Anomaly Detection Model for Predicting Hard Disk Drive Failures

The electromechanical design of the HDD (Hard Disk Drive) renders it more susceptible to failures than other components computer system. failure leads permanent data loss, which is typically expensive itself. SMART (Self-Monitoring, Analysis and Reporting Technology) system warns user if any parameter has exceeded predefined threshold value needed for safe operation. Machine learning methods take advantage dependence between multiple parameters in order make prediction precise. In this paper, we present a model based on anomaly detection method involving an adjustable decision boundary. are ranked by importance 13 most significant ones used as initial feature set our model. following stage, optimized removing those that have no major contribution model, forming final comprising seven features only. proposed achieved 96.11% rate average, with 0% false ten random tests. predicted 80% 24 hours before their actual occurrence, enables timely backup.