Credit Scoring for Loan Applicants Using Machine Learning Models

This study presents a comprehensive framework for credit scoring in loan application processes. The research presents a framework that uses data analysis and machine learning techniques to evaluate creditworthiness, in loan applications. By incorporating cutting-edge algorithms like Naive Bayes, SVM, Decision Tree, Multi Layer Perceptron (MLP), and Gradient Booster (GB) the goal is to revolutionize the decision-making process for loan approvals by using analytics. The model utilizes a dataset that includes payment statements enabling a thorough and randomized credit scoring system for each individual applying for a loan. By leveraging machine learning algorithms, the system examines past payment behaviors to extract patterns and information. These findings are then used to generate a credit score that is further classified as good, bad, or moderate. This approach provides an assessment of an applicant’s reliability. To improve the model’s accuracy, feature selection techniques and Synthetic Minority Oversampling Techniques (SMOTE) are utilized to compare models’ effectiveness, in predicting creditworthiness. The integration of data and machine learning not only enhances the efficiency of loan approval processes but also facilitates more informed financial decisions based on data analysis. On all the machine learning models which were performed, SVM performed best among the five models with an F-measure value of 0.86 incorporating feature selection and SMOTE.