Classifier training method, apparatus, electronic apparatus, and computer readable medium

Abstract

The invention provides a classifier training method, a device, an electronic device and a computer-readable medium, belonging to the technical field of the Internet. A method for train that classifierinclude training labeled samples in a plurality of data feature to obtain a primary classifier; Inputting an unlabeled sample from the plurality of data features to the primary classifier for training to output a label of the unlabeled sample, and calculating a probability that the unlabeled sample labels the label; According to the probability of the label, a part of the unlabeled samples is selected for expert labeling, and the samples labeled by the expert are retrained to obtain a new classifier. This method is based on active learning to update the classifier. Combined with a large number of data and artificial experience, the classifier model can make the classification results more accurate and improve the classification effect. There are only a small number of labeled samples anda small number of manual labeling in the training samples, which can quickly process the classification of a large number of unlabeled samples and improve the efficiency.

Description

technical field [0001] The present disclosure generally relates to the technical field of the Internet, and in particular, relates to a classifier training method, device, electronic equipment, and computer-readable medium. Background technique [0002] In recent years, with the rapid development of the Internet, there are more and more Internet transaction scenarios, and the number of merchants on various Internet platforms is huge, and some unscrupulous merchants also have fraudulent behavior. The behaviors of these fraudulent merchants include: selling stolen goods (accounts and bank card information stolen by Internet transactions), money laundering, cashing out, or secretly operating illegal businesses (such as gambling, etc.). The interests of the platform are damaged, and even illegal activities such as money laundering and gambling occur. [0003] The existing risk control methods for fraudulent merchants mainly include the following types: 1) Counter-check the asso...

AI Technical Summary

Problems solved by technology

These risk control methods also have their own shortcomings: 1) The method of correlation check can only be traced back passively, and cannot be actively identified

And the merchants purchased by fraudulent users are not necessarily fraudulent merchants, there is a certain degree of randomness, and it is difficult to fully control the risk

2) The rules customized by expert experience rely heavily on manual experience, and are relatively rigid, and are easily bypassed by temptation

The loopholes between the rules are not easy to be found. Once the number is large, the logical relationship is difficult to sort out, and the recognition effect cannot achieve the expected effect.

3) The method of naturally accumulating data and labels requires huge time and cost. Only when the cases are confirmed can effective case samples be left

In addition, fraudulent merchants are generally used to sell stolen goods and make false transactions. There are no users as victims, and no users report the case. It is difficult to find their fraudulent behavior, which makes it more difficult to accumulate samples in actual operation.

4) The unsupervised method requires a large amount of data, and the number of merchants is far smaller than the number of transactions, and there are huge differences between merchants in different industries, and the number of subdivisions is less, so it is difficult to model with an unsupervised method

Images