Advanced Investigation of Healthcare Fraud Detection Utilizing Machine Learning Algorithms

Healthcare fraud is a fast-growing issue that causes substantial financial loss and affects the quality of patient care. Conventional fraud detection techniques tend to be ineffective in detecting fraudulent claims because healthcare data is complex and enormous in volume. This research investigates the use of machine learning methods to enhance fraud detection within healthcare systems. We contrast the performance of Decision Tree, Random Forest, K-Nearest Neighbors (KNN), and Logistic Regression both prior to and post-hyperparameter tuning and feature selection. Forward feature selection was done with KNN and Logistic Regression to improve model performance by choosing the most salient features, whereas hyperparameter tuning was utilized to fine-tune all the models. Metrics of evaluation like accuracy, precision, recall, F1-score, confusion matrix, and ROC curves were employed to measure the effectiveness of the models. The outcome reveals that Logistic Regression had the highest accuracy following optimization and feature selection over other models in identifying fraudulent claims. The Voting Classifier, which is an ensemble learning, enhanced fraud detection by aggregating various models for enhanced predictive capability. Though Decision Tree and Random Forest performed well, tuning was not effective in improving their accuracy. These results indicate that machine learning methods, especially ensemble models and feature selection, can dramatically improve healthcare fraud detection. Subsequent studies need to integrate deep learning and advanced ensemble techniques to further enhance fraud detection accuracy and reduce false positives.