A Novel System Architecture for Anomaly Detection for Loan Defaults

Abstract

Given the rise in loan defaults, especially after the onset of the COVID-19 pandemic, it is necessary to predict if customers might default on a loan for risk management. This paper proposes an early warning system architecture using anomaly detection based on the unbalanced nature of loan default data in the real world. Most customers do not default on their loans; only a tiny percentage do, resulting in an unbalanced dataset. We aim to evaluate potential anomaly detection methods for their suitability in handling unbalanced datasets. We conduct a comparative study on different classification and anomaly detection approaches on a balanced and an unbalanced dataset. The classification algorithms compared are logistic regression and stochastic gradient descent classification. The anomaly detection methods are isolation forest and angle-based outlier detection (ABOD). We compare them using standard evaluation metrics such as accuracy, precision, recall, F1 score, training and prediction time, and area under the receiver operating characteristic (ROC) curve. The results show that these anomaly detection methods, particularly isolation forest, perform significantly better on unbalanced loan default data and are more suitable for real-world applications.