Prediction of Hard Drive Failures for Data Center Based on LightGBM

Today, industrial-scale organizations increasingly rely on data centers to store and process data. Therefore, it is of great value to protect data security and reduce the cost of data center if operators can accurately predict the failure of the hard disks. In this paper, we use data collected by SMART (Self-Monitoring, Analysis and Reporting Technology), which contains 63 attributes as preliminary features to propose a model for hard disk fault prediction using LightGBM (Light Gradient Boosting Machine) detection algorithm. After analyzing the features, we further process them and conduct comparative training. The experimental results show that a reliable prediction model is finally constructed, which has high accuracy in hard disk error prediction.