Advanced Healthcare Analytics using AI and ML for Insurance Pattern Detection

Abstract

Healthcare insurance fraud has emerged as a significant challenge for insurance providers due to the increasing volume and complexity of healthcare claim transactions. Fraudulent activities such as false claims, duplicate billing, exaggerated treatment costs, and unnecessary medical procedures result in substantial economic losses and reduce the efficiency of healthcare insurance systems. This study develops an advanced healthcare analytics framework for insurance pattern detection by integrating Isolation Forest, XGBoost (Extreme Gradient Boosting), and SHAP (SHapley Additive exPlanations). The novelty of the proposed framework lies in its unified integration of anomaly detection, fraud classification, and explainable artificial intelligence, enabling the identification of suspicious claim patterns, accurate fraud prediction, and transparent interpretation of model decisions within a single analytical pipeline. Initially, Isolation Forest is employed to identify anomalous claim behaviors and suspicious insurance patterns from healthcare claim records. The anomaly-informed data are subsequently analyzed using XGBoost to classify insurance claims into fraudulent and legitimate categories. To enhance transparency and interpretability, SHAP is incorporated to quantify the contribution of individual claim attributes and explain model predictions. Experimental evaluation was conducted using the publicly available Health Insurance Claims Data for Fraud Detection dataset and classification performance was evaluated using Accuracy, Precision, Recall, F1-Score, and ROC-AUC. Among the machine learning models evaluated, XGBoost demonstrated superior performance with an accuracy of 97.46%, precision of 97.12%, recall of 96.85%, F1-score of 96.98%, and ROC-AUC of 0.984. The results demonstrate that the developed framework provides accurate, reliable, and interpretable insurance pattern detection for healthcare fraud analytics.