HHM:A Hybrid Model for Predicting Medicare Violation Data

With the imbalance between medical supply and demand, coupled with the shortage of medicare funds, some hospitals engage in fraudulent practices by overcharging for unreasonable expenses. It is essential to identify such irregularities and provide feedback to the medicare authority for better supervision. In order to address this issue, this paper proposes a hybrid integrated model called the Hidden Hybrid Model (HHM). The model combines a BP neural network with a decision tree, utilizing the hidden layer to generate fused features that serve as inputs to the decision tree, resulting in a collaborative model. Subsequently, this collaborative model acts as a strong classifier and is integrated with individual BP neural networks and decision trees through voting to form the final integrated model. Experimental results on a real dataset from a municipal-level hospital showed that the model achieved an accuracy of 99.5% in predicting five-class violation outcomes with a data volume of 100,000. The model demonstrates strong robustness and performs well in predicting violation results.