Loan Eligibility System Using Machine Learning

Abstract

As people's demands grow, so does the need for loans to meet some of them. Every day, financial institutions receive numerous loan applications. The financial industry faces significant challenges in managing loan defaults, leading to increased fraud, bad debt, and financial losses. Financial institutions frequently lack reliable methods for determining a loan applicant's creditworthiness, particularly in Africa. The main objective of the study is to develop a machine learning-based loan eligibility system that accurately predicts the creditworthiness of loan applicants in the African financial sector. This study addresses a critical issue in the African financial industry by implementing a reliable predictive model that reduces risks and informs lending decisions. The methodology combines Design Science Research (DSR) and the Cross-Industry Standard Process for Data Mining (CRISP DM) framework. The data used includes parameters such as the applicant's net salary, loan amount, loan term, credit history, and employment status. Cleaning, standardization, and dividing the dataset into training and testing sets were all part of the data preparation process. The Logistic Regression algorithm was utilized for model training, implemented in Python using the Scikit-learn library. The Logistic Regression model's performance was evaluated using the accuracy metric, achieving an accuracy rate of 80%. The developed machine learning model demonstrated high accuracy in predicting loan eligibility, indicating its potential effectiveness in real-world applications. The system was deployed as a web-based application using the Streamlit framework, providing an accessible tool for financial institutions. Future research and improvements, such as data enrichment, advanced analytics, and adherence to financial regulatory compliance, are necessary to further enhance the system's accuracy and reliability.