Unbalanced industrial data classification method and device, and computer readable storage medium

Abstract

The invention relates to an unbalanced industrial data classification method and device, and a computer readable storage medium. The unbalanced industrial data classification method comprises the following steps: dividing an input unbalanced data set into a minority class sample set and a majority class sample set; in the sampling process, a plurality of balanced subsets are obtained in a mode of inducing missing values for a minority class sample set and estimating and complementing the missing values; training each balanced subset to obtain a sub-classifier; and integrating all the sub-classifiers to obtain a final data classifier. Missing values are induced for minority samples in the sampling process, estimation and complementation are carried out, and the problem of potential useful information missing caused by downsampling can be relieved.

Description

technical field [0001] The present disclosure relates to the field of emerging information technologies, and in particular to a method and device for classifying unbalanced industrial data, and a computer-readable storage medium. Background technique [0002] Aiming at the high complexity and high integration of industrial systems, the data-driven machine learning model can monitor the operating status of equipment, mine and discover the rules of industrial equipment operation, perform predictive maintenance on faults, and prevent accidents. Industrial production saves costs and improves efficiency. [0003] Unbalanced industrial data is ubiquitous in actual scenarios, which greatly affects the prediction effect. Contents of the invention [0004] The inventor found through research that the EasyEnsemble algorithm, which is a related technology to solve the problem of data balance, is a down-sampling technology based on integrated learning. Various types of samples are ta...

AI Technical Summary

Problems solved by technology

[0004] The inventor found through research that the EasyEnsemble algorithm, which is a related technology to solve the problem of data balance, is a down-sampling technology based on integrated learning. Various types of samples are taken as part of the training set, which makes the original few samples even fewer, resulting in potentially useful information. The lack of learning effect is poor, especially when the amount of data is small and the samples are extremely unbalanced. This method of integrated downsampling makes the sub-classifier contain too few training samples, which makes it impossible to learn the characteristics of the sample, resulting in poor performance of the classifier.

Images