Evaluation of Machine Learning Methods for Loan Default Prediction: A Case Study Using Peer-to-Peer Lending Data

Abstract

This diploma thesis investigates the transformative impact of digital technologies on the financial industry, particularly focusing on the role of data science and machine learning in banking. The study aims to determine how emerging technologies such as blockchain, big data analytics, artificial intelligence (AI), and machine learning (ML) are reshaping financial services. It highlights the fintech revolution and its disruptive influence on traditional banking models, with a particular focus on digital banks, neobanks, and the integration of technology in lending, payments, and investment management.

A comprehensive analysis is provided on advanced methods for credit risk assessment and fraud detection, leveraging machine learning algorithms to enhance predictive accuracy and operational efficiency. A significant portion of the study is dedicated to developing predictive modeling techniques for loan default prediction. Various machine learning algorithms, including logistic regression, decision tree, gradient boosting, random forest, and neural networks, are employed to evaluate and predict loan defaults. The research underscores the importance of model evaluation metrics such as accuracy, precision, recall, and especially F1 score in optimizing model performance.

The findings demonstrate that machine learning, especially models using ensemble learning, can effectively predict loan defaults, thereby aiding financial institutions in mitigating risk and improving decision-making processes. The study concludes with recommendations for future research, including exploring advanced neural network architectures and integrating alternative data sources like social media activity for enhanced predictive power.