Designing an Intelligent Fraud Detection System for Healthcare Insurance Claims Using a Machine Learning Approach

Health insurance is a valuable service that provides consumers with access to essential medical assistance during critical times. There is a substantial monetary effect from the complicated issue of health insurance fraud. Machine learning techniques have gained superiority over conventional methods to detect fraud because of advances in computational power together with the collection of large data quantities. The research brings a leading-edge ML method for detecting fraudulent claims within healthcare insurance systems. To solve class imbalance, the study uses a Kaggle dataset and applies strong preprocessing techniques such as feature scaling, data cleaning, and synthetic data balancing using ADASYN. Feature selection is performed using mutual information to enhance model efficiency. A fraud detection assessment involves evaluation of Extra Trees and ‘Random Forest’ and ‘Decision Tree’ classification models for the purpose of assessing ML models' efficacy in detecting fraud using the f1-score(f-measure), recall(rec), accuracy(acc), precision (prec), and ROC measures. The findings show that the Extra Trees Classifier outperforms the current logistic regression and XGBoost classifiers, as well as Random Forest (95%) and Decision Tree (93%), with the greatest accuracy of 96%. The results of the study highlight the value of ensemble learning strategies in raising the precision of fraud detection and bolstering the dependability of processing medical insurance claims.