A Graph-Based Method for Health Care Joint Fraud Detection
Ruicong Chen, Hao Zhang, Kai-Biao Lin · 2020
The development of health care insurance has brought many benefits to the Chinese people. However, the frequent occurrence of health care fraud has caused great damage to health insurance funds. Common fraud is mainly discovered by analyzing a patient's medication abnormalities and his/her abnormal costs. However, this individual fraud analysis method is not applicable to joint fraud. The fraud group needs to be explored through the potential relationships among patients. Nevertheless, due to the limited amount of available medical data, we cannot directly obtain the relationships among patients. Thus, this paper proposes a health care fraud detection method based on graph analysis to identify the joint fraud group. This method uses the similarity of patients' medical behaviors to mine the latent relationships between patients. Afterwards, it uses the community detection algorithm to find suspicious groups in the weighted patient network. Eventually, the clustering coefficient strategy is used in the community to find highly suspicious people. Experiments on a real medical dataset show that the proposed method can effectively identify fraud groups and achieves better performance than that of the existing fraud detection method.