Performance analysis of deep and machine learning algorithms for loan evaluation model

Abstract

In this study, we present a loan evaluation model that uses machine and deep learning algorithms using data obtained from a local bank in Ethiopia. We examined two important experiments: the first used a one-dimensional convolutional neural network deep learning method, while the second employed machine learning methods such as support vector machines, XGBoost, random forests, decision trees, and Naive Bayes classifiers. We train and implement the algorithms to decide whether to accept or reject a loan application. A comparison of the model performance under different performance metrics is provided. According to the experimental findings, machine learning algorithms outperform deep learning algorithms in terms of classification accuracy, precision, recall, and area under the curve (AUC). Therefore, from the experimental results, we draw the conclusion that Ethiopian banks should think about utilizing machine learning models for their loan evaluation process rather than relying on more subjective traditional methods.