how the "Cyber Risk Ratio" indicates cyber risk to affect major profitability measurements including Return on Assets (ROA), Return on Equity (ROE), Net Profit Margin (NPM), Operating Profit Margin (OPM), Gross Profit Margin (GPM), Bank Size and Capital Adequacy Ratio (CAR). The researchers used Linear Regression and XGBoost machine learning models to first estimate profitability measures depending on cyber risk levels. With its R-squared value, computed at 0.984, the Linear Regression model generated rather good forecasts since its proposed cyber risk ratio accounted for 98.4% of data changes in the anticipations. The model performed well in terms of accuracy through RMSE measurement of 0.019 and MAE measurement of 0.016. The Linear Regression model shows remarkable success as a technique for this dataset because it establishes a clear linear relationship between bank profitability and cyber risk. Excellent predictions were achieved through XGBoost which identified non-linear patterns while demonstrating R-squared of 0.914 together with RMSE value of 0.283. The specific application demonstrated XGBoost achieved slightly less precision than Linear Regression but added the capability to detect complex correlations in addition to resisting outliers. The XGBoost model proved able to make accurate predictions for non-linear patterns by reaching 85% success during binary classification evaluations. This research reveals that cyber risk introduces substantial effects on bank financial performance and Operating Profit Margin demonstrates the most intense relationship between these variables.