Disk failure prediction via Lightgbm model

Sitao Ji; Xingquan Zuo; Hai Huang

doi:10.1117/12.2652965

10 November 2022 Disk failure prediction via Lightgbm model

Sitao Ji, Xingquan Zuo, Hai Huang

Proceedings Volume 12331, International Conference on Mechanisms and Robotics (ICMAR 2022); 123313G (2022) https://doi.org/10.1117/12.2652965
Event: International Conference on Mechanisms and Robotics (ICMAR 2022), 2022, Zhuhai, China

Abstract

Disk failure prediction has become a major concern with the rapid expansion of data centers and the demand for reliable storage systems. Based on extremely unbalanced positive and negative samples in the disk data set, this paper proposes a fast and accurate disk failure prediction model. Target disks are selected from the actual disk data set, and the correlation matrix is used to filter the features with high correlation coefficients, which makes the model easy to fit. Aiming at the imbalance of positive and negative samples, an under-sampling strategy is implemented through Mini Batch K-Means clustering, and the Focal Loss function is introduced into the LightGBM model to reduce the impact of the imbalance data. Compared with several traditional machine learning models, the effectiveness of the proposed prediction model is verified on the public real data set.

Citation Download Citation

Sitao Ji, Xingquan Zuo, and Hai Huang "Disk failure prediction via Lightgbm model", Proc. SPIE 12331, International Conference on Mechanisms and Robotics (ICMAR 2022), 123313G (10 November 2022); https://doi.org/10.1117/12.2652965

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available