Loan Eligibility Prediction Using Deep Learning

Abstract

A loan is a financial transaction in which one party, often a lender, lends money, commodities, or services to another party, known as the borrower, with the expectation of repayment in the future. Loans are often made with the idea that the borrower would repay the loan amount plus any relevant interest or fees over a set period of time. And the success and failure of these lending sectors depend on the ability to evaluate the credit risk, as it has become a significant role of financial institutions/banking sector to sanction loans. Based on their requirements and business rules they approved the loan after doing the all process manually which is time taking and is also Inefficient. Automating the process to identify loan eligibility is an efficient way to reduce the amount of time it takes and the credit risk. So, predicting if the loan applicant is eligible or not helps the banks to decide to start/not to start the sanctioning loan process. There are a number of researches done in this area and each of them played a vital role in contributing something important by developing the best model to predict the eligible applicant. But most of them used traditional ML classification algorithms which are not latest algorithms when compared to deep learning algorithms for prediction. Convolutional Recurrent Neural Network (CRNN) is preferable for prediction eligible applicant from the given loan data-set, which is the credit history of bank’s or lending authority's customers by extracting and learning complex and relevant features. In this thesis study, the proposed model is to predict loan applicant using CRNN. Our dataset, which contains 461097 rows and 44 columns of lending club data-set collected from kaggle online data set, is used to train and evaluate the CRNN model. The application of CRNN improves the performance of several machine-learning classification methods significantly. Our proposed CRNN model has an accuracy of 99.67% and loss of 0.0081. In comparison to earlier research works on loan eligibility prediction, we were able to attain higher prediction accuracy using the developed CRNN model.