Data classification method and device, equipment and computer readable storage medium

Abstract

The invention provides a data classification method and device, equipment and a computer readable storage medium. On the condition that two classes of samples are unbalanded, a few of same-class sample sets are generated through downsampling for the samples with a large number, and new samples are generated through upsampling for the samples with a small number; the new samples are mixed with thesamples with the small number to form samples with a large number, so that the sample number of the sample sets with the small number and the sample number of the sample sets with the large number arebalanced, and the samples with the small number and the samples with the large number predict data through multiple times of modeling, and finally a prediction result with the number advantage is taken as a classification result; and the accuracy of data preciditon is improved by means of upsampling, downsampling, multiple times of modeling and multiple times of prediction.

Description

Technical field [0001] This application relates to the field of information processing technology, and in particular to a data classification method, device, equipment, and computer-readable storage medium. Background technique [0002] At present, in the process of data modeling and data classification, especially in the case of multi-classification, there are often problems of imbalance in the appearance of various types of samples. When the number of training samples of various types is quite different, unbalanced samples are directly used If the classification model is obtained by training, the result of model training may be very unsatisfactory due to the imbalance of the number of various samples. Then the prediction result obtained by using the trained model to predict is not ideal, or even the opposite. [0003] At present, the more common practice is to increase the number of samples by generating new samples with a smaller number of samples to achieve a level that is bala...

AI Technical Summary

Problems solved by technology

[0002] At present, in the process of data modeling and classification of data, especially in the case of multi-classification, there are often problems of class imbalance in various samples. When the number of training samples of various types is quite different, the unbalanced samples are directly used. If the classification model is obtained by training, the result of model training may be very unsatisfactory due to the imbalance of the number of samples of various types. Then the prediction result obtained by using the trained model for prediction is not ideal, or even the prediction result is opposite.

[0003] At present, the more common practice is to increase the number of samples by generating new samples for those samples with a small number, so as to reach the level of balance with the number of samples with a large number. The new samples often need to be as close as possible to the real samples, but After all, the new sample is not a real sample, and the model obtained by using it as a model training has a certain adverse effect on the prediction results of the data. If the one-time prediction result obtained by combining the new sample with the original sample for a single modeling prediction error, the result will be irreversible

Images