The scope of this study is to investigate the capability of AI methods to accurately detect and predict credit risks based on retail borrowers' features. The comparison of logistic regression, decision tree, and random forest showed that machine learning methods are able to predict credit defaults of individuals more accurately than the logit model. Furthermore, it was demonstrated how random forest and decision tree models were more sensitive in detecting default borrowers.
Author(s): Bohdan Popovych
Series: BestMasters
Publisher: Springer Gabler
Year: 2022
Language: English
Pages: 92
Foreword
Abstract
Contents
Abbreviations
List of Figures
List of Tables
1 Introduction
2 Theoretical Concepts of Credit Scoring
2.1 Literature Background
2.2 Theoretical Overview
2.2.1 Definition and Application
2.2.2 Advantages and Limitations
2.2.3 Comparison of Credit Scoring and Credit Rating
2.3 Regulatory Requirements
3 Credit Scoring Methodologies
3.1 Traditional Credit Scoring Techniques
3.1.1 Discriminant Analysis
3.1.2 Logistic Regression
3.1.3 Expert Systems
3.2 Artificial Intelligence Classification Methods
3.2.1 Overview of AI Classification Techniques
3.2.2 Decision Trees
3.2.3 Random Forest
3.3 Methods of Model Validation
4 Empirical Analysis
4.1 Data Analysis
4.2 Logistic Regression
4.2.1 Model Description
4.2.2 Model Interpretation
4.2.3 Model Performance
4.3 Decision Tree
4.3.1 Model Description and Interpretation
4.3.2 Model Performance
4.4 Random Forest
4.4.1 Model Description
4.4.2 Model Interpretation
4.4.3 Model Performance
4.5 Models Comparison
5 Conclusion
References
