A fraudulent data identification method and device

Abstract

The invention provides a fraudulent data identification method comprising the steps of training a preset training data set by using a preset successive type model training mode to build a successive type anti-fraud model; training to-be-tested data by using the successive type anti-fraud model to identify fraudulent data in the to-be-tested data. The invention also provides a fraudulent data identification device. Based on the characteristic that fraudulent data in to-be-tested data are non-balanced data, the successive type anti-fraud model is used for analyze and identify fraudulent data inthe to-be-tested data; compared with a common single model, the successive type anti-fraud model can improve the identification accuracy and recall rate of fraudulent data, more accurately judge fraudcases and reduce the range and cost of manual checking.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a method and device for identifying fraudulent data. Background technique [0002] At present, it is necessary to analyze and identify the fraudulent data in some fraudulent behaviors that are prone to occur. For example, there are some malicious or illegal card swiping and reimbursement behaviors in the social insurance and medical reimbursement system. The existence of these behaviors will waste medical resources and intensify social conflicts. . For tens of thousands of medical bills to be reimbursed, it is difficult to screen them one by one with limited human resources. Moreover, these data are unbalanced data, that is, fraudulent transaction data is relatively rare, and the common single model used for fraud data mining and prediction today only uses the maximum accuracy rate as the judgment standard, and the ratio of fraudulent transaction data to normal transacti...

AI Technical Summary

Problems solved by technology

[0002] At present, it is necessary to analyze and identify the fraudulent data in some fraudulent behaviors that are prone to occur. For example, there are some malicious or illegal card swiping and reimbursement behaviors in the social insurance and medical reimbursement system. The existence of these behaviors will waste medical resources and intensify social conflicts.

For tens of thousands of medical bills to be reimbursed, it is difficult to screen one by one with limited human resources

Moreover, these data are unbalanced data, that is, fraudulent transaction data is relatively rare, and the common single model used for fraud data mining and prediction today only uses the maximum accuracy rate as the judgment standard, and the ratio of fraudulent transaction data to normal transaction data is very high. When it is rare, the normal single model has a high correct estimation rate for the data set that needs to be judged fraudulent, which makes it difficult to identify and display the fraudulent transaction data in the data set, which makes the recognition accuracy and recall rate of the fraudulent transaction data biased by the model. Low

Images