Healthcare Fraud Detection Based on Trustworthiness of Doctors

Big data is now rapidly expanding into various domains such as banking, insurance and e-commerce. Data analysis and related studies have attracted more attentions. In health insurance, abuse of diagnosis is one of the key fraud problems, which damages the interests of insured people. To address this issue, numbers of studies have focused on this topic. This paper develops a healthcare fraud detection approach based on the trustworthiness of doctors to distinguish fraud cases from normal records. Compared to conventional methods, our approach can detect healthcare fraud in a good accuracy by only little feature information from healthcare data without the violation of privacy. This approach combines a weighted HITS algorithm with a frequent pattern mining algorithm to calculate a rational treatment model of a certain disease. In addition, this paper also introduces the copy precision behavior in the treatment sequences of patients, which is a critical metric to learn the trustworthiness of doctors. The numerical validation with a healthcare dataset demonstrates that healthcare fraud by misdiagnosis in healthcare treatments can be successfully detected by employing the developed fraud detection approach.