Enhancing Loan Default Prediction Accuracy in Nigerian Banks: A Hybrid ANN-LSTM Approach

Abstract

Loan defaults pose a significant challenge to Nigerian banks, threatening financial stability and profitability. Existing predictive models often lack the accuracy needed for effective risk management. This study proposes a hybrid Artificial Neural Network-Long Short-Term Memory (ANN-LSTM) model to enhance loan default prediction accuracy. Leveraging a dataset of 148,670 loan records with 37 features from Nigerian banks, the model integrates ANN's ability to capture complex patterns with LSTM's proficiency in processing sequential data. Performance evaluation using accuracy, precision, recall, F1-score, and Area Under the ROC Curve (AUC) demonstrates the hybrid model's superiority over traditional approaches. The ANN-LSTM model achieved 99.7% accuracy and 100% AUC, significantly outperforming Naive Bayes and Logistic Regression models. These results suggest that the proposed hybrid approach can substantially improve risk assessment and decision-making processes in Nigerian banks, potentially reducing loan default rates and enhancing overall financial stability.