Comparative Analysis of Methods for Detecting Social Engineering When Taking Out a Loan

This thesis presents a Comparative Analysis of Methods for Detecting Social Engineering When Taking Out a Loan, focusing on loan applications processed through the mobile app of the Yellow bank. The study addresses the challenge of identifying instances of social engineering within the financial sector, posing risks of financial losses, reputational damage, and regulatory non-compliance. Through extensive literature review and empirical analysis, the research explores statistical methodologies and machine learning models for fraud detection in imbalanced datasets. Key findings indicate the efficacy of Logistic Regression (LogReg) and Random Forest (RF) models, further optimized through ensemble learning and stacking techniques. Practical considerations drive the selection of a simplified model for production deployment, supported by rigorous A/B testing against existing systems. The research underscores the significant business value of advanced fraud detection systems in mitigating financial risks and enhancing profitability.