Imbalanced data industrial fault classification method based on k-means

Abstract

The invention discloses an imbalanced data industrial fault classification method based on k-means. The method comprises the following steps: first, utilizing the k-means; based on the imbalance degrees, clustering the classes with relatively big data; dividing the majority classes into N sub-classes; combining the N sub-classes with M minority classes to serve as a multi-classification problem for an (M+N) classification; and finally, performing learning according to a naive bayes classifier. Compared with other existing methods in prior art, the method of the invention keep the information of the original data to the largest extent and better resolves the problem with imbalanced class data classification under the condition that over-fitting is prevented. Therefore, compared with other methods, the classification precision is increased, and the phenomenon of over-fitting can be reduced.

Description

technical field [0001] The invention belongs to the field of industrial process control, in particular to an industrial process fault classification method for unbalanced data. Background technique [0002] In the work of industrial fault classification, some commonly used classification methods have a prerequisite for use, that is, the amount of data in the training set is equivalent. But the reality is often not the case. When a certain type of data is very large, or a certain type of data is rare, that is, unbalanced data appears, directly using the traditional classification method will produce a large classification error. [0003] In recent years, the research on unbalanced data has been a hot spot. The existing methods are mainly solved from two directions, one is from the algorithm level, and the other is from the sampling level. The present invention mainly improves the traditional classification method at the sampling level. . The improvement methods for sampling...

AI Technical Summary

Problems solved by technology

The improvement methods for sampling are mainly divided into two categories. One is oversampling, that is, resampling the minority class to achieve data balance. A major drawback of this method is that it will increase the system method and cause overfitting. The application effect is not very ideal; the other type is under-sampling, that is, according to certain rules, a part of the majority class is selected as training data, and other data are discarded to achieve data balance. This method ignores part of the Most class data information will lead to insufficient accuracy of the trained classifier

Images