An abnormal surgical record recognition model with keywords combination patterns based on TextRank for medical insurance fraud detection.

Increasing insurance fraud has resulted in the loss of large amounts of money, making it difficult to expand insurance coverage and scale. This phenomenon is particularly acute in the field of health insurance. Medical insurance fraud is the falsification of medical records to obtain medical insurance funds or medical insurance benefits. Therefore, effective detection of health insurance fraud is of great importance for the rational use of health insurance funds. To address the frequent violations and frauds in health insurance, this paper proposes a keyword combination-based approach for health insurance fraud detection. First, a medical dictionary is built by TextRank to segment the surgical procedure text and extract the surgical keywords, then, the synonyms corresponding to each keyword are extracted from the electronic medical record data to form a keyword combination pattern as the final detection rule, and finally, a medical insurance fraud detection model is built on this basis. In this paper, data on acute myocardial infarction and unstable angina were selected for examination, with 1371 and 1787 cases, respectively. The performance of the model was evaluated by the coverage rate and compared experimentally with the TF-IDF and LDA algorithms. The experiments also prove the efficiency and advancedness of the algorithm in this paper. In the case of acute myocardial infarction, the method in this paper improved the coverage rate by 23.77% and 9.4% compared with the TF-IDF and LDA methods respectively. In the case of unstable angina, the coverage of the method in this paper was improved by 20.21% compared to both TF-IDF and LDA methods. [ABSTRACT FROM AUTHOR]