Article
Automation & Control Systems
Yu Wang, Long He, Shan Jiang, Tommy W. S. Chow
Summary: An adaptive error tracking method is proposed for HDD failure prediction, which is validated through model estimation and accelerated degradation testing, demonstrating better performance in failure prediction and alarm distance compared to previous methods.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2021)
Article
Chemistry, Multidisciplinary
Mingyu Zhang, Wenqiang Ge, Ruichun Tang, Peishun Liu
Summary: This paper proposes a novel failure prediction method that combines machine learning algorithms and neural networks. Through experiments on the publicly available BackBlaze hard disk datasets, it demonstrates the superiority of this method in hard disk failure prediction and solves the problem of low robustness and generalization in traditional machine learning methods.
APPLIED SCIENCES-BASEL
(2023)
Article
Engineering, Electrical & Electronic
Guochao Wang, Yu Wang, Xiaojie Sun
Summary: A method based on LSTM network and attention mechanism for multi-instance long-term data classification is proposed to predict HDD failures, achieving better results than other methods through analysis of HDD data from a communication company and Backblaze data center.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Chemistry, Multidisciplinary
Federico Gargiulo, Dirk Duellmann, Pasquale Arpaia, Rosario Schiano Lo Moriello
Summary: The paper examines methods for automated proactive disk replacement in large-scale computer centers, as well as the application of supervised machine learning to predict disk failures. It presents a strategy for optimizing hyperparameters of the machine learning classifier while also automatically labeling disk status (healthy/at-risk) during training and validation stages. The approach is tested against 65,000 hard drives in the CERN computer center and the results are discussed.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Hardware & Architecture
Fernando D. S. Lima, Francisco Lucas F. Pereira, Iago C. Chaves, Javam C. Machado, Joao Paulo P. Gomes
Summary: Predicting failures in Hard Disk Drives (HDD) is a major challenge faced by industry and academia. Health degree prediction is a popular strategy, but practical details have been neglected in previous works. A framework based on Deep Recurrent Neural Networks (DRNN) was proposed, taking into account the ordinal nature of the problem and different costs associated with mis-classifications.
IEEE TRANSACTIONS ON COMPUTERS
(2021)
Article
Food Science & Technology
Huili Zhu, Minyan Wang, Jing Zhang, Fengwang Ma
Summary: This study utilized the Random Forest algorithm to construct a prediction model of aroma components in apple hybrid offspring, with different preprocessing methods tested, showing that SNV is the most effective in noise removal. The characteristic wavelength-aroma chemical group model can accurately predict the aroma components in apples.
Article
Genetics & Heredity
Jin Li, Wenjie Liu, Luolong Cao, Haoran Luo, Siwen Xu, Peihua Bao, Xianglian Meng, Hong Liang, Shiaofen Fang
Summary: This study proposes a novel method based on random forest for identifying important features by jointly analyzing multimodal data. Significant subregions and genes were identified, providing new candidate genes for AD and demonstrating the contribution of hippocampal subregions and genes to AD.
Article
Engineering, Electrical & Electronic
Junjie Shi, Jing Du, Yingwen Ren, Boyu Li, Jinwei Zou, Anyi Zhang
Summary: This paper proposes an improved algorithm for early warning of mechanical hard disk failures, using feature selection, Generative Adversarial Networks (GAN), and Convolution-LSTM (C-LSTM) to enhance the data availability of the IaaS cloud platform.
JOURNAL OF SENSORS
(2022)
Article
Chemistry, Physical
Botao Li, Yoshihiko Nishikawa, Philipp Hollmer, Louis Carillo, A. C. Maggs, Werner Krauth
Summary: This paper discusses pressure computations for the hard-disk model since 1953 and compares them with the results obtained from advanced Monte Carlo and Metropolis algorithms. The difficulty of estimating pressure in the hard-disk phase-transition scenario has not been fully recognized in the long-standing controversy. The authors present the physics of the hard-disk model, the definition of pressure, unbiased estimators, and different sampling algorithms and criteria for bounding mixing times.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Materials Science, Composites
Sheng Sang, Chen Xu, Ziping Wang, Conner Side, Brent Fowler, Jiadi Fan, Daniel Miao
Summary: In this article, a methodology using machine learning and propagation of elastic waves was proposed to accurately determine the topology of binary composite plates. The study indicated that elastic waves propagated through composites can efficiently collect microstructure information, and multiple RF models can accurately predict composite configurations by learning from plate topologies and their corresponding output waves.
COMPOSITES COMMUNICATIONS
(2023)
Article
Green & Sustainable Science & Technology
Hamdi Ozaktan, Necati Cetin, Sati Uzun, Oguzhan Uzun, Cemalettin Yasar Ciftci
Summary: This study used image processing techniques to determine the physical attributes of 20 different bean genotypes and analyzed their color characteristics. Four machine learning algorithms were employed to predict seed mass. The results showed that machine learning techniques improved the efficiency of related machinery and saved time and labor.
ENVIRONMENT DEVELOPMENT AND SUSTAINABILITY
(2023)
Article
Chemistry, Multidisciplinary
Han Wang, Qingfeng Zhuge, Edwin Hsing-Mean Sha, Rui Xu, Yuhong Song
Summary: To improve the accuracy and efficiency of hard disk failure prediction, a two-layer classification-based feature selection scheme is proposed in this paper. The proposed technique improves the prediction accuracy of machine learning and artificial intelligence models, with random forest and long short-term memory showing the best results. Furthermore, this scheme significantly reduces training and prediction latency compared with baseline methods.
APPLIED SCIENCES-BASEL
(2023)
Article
Chemistry, Medicinal
Robert P. Sheridan
Summary: Similar predictions are observed for the majority of molecules across different versions of the random forest models for ADMET end points. However, a small minority of molecules show substantial shifts in predictions over a few versions. These shifting molecules tend to have more accurate predictions in later versions. This Perspective investigates metrics to identify and predict substantial shifts in molecule predictions.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2022)
Article
Materials Science, Multidisciplinary
Bor-Yuan Jiang, Kunliang Zhang, Takahiko Machita, Wenyu Chen, Moris Dovek
Summary: Hard Disk Drive read heads switched from GMR to TMR technology to improve signal-to-noise ratio for advanced linear densities. The use of crystalline MgO barriers matched to BCC magnetic layers increased the MR ratios. Enhanced side shielding and reduced shield spacing further improved the track and linear density capabilities of TMR heads. Introduction of two-dimensional recording heads boosted head SNR, but overcoming spin torque noise remains a challenge. Magnetic thermal noise can only be overcome by increasing the moment of the free layer, potentially setting the final density achievable by TMR heads.
JOURNAL OF MAGNETISM AND MAGNETIC MATERIALS
(2023)
Article
Computer Science, Information Systems
Changhao Wang, Feng Tian, Yan Pan
Summary: This paper explores the use of machine learning and data mining techniques to predict and reduce urban crime rates through the implementation of a random forest algorithm and a multi-drone patrol response strategy.