IMPROVING LOAN PREDICTION ACCURACY USING ENSEMBLE MODELS AND IMBALANCE LEARNING METHODS

Abstract

Accurate loan prediction is essential for financial institutions to reduce credit risk and improve decisionmaking
in loan approval processes. Traditional credit evaluation methods often rely on manual
assessment and basic statistical techniques, which may fail to effectively analyze large volumes of
financial data and identify complex patterns related to borrower behavior. In recent years, machine
learning techniques have been widely adopted to automate loan prediction and improve the accuracy of
credit risk assessment. However, one of the major challenges in loan prediction datasets is class
imbalance, where the number of approved loans is significantly higher than default cases, leading to
biased model performance. This study proposes an approach to improve loan prediction accuracy by
combining ensemble learning models with imbalance learning techniques. Ensemble models such as
Random Forest, Gradient Boosting, and AdaBoost are used to enhance predictive performance by
integrating multiple learning algorithms. In addition, imbalance learning methods such as Synthetic
Minority Over-sampling Technique (SMOTE) and other resampling strategies are applied to address the
imbalance problem in loan datasets. The proposed framework evaluates model performance using metrics
such as accuracy, precision, recall, and F1-score to determine the most effective prediction strategy.
Experimental results demonstrate that the combination of ensemble models with imbalance learning
methods significantly improves loan prediction accuracy and provides a more reliable system for financial
decision-making. This approach can assist banks and financial institutions in minimizing credit risk while
ensuring fair and efficient loan approval processes