Training method of machine learning model, exception prediction method and related devices

Abstract

The embodiment of the invention discloses a training method of a machine learning model, an exception prediction method and a related device. The method includes: processing The training data of the T-M time to the T time by the trained first machine learning model to obtain the predicted data of the T+N time;; Determining credible abnormity according to the prediction data at the T + N moment andthe real data at the T + N moment, and processing The training data of the T-M time to the T time by using the second machine learning model to obtain a prediction abnormality,; and according to theerror of the prediction abnormity degree and the credible abnormity degree, transmitting the error back to the second machine learning model, adjusting the current model parameters of the second machine learning model, and achieving unsupervised learning of the second machine learning model capable of predicting the data abnormity at the future moment. Moreover, the relationship between continuousdata is considered in training of the training data based on the time period, and the accuracy of model prediction can be improved.

Description

technical field [0001] The present application relates to the technical field of machine learning, and in particular to a training method of a machine learning model, an abnormality prediction method and related devices. Background technique [0002] In the production environment, the normal state data of the device is widespread, and the abnormal state data is less. For machine learning algorithms, the higher the proportion of similar data distribution, the easier it is to learn. The higher the sample distribution ratio, the easier it is to learn, and the result of easy learning is that the learning error is smaller, and vice versa. [0003] At present, in the field of fault prediction, the fault mode and normal mode are mainly learned through a supervised method. This method has the following two problems: First, a large amount of fault label data is required, and the label data is difficult to obtain in actual production. ; Second, there are many kinds of faults, it is ...

AI Technical Summary

Problems solved by technology

[0003] At present, in the field of fault prediction, the fault mode and normal mode are mainly learned through a supervised method. This method has the following two problems: First, a large amount of fault label data is required, and the label data is difficult to obtain in actual production. ;Second, there are many kinds of faults, it is difficult to have a certain class pattern, and it is difficult to learn

Images