An Integrated Approach Using Ensemble Machine Learning and Deep Learning for Loan Approval Prediction

The suggested loan prediction system overcomes the drawbacks of conventional selection processes by utilizing machine learning (ML) to enhance the identification of creditworthy borrowers. A Random Forest (RF) model to predict loan is evaluated in our study along with data pretreatment using the Synthetic Minority Over-Sampling Technique (SMOTE) to solve the class imbalance. Nine ML models in all, including Deep Learning (DL) approaches, ensemble learning strategies, and conventional classifiers, were used in this work to reliably and successfully forecast loan acceptance results. According to a number of characteristics, the dataset utilized in this study aims to predict whether a loan application will either be approved or denied. The RF model obtained 94.10% of F1-Score, an Accuracy of 94.12%, Precision and Recall values of 94.24% and 94.12%, respectively, following the use of SMOTE for maintaining equilibrium of data.These findings demonstrate RF's effectiveness in managing unbalanced datasets and its capacity to forecast loan approval outcomes. Because they automate the approval process, lower human error, and increase decision-making efficiency, loan prediction models are essential to financial organizations.