Loan Prediction System with Exploring Explainable AI Transfer Learning with SHAP

This paper introduces a versatile system for analyzing loan approval datasets using Python libraries. The system integrates GUI components via Tkinter, enabling file selection anddataset loading. It incorporates data preprocessing techniques, transforming non-numeric data into numeric formats and visualizing loan status and rejection reasons through bar graphs. The datasetundergoes normalization and splitting, facilitating training and testing using a Random Forest Classifier from the Scikit-learn library. Performance metrics such as accuracy, precision, recall, and F1-score are calculated and displayed along with confusion matrices for insightful model evaluation. The system extends its functionality by allowing users to predict loan approval status and rejection reasons for new datasets. Furthermore, it integrates SHAP (SHapley Additive exPlanations) to offer explainable AI, visualizing the impact of features on loan approval predictions. This comprehensive system offers a holistic approach to loan approval analysis, providing prediction capabilities and interpretability through explainable AI techniques.