Data-Driven Lending Decisions: Financial Parameter Modeling using ML Techniques
Abstract
This work involves depth exploration on loan eligibility classification with the aid of machine learning algorithms, i.e., Support Vector Machine (SVM) and K-Nearest Neighbors (KNN). With the aid of secondary set comprising loan applicants' financial and demographic features, both algorithms were designed, trained, and tested to determine their expertise in binary classification procedures. The given set included prominent features such as applicant and co-applicant income, loan amount, loan term, credit history, and location of property. The SVM model, designed with a radial basis function (RBF) kernel and standardized features, demonstrated improved accuracy (95%), precision (1.0), and F1 score (0.77) against others. Nevertheless, the KNN model, initialized with 5 neighbors, demonstrated comparatively higher AUC (0.87), indicating better probabilistic classification. Performance comparison tables, confusion matrices, and ROC curves were utilized to examine model behavior. Analysis reveals the strengths and trade-offs of both algorithms in realistic financial classification, portraying insightful information towards automatic loan screening devices.