Method and apparatus for optimizing training data of supervised learning, electronic device and medium

Abstract

The invention provides a method and a device for optimizing training data of supervised learning, an electronic device and a medium. Only a small amount of but not a full amount of the training data needs to be remarked. The method comprises the steps of 1, judging whether the quality of the training data reaches the standard or not, and if the quality of the training data reaches the standard, applying the training data to the training of a classification model, otherwise, entering the step 2; 2, dividing the training data into N parts, wherein N is an integer greater than 1; 3, selecting N-1parts in the N parts as a training set for training the classification model, taking the rest of one part as a test set, then estimating a classification result of the training data in the test set by utilizing a trained classification model, and screening the training data needed to be remarked according to the classification result; and 4, judging whether the test set in the step 3 is the lasttest set or not, and if yes, ending the optimization, otherwise, repeating the steps 3 and 4, until each part of the training data in the N parts is estimated as the test set.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a method, device, electronic device and medium for optimizing training data of supervised learning. Background technique [0002] Supervised learning refers to learning a function from the given training data. When new data arrives, the result of the new data can be predicted according to this function. The training data requirement for supervised learning is to include input and output (that is, classification values), which can also be said to be features and targets. Objects in the training data are annotated by humans. [0003] According to the foregoing description, it can be seen that supervised learning requires certain known categories of labeled data. In the existing supervised learning method, it is mainly to obtain a batch of data, manually mark each data category to obtain training data, and then use the training data combined with a specific algorithm to ...

AI Technical Summary

Problems solved by technology

And this kind of audit is an audit without focus, it is often difficult to find those problematic marked data, and the audit quality is not high

Images