An Intelligent Hybrid Data Mining Framework for Healthcare Fraud Detection Using Machine Learning and Association Rule Mining

Abstract

Healthcare insurance fraud is one of the biggest healthcare challenges today and all over the world. The conventional fraud detection approaches are based on manual auditing and rule-based systems and tend to be ineffective in detecting fraudulent behaviors as they change over time. In this paper, Association Rule Mining (ARM), Random Forest (RF) and Isolation Forest (IF) algorithms are combined to provide a novel hybrid data mining for intelligent healthcare fraud detection. The novelty of the proposed approach is the fusion of supervised and unsupervised learning techniques as well as of the adaptive risk scoring, which will enhance the detection accuracy and reduce false positives. The framework uses an insurance claim data set from healthcare to identify any unusual patterns and transactions which could be suspicious. The experimental results show that the proposed hybrid model outperforms the conventional machine learning models in terms of accuracy, precision, recall, and F1-score. Under highly imbalanced datasets, the proposed framework achieved an accuracy of 97.2% and significantly enhanced the fraud detection performance. Research provides a scalable and explainable data mining solution appropriate for real-world health care systems.

Keywords