Home Business & Economics Chapter 12 Optimization of debt collection strategies for South African banks with machine learning models
Chapter
Licensed
Unlicensed Requires Authentication

Chapter 12 Optimization of debt collection strategies for South African banks with machine learning models

  • Jabulani Monchwe , Elias Munapo and Martin Chanza
Become an author with De Gruyter Brill
Explore this Subject How to publish with us
Smart Green Energy Production
This chapter is in the book Smart Green Energy Production

Abstract

Once banks are no longer able to collect money from defaulted accounts, they turn to third-party debt collectors to recover the funds. These debt collectors use aggressive tactics, such as contacting the debtor’s family and friends, to try to locate the money. Because of this, many people have become afraid of ever becoming delinquent on their loan payments again. This has led to a decrease in the number of people who are able to become debt-free. This chapter presents machine learning models for optimization debt collection for South African banks. The main objective of this study is to compare the accuracy and cost optimization of logistic regression (LR), random forest (RF), multilayer perceptron (MLP) and decision tree (DT) models using bank delinquency data. Findings reveal that the LR and RF classifiers achieved the highest debt recovery rates. Since the LR model only considers SMS as a communication mode compared to RF, which has diverse communication modes, the RF model is selected as the optimal model to prescribe communication strategies. Communication strategies are based on classifiers that have the highest Hamming score, high subset accuracy, lowest action cost and diversity of communication modes.

Abstract

Once banks are no longer able to collect money from defaulted accounts, they turn to third-party debt collectors to recover the funds. These debt collectors use aggressive tactics, such as contacting the debtor’s family and friends, to try to locate the money. Because of this, many people have become afraid of ever becoming delinquent on their loan payments again. This has led to a decrease in the number of people who are able to become debt-free. This chapter presents machine learning models for optimization debt collection for South African banks. The main objective of this study is to compare the accuracy and cost optimization of logistic regression (LR), random forest (RF), multilayer perceptron (MLP) and decision tree (DT) models using bank delinquency data. Findings reveal that the LR and RF classifiers achieved the highest debt recovery rates. Since the LR model only considers SMS as a communication mode compared to RF, which has diverse communication modes, the RF model is selected as the optimal model to prescribe communication strategies. Communication strategies are based on classifiers that have the highest Hamming score, high subset accuracy, lowest action cost and diversity of communication modes.

Chapters in this book

  1. Frontmatter I
  2. Contents V
  3. List of authors VII
  4. Chapter 1 Use of digital systems in the design system of photovoltaic solar stations 1
  5. Chapter 2 Potential wind energy in Turkmenistan 21
  6. Chapter 3 Potential of using biogas technology in Turkmenistan 31
  7. Chapter 4 Energy efficiency 45
  8. Chapter 5 Latent renewable energy in Turkmenistan 57
  9. Chapter 6 Approximate stochastic simulation algorithms 67
  10. Chapter 7 The role of supply chain management in the construction industry 95
  11. Chapter 8 Selection of threshold in binary graphs of biological networks 121
  12. Chapter 9 Model selection criteria with bootstrap algorithms: applications in biological networks 133
  13. Chapter 10 Technocracy in Governance: new directions in city functioning and urban planning 149
  14. Chapter 11 Outlier detection in biomedical data: ECG-focused approaches 161
  15. Chapter 12 Optimization of debt collection strategies for South African banks with machine learning models 183
  16. Chapter 13 Performance of six turbulence models in predicting two-phase flow on a hydraulic test bench 209
  17. Index 231
https://doi.org/10.1515/9783110717655-012

Chapters in this book

  1. Frontmatter I
  2. Contents V
  3. List of authors VII
  4. Chapter 1 Use of digital systems in the design system of photovoltaic solar stations 1
  5. Chapter 2 Potential wind energy in Turkmenistan 21
  6. Chapter 3 Potential of using biogas technology in Turkmenistan 31
  7. Chapter 4 Energy efficiency 45
  8. Chapter 5 Latent renewable energy in Turkmenistan 57
  9. Chapter 6 Approximate stochastic simulation algorithms 67
  10. Chapter 7 The role of supply chain management in the construction industry 95
  11. Chapter 8 Selection of threshold in binary graphs of biological networks 121
  12. Chapter 9 Model selection criteria with bootstrap algorithms: applications in biological networks 133
  13. Chapter 10 Technocracy in Governance: new directions in city functioning and urban planning 149
  14. Chapter 11 Outlier detection in biomedical data: ECG-focused approaches 161
  15. Chapter 12 Optimization of debt collection strategies for South African banks with machine learning models 183
  16. Chapter 13 Performance of six turbulence models in predicting two-phase flow on a hydraulic test bench 209
  17. Index 231
Sign up now to receive a 20% welcome discount
Institutional Access
How does access work?
Have an idea on how to improve our website?
Please write us.
© 2025 De Gruyter Brill
Downloaded on 12.12.2025 from https://www.degruyterbrill.com/document/doi/10.1515/9783110717655-012/html
Scroll to top button