Model training method, disk prediction method and device and electronic equipment

Abstract

The invention relates to a model training method, a disk prediction method and device and electronic equipment, and belongs to the field of computers. The method comprises the steps that negative sample data representing disk faults and positive sample data representing disk normality are acquired; performing N times of iterative training on the initial model based on the negative sample data andthe positive sample data to obtain a prediction model capable of predicting whether the disk fails or not; and in the ith iterative training process, performing down-sampling on the positive sample byusing a loss value generated by the model obtained by the i-1th iterative training on the positive sample data, and training the model obtained by the i-1th iterative training by using the positive sample data and the negative sample data obtained by sampling. According to the method, when down-sampling is carried out on the positive sample, the positive sample data sampled each time is differentdue to the fact that the loss values generated by the model obtained through iterative training each time on the positive sample data are different, more attention can be paid to difficult samples based on classification difficulty during sampling, so that the precision of the model is improved.

Description

technical field [0001] The present application belongs to the field of computers, and in particular relates to a model training method, a disk prediction method, a device and electronic equipment. Background technique [0002] In recent years, with the development of emerging technologies such as cloud storage, mass data storage technology is developing faster and faster. As the final storage place of data, the disk is one of the most important network devices, and it is also the device that fails most often. The disk failure prediction method based on machine learning has achieved satisfactory prediction results with the help of the powerful learning ability of machine learning algorithms. [0003] At present, the model training in the disk failure prediction method: mostly by marking the disk Self-Monitoring Analysis and Reporting Technology (SMART) log data as normal samples and faulty disk samples, and according to the attribute value of the sample Divide the samples i...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the purpose of this application is to provide a model training method, disk prediction method, device and electronic equipment to improve the prediction accuracy of the existing disk prediction model not high problem

Images