Year: 2024
Type: article
Abstract: This study investigates the application of machine learning methods to identify fraudulent claims in animal healthcare. The researchers utilized publicly available veterinary claims data and regulatory exclusion databases to label fraudulent claims. Three supervised machine learning models were employed: C4.5 Decision Tree, Logistic Regression, and Support Vector Machine. Their performance was evaluated using metrics such as Area Under the ROC Curve, False Positive Rate, False Negative Rate, and Precision-Recall. The findings demonstrate that the C4.5 Decision Tree outperformed the other two models in terms of AUC, recall, and FNR, making it the most effective approach for detecting fraudulent claims in veterinary healthcare. The C4.5 model achieved an AUC of 0.883 at an 80:20 class distribution and exhibited the lowest FNR, successfully identifying fraudulent claims without missing significant instances of fraud. Although Logistic Regression showed high precision, it had a higher FNR, indicating a trade-off between precision and recall. SVM exhibited lower overall performance compared to the other models, particularly in AUC and FNR. The results highlight the potential of machine learning to enhance fraud detection systems in animal healthcare, providing a robust approach to identifying fraudulent claims that may otherwise be overlooked. Future research could focus on exploring additional data sources, feature engineering, and alternative sampling techniques like Synthetic Minority Over-sampling Technique to further improve the detection process. This study contributes to the growing body of work aimed at leveraging machine learning to detect fraud and ensure the proper allocation of resources in animal healthcare.
Author: Dr. Kondragunta Rama Krishnaiah
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4410059818