Comparative Analysis of Cost Sensitive Ensemble Classifiers for Reducing False Negative Predictions in Consumer Loan Defaults

Author

Abdulrahman Adebola Iyaniwura


Abstract

 Consumer lending requires reliable default prediction to balance credit access, portfolio stability and financial losses. Conventional classifiers can overlook borrowers who subsequently default, particularly when default observations are uncommon and evaluation prioritises overall accuracy. This proposed study will compare cost-sensitive ensemble classifiers for reducing false-negative predictions in consumer loan default assessment. Default will constitute the positive class, making missed defaults the primary error of interest. Random forests, gradient boosting and adaptive boosting will be evaluated using consistent predictors, temporal partitions and development procedures. Misclassification costs, class weighting and decision thresholds will be tuned within training data, while independent evaluation will preserve observed default prevalence. Performance assessment will examine default recall, precision, precision–recall discrimination, calibration and expected lending costs alongside false-positive rates and approval implications. Sensitivity analyses will vary cost assumptions, default prevalence and threshold choices, with subgroup assessment examining uneven error burdens. Comparisons will distinguish improvements in detecting defaults from indiscriminate rejection of applicants. The study will identify trade-offs between missed-default reduction, predictive reliability and lending feasibility, without assuming that higher recall alone establishes economic value or equitable credit decisions.



Keywords

Consumer loan default; Cost-sensitive learning; Ensemble classifiers; False-negative reduction; Credit risk assessment; Decision thresholds.



Full Text:

Download Paper PDF


References


  1. Aruleba I, Sun Y. Effective credit risk prediction using ensemble classifiers with model explanation. IEEE Access. 2024 Aug 16;12:115015-25.
  2. IDOWU GA, OWOLABI JE. ENSEMBLE-BASED FRAUD DETECTION IN NIGERIAN BANKING: A SYNTHETIC DATA BENCHMARK AND COST-SENSITIVE ANALYSIS. 2026 Mar 9;6(1):107-29.
  3. Trinh LT. A comparative analysis of consumer credit risk models in Peer-to-Peer Lending. Journal of Economics, Finance and Administrative Science. 2024 Oct 11;29(58):346-65.
  4. Singh A. Peer-to-peer loan default prophecy in fintech: A comparative analysis of the predictive performance of machine learning models. Corporate Governance. 2024 Dec;6(2):1-2.
  5. Petrides G, Verbeke W. Cost-sensitive ensemble learning: a unifying framework. Data Mining and Knowledge Discovery. 2022 Jan;36(1):1-28.
  6. Ong P, Lee CN. A Comparative Study of Credit Default Risk Modelling with Ensemble Machine Learning Methods. InComputational Intelligence for Business Analytics and Decision Systems 2026 May 14 (pp. 239-249). Cham: Springer Nature Switzerland.
  7. Chukwunweike J. Coordinating PLC-based load shedding and variable-speed drives to prevent transformer overloading during industrial production peaks. Int J Electr Data Commun. 2025;6(2 Pt B):118-135. doi:10.22271/27083969.2025.v6.i2b.109.
  8. Opayemi OO. Credit Risk Default Prediction: A comparative analysis of Machine Learning Techniques(Doctoral dissertation, Dublin, National College of Ireland).
  9. Akuji II, Olanrewaju AB, Ahmed TA, Alabi AJ, Adeyemi IB. A comparative study of classical, bagging, and hybrid methods for optimizing loan default prediction. Computer Science and Information Technologies. 2026 Jul 1;7(2):179-95.
  10. Ogan RK. Machine-learning-based API failure prediction and root-cause analytics for improving reliability of enterprise banking integration architectures at scale. Int J Comput Artif Intell. 2021;2(2):142-153. DOI:10.33545/27076571.2021.v2.i2a.378.
  11. Ekinci A, Sen S. Forecasting bank failure in the US: A cost-sensitive approach. Computational Economics. 2024 Dec;64(6):3161-79.
  12. Cao J, Lu H, Wang W, Wang J. A loan default discrimination model using cost-sensitive support vector machine improved by PSO. Information Technology and Management. 2013 Sep;14(3):193-204.
  13. Alebiosu O. Engineering stakeholder alignment frameworks for resolving land rights, environmental compliance, and project delivery conflicts. Int J Appl Res. 2024;10(12):401-412.
  14. Kim J, Choi K, Kim G, Suh Y. Classification cost: An empirical comparison among traditional classifier, Cost-Sensitive Classifier, and MetaCost. Expert Systems with Applications. 2012 Mar 1;39(4):4013-9.
  15. Akinjole A, Shobayo O, Popoola J, Okoyeigbo O, Ogunleye B. Ensemble-based machine learning algorithm for loan default risk prediction. Mathematics. 2024 Oct 31;12(21):3423.
  16. Bhandari, Ambika Baniya. 2026. “Bridging the Public Health–Cloud Security Divide: Workforce Implications of the HIPAA Security Rule Modernization for Public Health Informatics”. Current Journal of Applied Science and Technology45 (8):33-44. https://doi.org/10.9734/cjast/2026/v45i84730.
  17. Taha AH, Abu-Naser SS. Predicting Loan Defaulters: A Comprehensive Analysis and Comparative Study of Machine Learning Algorithms Using a Large-Scale Loan Default Dataset. InProceedings of Eighth International Conference on Information System Design and Intelligent Applications 2025 Jan 3 (pp. 15-28). Singapore: Springer Nature Singapore.
  18. Ramos SM. Cost-Sensitive Machine Learning for Peer-To-Peer Credit Scoring: Balancing Predictive Accuracy, Risk and Regulatory Compliance(Master's thesis, Universidade NOVA de Lisboa (Portugal)).
  19. Bhandari, Ambika Baniya. 2026. “Quality Control Frameworks for Electronic Health Record Data Workflows in Resource-Constrained Settings: A Structured Review With Implications for Global Health Information Management”. Current Journal of Applied Science and Technology45 (8):1-16. https://doi.org/10.9734/cjast/2026/v45i84728.
  20. Neema S, Soibam B. The comparison of machine learning methods to achieve most cost-effective prediction for credit card default. Journal of Management Science and Business Intelligence. 2017 Aug;2(2):36-41.
  21. Papouskova M, Hajek P. Two-stage consumer credit risk modelling using heterogeneous ensemble learning. Decision support systems. 2019 Mar 1;118:33-45.
  22. Iyanuoluwa Oluwafemi. (2024). ENVIRONMENTAL MONITORING STRATEGIES INTEGRATING MOLECULAR DETECTION TECHNOLOGIES TO ENHANCE AGRICULTURAL BIOSECURITY AND SUSTAINABLE CROP PRODUCTION SYSTEMS. International Journal Of Engineering Technology Research & Management (IJETRM), 10(12), 757–772. https://doi.org/10.5281/zenodo.21152414
  23. Li Z, Zhang J, Yao X, Kou G. How to identify early defaults in online lending: A cost-sensitive multi-layer learning framework. Knowledge-Based Systems. 2021 Jun 7;221:106963.
  24. Aleksandrova Y, Armianova M. Evaluation of cost-sensitive machine learning methods for default credit prediction. In2022 International Conference Automatics and Informatics (ICAI) 2022 Oct 6 (pp. 89-94). IEEE.
  25. Bhetwal S, Bastakoti R, Acharya N, Gupta GK, Bhandari AB. Benchmarking Local LLMs for Natural-Language-to-SQL Querying in Biopharmaceutical Manufacturing: An Empirical Benchmark on Consumer-Grade Hardware. arXiv preprint arXiv:2606.01338. 2026 May 31.
  26. Chen N, Ribeiro B, Chen A. Comparative study of classifier ensembles for cost-sensitive credit risk assessment. Intelligent Data Analysis. 2015 Jan;19(1):127-44.
  27. Gaffari MA. Explainable machine learning for predicting depression risk using sleep patterns, self-esteem, and psychosocial factors. International Journal of Computer Applications Technology and Research. 2026;15(11):26–37. doi:10.7753/IJCATR1511.1003.
  28. Xiao J, Li S, Tian Y, Huang J, Jiang X, Wang S. Example dependent cost sensitive learning based selective deep ensemble model for customer credit scoring. Scientific Reports. 2025 Feb 18;15(1):6000.
  29. Luo Z. An Adaptive Ensemble Learning Approach with Cost-Sensitive Optimization for Credit Default Prediction. InProceedings of the 2026 5th International Conference on Cyber Security, Artificial Intelligence and Digital Economy 2026 Mar 13 (pp. 523-531).
  30. Sandra Darkwa. Transformer-Driven Strategic Communication Analytics for Identifying Synthetic Media Manipulation and Coordinated Narrative Amplification Networks Digitally. Int J Literacy Educ 2024;4(2):333-346. DOI: 22271/27891607.2024.v4.i2d.434
  31. Yotsawat W, Wattuya P, Srivihok A. A novel method for credit scoring based on cost-sensitive neural network ensemble. Ieee Access. 2021 May 25;9:78521-37.

Share your valuable work from Social Media Buttons