Using Boosting and Neural Networks Methods to Detect Healthcare Fraud

Global demographic change has played a key role in transforming hospital operations, especially in meeting the increased demand while maintaining efficiency and integrity. The introduction of Artificial Intelligence (AI) in healthcare offers a promising path for addressing these challenges, enabling improved operational efficiency and better quality of care. Incorporating AI enhances the Supply Chain of Health Units, that is crucial for their operational success. Additionally, the combination of AI and Machine Learning (ML) within healthcare is not just about improving operations but also plays an important role in financial oversight, as there is a pressing need to reduce the economic burden on governments and prevent the misuse of the system through fraudulent activities due to rising healthcare costs. Our research uses advanced ML models, such as XGBoost, CatBoost, LSTM Neural Networks and MLP Classifier, evaluated through Accuracy, ROC-AUC and F1-Score metrics, to develop predictive models that can identify potential fraud in hospital operations and accurately estimating annual deductibles for inpatient care. In this effort, the CatBoost model proved to be the best model with an accuracy of 93.99%, ROC-AUC of 94.86% and F1-Score of 66.32%, while second came the Logistic Regression model with slightly lower performance. The MLP Classifier and the LSTM Neural Network followed, having partly lower evaluations, showing the superiority of the boosting methods.