To Fail or Not to Fail: Predicting Hard Disk Drive Failure Time Windows

Züfle, Marwin; Krupitzer, Christian; Erhard, Florian; Grohmann, Johannes; Kounev, Samuel

doi:10.1007/978-3-030-43024-5_2

Cited by 10 publications

(7 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The solution in [ 40 ] has the ability to make real-time predictions by utilising technologies such as Apache Hadoop ( (accessed on 2 June 2021)) and Apache Spark ( (accessed on 6 September 2021)). The use of a RF classification algorithm for HD failure prediction can also be seen in [ 41 , 42 ]. However, in addition to using the same ML algorithm, Shen et al [ 41 ] use a sliding window to reduce the effect of noise and an additional part voting strategy to improve the prediction accuracy.…”

Section: State Of the Artmentioning

confidence: 99%

“…However, in addition to using the same ML algorithm, Shen et al [ 41 ] use a sliding window to reduce the effect of noise and an additional part voting strategy to improve the prediction accuracy. On the other hand, Züfle et al [ 42 ] use additional techniques such as synthetic minority oversampling technique (SMOTE) [ 43 ] and enhanced structure-preserving oversampling (ESPO) [ 44 ] together with random forest. Similar to Su et al, Züfle et al and Shen et al, Mashhadi et al [ 45 ] also make use of RF algorithm.…”

Section: State Of the Artmentioning

confidence: 99%

“…In addition to comparing the accuracy, precision, and recall of our results, we also compared the training time of our solution to that of related work (see Table 7 ). As can be seen in Table 7 , half of the studies we considered for comparison, with the exception of Züfle et al [ 42 ], Han et al [ 47 ], and Mamoutova et al [ 24 ], did not report training time. Our experiment’s training time is significantly longer than Züfle et al [ 42 ] and Han et al’s [ 47 ] work.…”

Section: Performance Evaluationmentioning

confidence: 99%

“…As can be seen in Table 7 , half of the studies we considered for comparison, with the exception of Züfle et al [ 42 ], Han et al [ 47 ], and Mamoutova et al [ 24 ], did not report training time. Our experiment’s training time is significantly longer than Züfle et al [ 42 ] and Han et al’s [ 47 ] work. One of the reasons for the longer time, in our case, is the sample size.…”

Section: Performance Evaluationmentioning

confidence: 99%

“…One of the reasons for the longer time, in our case, is the sample size. We used a sample size of 13,553,809, while Züfle et al [ 42 ] used 68,411. The other reason is that the H2O AutoML approach performs extensive hyperparameter optimisation of multiple algorithms, whereas the KG-based approach requires learning relationships.…”

Section: Performance Evaluationmentioning

confidence: 99%

See 4 more Smart Citations

Knowledge Graph Based Hard Drive Failure Prediction

Chhetri

Kurteva

Adigun

et al. 2022

Sensors

View full text Add to dashboard Cite

The hard drive is one of the important components of a computing system, and its failure can lead to both system failure and data loss. Therefore, the reliability of a hard drive is very important. Realising this importance, a number of studies have been conducted and many are still ongoing to improve hard drive failure prediction. Most of those studies rely solely on machine learning, and a few others on semantic technology. The studies based on machine learning, despite promising results, lack context-awareness such as how failures are related or what other factors, such as humidity, influence the failure of hard drives. Semantic technology, on the other hand, by means of ontologies and knowledge graphs (KGs), is able to provide the context-awareness that machine learning-based studies lack. However, the studies based on semantic technology lack the advantages of machine learning, such as the ability to learn a pattern and make predictions based on learned patterns. Therefore, in this paper, leveraging the benefits of both machine learning (ML) and semantic technology, we present our study, knowledge graph-based hard drive failure prediction. The experimental results demonstrate that our proposed method achieves higher accuracy in comparison to the current state of the art.

show abstract