Stacking Ensemble Approach for Predicting Loan Approval Using Machine Learning Techniques

Area of Science:

• Fintech and Digital Lending
• Machine Learning in Finance
• Credit Risk Assessment

Background:

• Digital lending and fintech innovations are transforming global financial inclusion and credit availability.
• Peer-to-peer (P2P) and digital lending platforms leverage technologies like AI and machine learning for loan approvals.
• Challenges in digital lending include algorithmic risk, customer trust, financial exclusion, and regulatory gaps.

Purpose of the Study:

• To examine the evolving landscape of digital lending and fintech.
• To propose a robust machine learning approach for accurate loan approval forecasting.
• To address challenges within the digital lending ecosystem through advanced analytics.

Main Methods:

• A stacking ensemble machine learning model was developed for loan approval prediction.
• Data preprocessing involved train-test partitioning, exploratory analysis, and label encoding.
• The ensemble model utilized XGBoost as a meta-learner with Gradient Boosting, Efficient Gradient Boosting, AdaBoost, and Extra Trees as base learners.

Main Results:

• The stacking ensemble model achieved a high accuracy of 98% in predicting loan approvals.
• Key factors influencing loan approvals include assets, income, and CIBIL scores.
• The model demonstrated superior performance and generalization compared to conventional methods.

Conclusions:

• The proposed machine learning model offers a powerful tool for automated, data-driven credit decisions.
• This approach can enhance efficiency and accuracy in the digital lending sector.
• The study highlights the potential of advanced ML techniques to improve financial inclusion and credit accessibility.