Class Balancing and Random Subspace based Ensemble Learning for Credit Scoring
Qaisar, Rabia Haider, Usman Khalid, Fatima Baig, Muhammad Muneeb
Qaisar, Rabia
Haider, Usman
Khalid, Fatima
Baig, Muhammad Muneeb
Type
Supervisor
Subject
Date
2026-01-13
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
The banking industry faces immense difficulty when evaluating credit risk. The process of manually assessing a client's creditworthiness involves intricate steps and demands significant time investment. The financial sector requires an automated model that simultaneously enhances prediction accuracy and shortens application processing times. This research document presents an innovative credit scoring system intended to improve credit risk evaluation. The dataset's class imbalances are managed through the application of an effective oversampling method. An ensemble learning strategy utilizes the Random Subspace strategy to integrate multiple base learners for enhanced performance. The model's performance evaluation on the Australian creditworthiness dataset yields remarkable outcomes. The model attains a highest accuracy score of 87.54% with 87.34% F-measure, and 92.50% AUC values.
Department
Publisher
Sponsor
Effat University
Copyright
CC0 1.0 Universal