An anti-fraud method for behavior recognition based on sensor data

Abstract

The invention relates to the technical field of behavior recognition, and provides an anti-fraud method for behavior recognition based on sensor data, and the main technical scheme comprises the steps: firstly, processing original data by adopting a standardized data preprocessing scheme, and classifying the data by using a standardized offline model to obtain a preliminary classification result and a classification confidence coefficient; Reserving the sample tags with higher total recognition confidence of the preliminary classification result; data samples with lower identification confidence are identified according to timestamps; mapping the data back to the original data, and performing data preprocessing again by using a non-standardized data preprocessing scheme corresponding to the classification label of the original data; and identifying the obtained data again by using the corresponding non-standard offline model, marking the samples with the classification confidence lowerthan a preset threshold as unknown behaviors, and reserving the sample labels with the confidence higher than the threshold. And performing dynamic time sequence feature capture on the behavior sequence input online model obtained after the two times of recognition to obtain a classification result.

Description

technical field [0001] The present application relates to the field of data mining, in particular to an anti-fraud method for behavior recognition based on acceleration sensor and gyroscope sensor data. Background technique [0002] In recent years, with the rapid development of mobile communications and the Internet, as well as the high popularity of smart mobile terminals, more and more scenes in life can be developed efficiently and conveniently in a purely online manner. However, due to the fact that the purely online mode of operation cannot directly contact users, some organizations that adopt this mode of operation cannot verify the authenticity and validity of customers when facing customers' incoming documents, and various malicious fraud methods have been bred. . Fraud can take different forms, the simplest one includes using multi-control machines to manipulate multiple accounts at the same time, simulating real user interaction situations, and obtaining a large ...

AI Technical Summary

Problems solved by technology

[0006] The purpose of this application is to address the problems existing in the existing technical means, to provide a dynamic, non-perceptual anti-fraud method, to improve the flexibility and traceability of anti-fraud, and to raise the issue of the difficulty in obtaining training samples for behavior recognition models. solution

Images