Disk failure predictions are essential in contemporary storage solutions for ensuring data integrity. Machine learning was shown to be an effective strategy for addressing the issue of disc failure forecasting. To better anticipate when a disc may fail, researchers have recently begun to include the idea of time windows in the field. By analyzing the correlation between the value of a disc and its age, these studies have been able to make some good predictions. Storage devices are among the most regularly replaced hardware devices and continue to provide difficulties in predicting failure. There have been several promising suggestions for developing a hard drive forecasting model based on SMART features, including the use of statistical and machine learning techniques. Although these models are used in production data centers, they were not evaluated under production conditions. In addition, hard drives deteriorate with time, although this slow decrease is not fully explained by current theories. To address the shortcomings of the currently available technology, this study introduces a novel hard drive failure prediction model that is based on Ensemble Learning. This model outperforms its predecessor, which used a Back propagation artificial neural network, in terms of prediction accuracy, model stability, and explanatory lucidity. With an overall forecast of 15 days, our proposed ensemble learning-based model is 96% accurate in predicting disc failures.

Ensemble Learning, Back blaze predictions, Hard-drive failure, Machine learning, Data science.