An Explainable Machine Learning Framework for Behavioural Risk Classification in Financial Services
Main Article Content
Abstract
Financial solutions are getting to ever increasing demands and requirements for using a transparent data-driven approach to assess behavioral risk from complex financial customer data. This study presents an explainable machine learning model based on demographic, financial and transactional data, applied to explaining how to identify engineered behavioral risk states. Processing the models like Logistic Regression, Random Forest and XGBoost are evaluated on an 80:20 stratified traintest split and by five folds of test set cross-validation, and the model is interpreted using SHAP. Logistic regression had the best test accuracy at 79.55% and the mean and cross-validation accuracy is at 80.01%. The results show similar model performance and underline the importance of explainability for uncovering salient financial risks features and for a transparent financial decision-making.