Tracking the Best Machine Learning Strategy for Hard Drive Failure Prediction - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Tracking the Best Machine Learning Strategy for Hard Drive Failure Prediction

Nour El Houda Benlaribi
  • Fonction : Auteur
Michel Batteux

Résumé

In order to reduce the consequences of hard drives failures which can lead to data losses, several failure prediction models have been proposed in the literature. In this paper, we study the effect of integrating methods designed for time series to a classification model. We consider Exponential Smoothing and ARMA (Autoregressive Moving Average) as time series methods, and Random Forest as a classification model. We use SMART (Self-Monitoring, Analysis and Reporting Technology) attributes as parameters for the classification model. Our study relies on operational data published by the Backblaze company for the year 2014. This dataset is collected from over 47000 hard drives with 81 models from 5 manufacturers. To analyze and compare the performances of the selected techniques, we compute two performance measures, the precision of the predictions and the recall.
Fichier non déposé

Dates et versions

hal-03784402 , version 1 (23-09-2022)

Identifiants

  • HAL Id : hal-03784402 , version 1

Citer

Nour El Houda Benlaribi, Leila Kloul, Michel Batteux. Tracking the Best Machine Learning Strategy for Hard Drive Failure Prediction. 32nd European Safety and Reliability Conference (ESREL 2022), Aug 2022, Dublin, Ireland. ⟨hal-03784402⟩
42 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More