Neuro Quantology | September 2022 | Volume 20 | Issue 9 | Page 4146-4154 | doi: 10.14704/nq.2022.20.9.

NQ44475
Dr. K. Kishore Anthuvan Sahayaraj, Dr. R. Chithra, Sankar R, Satishkumar J, Dr.Vilas Ramrao Joshi, Dr.R.Thiagarajan/ Virtual Storage Failure
Prediction model using supervised machine learning

Virtual Storage Failure Prediction model using supervised

machine learning 4146
Dr. K. KishoreAnthuvan Sahayaraj1,
Dr. R. Chithra, Sankar Professor2, R3, Satishkumar J 4,
Dr.Vilas Ramrao Joshi5, Dr.R.Thiagarajan6
1Assistant
Professor, Department of Computing Technologies,School of Computing,College of
Engineering and Technology, SRM Institute of Science and Technology, [email protected]
2Department of Information Technology, K S Rangasamy College of Technology, [email protected]
3Assiatant Professor, Dept of MCA, S.A.Engineering College ( Autonomous), [email protected]
4Assistant professor, Dept of Information Technology, Kongunadu College of Engineering and
Technology [email protected]
5Assistant professor, Computer Engineering Department, Navsahyadri Education society Group of
Institution Naigaon Pune, [email protected]
6Associate professor, Dept. of Information Technology, Prathyusha Engineering College,
[email protected]

Abstract
Hard disk failure occurs due to the malfunctioning in the operation of configured computers. Due to
some of the external factors such as natural calamities, electrical disturbance can cause impact over the
failure of a hard disk. Data corruption, distortion and disruption in the computer’s hard drive due to the
malware infection in the disk. If the hard disk fails, the PC can stop its function by deriving any noise.
Due to the hard disk failure, the system halts the process due to the triggered physical failure. As the
data centers are predominantly increasing, the hard disk space complexity plays a vital role. Continuous
monitoring of the computer can reduce the efficiency of the lack of control over the security aspects. In
this paper, the artificial intelligence with machine learning is used to predict the potential features in
the disk. Detection of the hard disk features by determining the attributes with the statistical tests of
compatibility. Different methods are used to analyze the test parameters using the svm, random forest
and naïve bayes classifier to analyze the accuracy over the approximation results. Prediction of failure
test can improve the reliability in the storage system. Different techniques are used to analyze the signs
of the failure. The warning signs are detected to identify the early detection of the failure in the hard
disk. Some of the viruses or malware infected files can destroy the data in the hard disk, which can
cause potential data losses. A hard drive crash can fail up the boot process where it gets corrupted.
Some of the corruptions are firmware, ransom ware corruption which causes unreadable which leads to
the data loss. Electronic failure or power surge can cause the system to halt process. The internal failure
such as corruption of files, overheating and external factors such as human errors can cause permanent
damage to the data loss medium. The AI is used to predict the failure over the hard drive model for
identifying the exact precision. An external hard drive has to be checked monitoring based upon the
failure statistical report. In some data centers, the HDD will overheat due to the high consumption of
energy and overheating due to the maximum range. So, to avoid such disruption, the failure has to be
analyzed by using AI with machine learning. The comparative results are analyzed to predict the HDD
failure using AI and MI methodologies.
DOI Number: 10.14704/nq.2022.20.9.NQ44475 Neuro Quantology 2022; 20(9):4146-4154
Neuro Quantology | September 2022 | Volume 20 | Issue 9 | Page 4146-4154 | doi: 10.14704/nq.2022.20.9.NQ44475
Dr. K. Kishore Anthuvan Sahayaraj, Dr. R. Chithra, Sankar R, Satishkumar J, Dr.Vilas Ramrao Joshi, Dr.R.Thiagarajan/ Virtual Storage Failure
Prediction model using supervised machine learning

Introduction machine learning approach is improvised. By

The real-world recognition problems are using the runtime and training up of the data, the
implemented due to the internal and external data gets correlated in a systematic way. To
failures in the computer system. In some of the analyze the specific constraints, the machine
external drives, the failure over the hard disk learning formulates the derivation of the failure
occurs due to the potential causes of losses which detection in the proactive aspects. Both the
occur in the significant data[1]. The data gets training and testing model are correlated to 4147
corrupted within the hard disk, which damages attain the functional way of predicting the data.
the data. These problems are characterized Statistical hypothesis is used to predict the hard
according to the state of events. Self-monitoring drives which fail during the progress. Benchmark
of the hard disk can be able to analyze the failure, approach is used to analyze the evaluation of the
and it is considered an early way of identifying different machine learning algorithms
the failure. These allow the user to backup the methodology. Storage systems stores up the data
data automatically by detecting the data. The efficiently where it emphasizes the data into
smart way of analyzing the failure over the hard incidence in the hard disk. It enhances the
disk is characterized using the signal by the use of performance by checking the stability and
a predictive flag. Some of the errors are read-type reliability of the hard disk. The trained samples
errors, track retrieving the data, pattern with are used to classify the data by classifying the
positive values by attaining the failure state. This approach where it utilizes the randomization of
warning or alarm can give the attributes data. Based upon the optimization of the trained
validation based on the impending-type failure. data and samples, the data are parameterized
Threshold range is used to approximately with the optimization method. These attributes
identify the exceeding of the predefined value. are advanced using the specialization of the
This contribution enhances the performance by model by distinguishing the drive characteristics.
using the machine learning with AI technique. The drive characteristics are analyzed using the
The hard drive computates, along with the different parameters which identify the good
interactive type model of the OS of the system. To drives and failure type drives. Data which are
attain the outline over the data values with imbalanced are improvised using the predictive
enhanced accuracy performance, the supervised analysis. These computates the data by detecting
and unsupervised learning techniques. To avoid the overhead data and enhances the efficiency of
the disruption of the data, the reliability of the the different failure aspects. Memory spaces and
data has to be maintained[2]. Some of the storage space play a vital role in the component
complex type systems can cause failure in the of the system. It enhances the data memory by
data while communicating with each other. categorizing the data according to the three
Traditional way of approach does not provide a different memory devices such as primary,
proper way to analyze the data privacy and tertiary and secondary devices respectively.
confidentiality. Some devices need less computational space to
Proactive failure in the hard disk occurs due to execute the instructions. Hard disk drives can
the improper way of handling the system. It store up the data in terms of the storage of the
indicates the unusual or abnormal type of data where it has both the data and information.
parameters in the system by analyzing the Hard disk drives are the secondary memory
behaviors such as by analyzing the system devices which are used to store up the memory. It
loading performance, network traffic mechanism, has rigid and high tolerance in performance such
and CPU utilization speed and memory structure as read/write type operations. These are reliable
[3]. This failure has to be analyzed so that the in cost and less expensive in nature. These
recovery actions can be taken as soon as possible. circular type disks which are assembled in the
In some extreme cases, the failure cannot be computer are composed of the magnetic material.
avoided due to the service outage in the system. A In the data center, the hard disk is used to store
machine learning technique is used to analyze the the data memory for the easy access purpose,
data labeling and prediction purpose. To enhance which enhances the reliability of the storage
the quality and performance over the data system. If the hard disk failure occurs, then the
labeling and classifying the predicted data, the memory storage in the data is also affected. These
Neuro Quantology | September 2022 | Volume 20 | Issue 9 | Page 4146-4154 | doi: 10.14704/nq.2022.20.9.NQ44475
Dr. K. Kishore Anthuvan Sahayaraj, Dr. R. Chithra, Sankar R, Satishkumar J, Dr.Vilas Ramrao Joshi, Dr.R.Thiagarajan/ Virtual Storage Failure
Prediction model using supervised machine learning

data are monitored using the electro-mechanical backup of the data is for a futuristic purpose. It
defective mechanism. Machine learning performs helps to support the drives by creating the
the high level of accuracy in predicting the disk interface between the computer medium. These
failures by attaining the futuristic model. external drives are portable and consist of more
Literature Review space to store the data. Hard disks generally can
store upto terabyte of data where it encloses the
A hard drive is used to store the data using the data in a sequential way. They are made up of 4148
electro-magnetic storage which can retrieve or circular plates along with the magnetic material.
gather the data for further processing. These In this approach, the data are stored in the
drives have the rigid type platters along with the surface medium using the concentric tracks.
magnetic material. The magnetic material is These magnetize the tiny spots by detecting the
paired along with the actuators with the platter read/write data using the spinning of the
to read and write the data. The data are direction. A hard drive uses read only memory
sequentially read/write using the sequential to give instruction to read/write the disks. These
functionality of electro-magnetic storage. These are stacked according to their storage of the data
data are stored in the random access manner periodically. These tracks periodically
where it has non-volatile storage. These are used emphasize according to the distance from the
to store the vast data to enhance the speed and centre. The main storage of the computer is the
performance. The file system is improved using hard disk drive to store up the data memory.
the error correction by following redundancy The hard Disk Drive stores up the data of the
and recovery[4]. In the modern hard drive disk, operating system and applications in the
the progress is checked using the following software within it. A hard disk drive is generally
block size. The block size is specified in the used in different applications to store the data
modern technique where it has the product's memory in vast speed and reduces the
manufacturer specification. These use low drive complexity. According to the positioning of the
type commands in improving the data. To spindle inside the chamber of the head disk
reduce the delay over the minimization delay, drive, can attain the read and write motors state
the defragmentation method is used. These and its direction. External hard drives can
reduce the delay in the access over speed. The expand the storage capacity using the device
hard disk storage is a type of non-volatile which is connected. These are generally slower
memory which is internally stored in the in data transfer when compared to the internal
computer or even data centers. Some of the hard hard drives. These uses the slower data transfer
disks use the motherboard, namely, SATA cable due to the slower transfer rates. The drive
or ATA cable. These use the connectivity to the capacity increases as it is used internally.
power supply using the PSU method. The data is
stored in the hard disk, where there are some of A hard disk drive consists of malware which can
the confidential files. These files are retrieved damage the computer code and software by
for later purpose. Computer tries to interact using the self-replicating virus. This self-
with the operating system to attain the data replicating virus has the ability to replicate itself
which are stored in the memory. It analyzes the by corrupting the files inside the memory of the
data storage by checking the storage medium computer. The smart way of identifying the hard
where the OS has been installed. These drives disk malfunction using the artificial intelligence
also require certain installation of software to can reduce the complexity in the data. This virus
download the files inside the computer. A hard can infect the system software and application
disk can store the data permanently inside the by acknowledging as a user and infects the
storage medium. Data which are used to computer without the user knowledge [4]. The
read/write are interpreted using the disk ability to replicate itself is one of the major
controller, the disk controller is used to instruct characteristics. Malware on the hard disk can
the actuators. These hard drives are of two damage or corrupt the data in the hard disk.
different types, internal drives and external These can steal the data confidentiality of the
drives. Mostly, all the hard disk drives are private information which can infect the files.
internal, but some of them are stand-alone Boot sector virus is one of the cyber attacks
devices such as external hard drives. This which can damage the system’s confidentiality
Neuro Quantology | September 2022 | Volume 20 | Issue 9 | Page 4146-4154 | doi: 10.14704/nq.2022.20.9.NQ44475
Dr. K. Kishore Anthuvan Sahayaraj, Dr. R. Chithra, Sankar R, Satishkumar J, Dr.Vilas Ramrao Joshi, Dr.R.Thiagarajan/ Virtual Storage Failure
Prediction model using supervised machine learning

and privacy measures. These trigger the victim's using the external drives, these boot sector
action by creating the malicious activity. These viruses can occur which can potentially damage
can inject the data by infecting the software the computer system.
programs and applications. Some of the boot
virus injects the victim where it exists within the
program. Virus can replicate and infect the
victim's system and its application. This 4149
malware can take over the control of the entire
host system. The hard drive malware can infect
the system memory by reducing the payload and
increasing the consumption of traffic mechanism
[5]. These types of different viruses can easily
stop the process by temporarily disabling the
hard drive. Boot sector virus is one of the types
of virus which can damage the hard drive by
infecting the files and their information. This
stored virus infects the operating system by
controlling the loading information. These
viruses can reformat the disk by cleaning up or
Figure 1: Hard Disk Drive
removing the data from the file without
acknowledging it. Traditional way of approach
does not provide a proper way to analyze the
data privacy and confidentiality [6-10]. Mainly
within the HDD. Using the self-monitoring,
Proposed System analyzing and report method, the HDD can be
monitored. It checks and monitors various
A hard disk can cause the physical failure due to attributes in the hard disk drive. The smart way
the physical damage where the stored data gets of using the Artificial Intelligence can check the
repaired and cannot be recovered properly. drive reliability and determines the error rate. It
Some of the different failures are different as checks the different attributes and identifies the
logical failure occurs where the data can be error rate. Using the supervised learning
recovered by using the software. By detecting approach, different attributes are compared to
the noise and unwanted bad performance, the analyze the accuracy in performance over the
logical failure can be analyzed. The hard results. These trains up the dataset and features
drivefailure occurs due to the malfunctioning in are extracted. The hard Disk Drive stores up the
the drive data or corrupted data. A data center data of the operating system and applications in
stores up the memory for the processing of the the software within it. The extracted results are
storage of data. These data are stored in the hard progressed where the data gets preprocessed,
disk drive. System peripherals are connected to trained and validated to attain the exact results.
the hard drive to store the vast amount of data. The resulted data are classified according to
The smart AI technology validates the data their classified approach in the trained data.
Neuro Quantology | September 2022 | Volume 20 | Issue 9 | Page 4146-4154 | doi: 10.14704/nq.2022.20.9.NQ44475
Dr. K. Kishore Anthuvan Sahayaraj, Dr. R. Chithra, Sankar R, Satishkumar J, Dr.Vilas Ramrao Joshi, Dr.R.Thiagarajan/ Virtual Storage Failure
Prediction model using supervised machine learning

4150

Figure 2: Systematic Representation of Analyzing Malfunction in HDD

Data Preprocessing along with the viruses. Using the machine
Data which are generally taken are inconsistent learning algorithms along with the AI, it detects
and incomplete. These raw data has to be the malfunction by recognizing the attributes
processed for resolving the data issues into an and predictive way to detect the comparison
understandable format. Data gets transformed over the dataset.
where it is encoded into an understandable
format. Parsing the data into a meaningful Data Training
format easily features up the data. Data which The training up of the dataset can predict the
are interpreted are in an encoded form. To avoid model data more precisely and accurately. It
some of the using attribute, values and predicts test data to normalize the performance,
unfiltered data in the raw data, it has to be efficiency to train the model. Initial set of data
avoided. Due to the presence of unfiltered which trains up the program to understand and
attribute, the quality of the performance over get sophisticated results. Based upon the
the data gets reduced. In the data processing, the parameters categorization, the models are
statistical type of parameters is trained. Data characteristically fit in type and range to get
which are extracted are featured, normalized, access to the performance of the model. Initially,
and transformed into a meaningful format. these applications such as ML and AI need
Organize each data and create valuable initialization to acquire the data to accurately
information by checking the failed attributes label the data. The labeling of the data can
where null is used to fill up those missing values. accurately increase the reliability of the
A smart way of reporting the data checks the information which is processed for the further
data monitoring where it automatically detects processing usage. The training up of the data is
the malfunctioning in the hard disk. By using the by the iterative approach which eventually
artificial intelligence, the smart way to detect the categorizes according to the recognition of the
malfunction in HDD reports various attributes features. This training set composes eventually
Neuro Quantology | September 2022 | Volume 20 | Issue 9 | Page 4146-4154 | doi: 10.14704/nq.2022.20.9.NQ44475
Dr. K. Kishore Anthuvan Sahayaraj, Dr. R. Chithra, Sankar R, Satishkumar J, Dr.Vilas Ramrao Joshi, Dr.R.Thiagarajan/ Virtual Storage Failure
Prediction model using supervised machine learning

according to their futuristic type. The correlated data which have generally occurred are split into
dataset are classified into their characteristic small types of fractions.
features. After the training up of the data, the
functional testing is carried out to verify the Supervised learning approach
data. The set of input validates the input to In the machine learning algorithms, the
attain the result outcomes. This testing model supervised approach is used to predict the
evaluates the performance metrics over the model where it trains up the data and values to 4151
observation results. It generalizes the deliver the necessary results. In this predictive
performance of the task within the new data. approach, the model analytics used there is the
The testing model parameterizes the complex predictive modeling. This predictive modeling
type model which memorizes the data. It which is used is to build the model for deriving
predicts the values along with the response the predictions. This trains up the data, using the
variables. It checks over fitting variables such as certain properties along with the dataset to get
memorizing that trained set. The trained set overall results. The support vector machine
identifies the prediction error which determines method is one of the types of classification
the percentage error and high variance error. methodology where it samples the hyper planes
Unlabeled data are typically augmented where into two or multiple streams of classes. These
the labeled data are tagged according to the consist of the maximum distance between the
sampling of data. Labeled data are significantly points of the hyper plane. Decision tree
taken from the raw data of unlabeled data. classification approach is a tree-type approach
Identifying the raw data into a meaningful range which uses the conditions to represent the data
of labels, this provides the descriptive content. according to their particular set of events. A
Labeling combines two or more samples of class label classifies the dataset according to
labels into one tagged one. Test data produces their set of features. The random Forest
the systematic way of generating the data. A data classifier is used to predict the multiple decision
generation tool systematically generates and trees which eventually distribute random
symbolizes the data. The data are tested by variables. This predicted tool uses the different
determining the considerations based on the matrix evaluation to analyze the number of
quality and cost. Using the equivalence class samples where it lists out the predicted and
model, the data are validated towards the actual classes respectively. The accuracy defines
particular set of reports. Reports are the total classifications according to the number
sequentially derived based upon the tested of samples. It is also used in the specificity and
results. precision where both are calculated differently.
In this approach, the accuracy is calculated using
Data Validation the following equations. The TP, TN, FP and FN
To increase the accuracy and performance over can be represented as follows,
the source of the data, the data are validated for 𝑇𝑃
further process. Depends upon the constraints True Positive Rate (TP) =𝑇𝑃+𝐹𝑁
such as the objectives, the data validation takes False Positive Rate (FP) =𝑇𝑁+𝐹𝑃
𝐹𝑃
place. These data are validated to attain the data
cleansing process which guarantees the fitness Precision =𝑇𝑃+𝐹𝑃
𝑇𝑃
and develops the consistency of the data
application. These are implemented using The above equations are calculated to derive the
different methodologies where it progress with validation over the performance using the test
the data integrity methods. Data validation sets. These are over-fitted which are validated to
checks the error and misspells the error in the avoid any type of potential boas. These have the
data. It checks and integrates the error in the normalized values where it uses the cross
data and determines the accuracy of the system. validation to validate the performance.
In terms of cross-validation model, the Sensitivity defines the probability in classifying
performances are based upon the statistical the problem, i.e., True positive value. Specificity
prediction type. Based upon the iteration, the specifies the probability of predicting the false
cases, i.e., true negative value respectively.
Neuro Quantology | September 2022 | Volume 20 | Issue 9 | Page 4146-4154 | doi: 10.14704/nq.2022.20.9.NQ44475
Dr. K. Kishore Anthuvan Sahayaraj, Dr. R. Chithra, Sankar R, Satishkumar J, Dr.Vilas Ramrao Joshi, Dr.R.Thiagarajan/ Virtual Storage Failure
Prediction model using supervised machine learning

Recognition of Failure Drive Failure- State of the HDD drive fails or

Drive Size- Indicates the size of the HDD Total halts condition
Size Failure Rate- Percentage of the failure condition
Drive Days Count- Total amount of time, the Defective Level- Low/High/Medium State
HDD is used Unsupported File- Yes/No
Drive Count-Ranges of Drive
4152

Drive Size Drive Days Drive Drive Failure Defective Unsupported

Count Count Failure Rate Level File

4 Tb 80,689 3,953 9 0.28% Medium No

4 Tb 60,638 1,500 0 0% - -
8 Tb 1,60,009 9,575 16 0.60% High Yes
8 Tb 1,36,313 13,363 13 0.30% Medium No
4 Tb 1,73,954 19,572 12 0.12% Medium No
6 Tb 80,657 896 14 0.35% Medium No
10 Tb 1,75,388 1400 82 0.90% High Yes
12 Tb 3,19,001 15,655 26 0.76% High Yes
14 Tb 6,65,477 8,343 20 1.15% High Yes
Table 1: Various attributes for the recognition of failure in HDD from the dataset example

Different Algorithm Accuracy Precision Recall F-measure

Support Vector Machine 0.868 0.985 0.006 0.015
Decision Tree Algorithm 0.676 0.787 0.665 0.756
Random Forest (Multiple 0.987 0.961 0.946 0.967
Decision Trees)
Table 2: Cross-Validation Ranges of Different Algorithm

1.2
0.987 0.985 0.961
1 0.946
0.967
0.868
0.8 0.787 0.756
0.676 0.665
0.6

0.4

0.2

0 0.006 0.015
Accuracy Precision Recall F-measure

Accuracy, Precision, Recall and F-measure

Neuro Quantology | September 2022 | Volume 20 | Issue 9 | Page 4146-4154 | doi: 10.14704/nq.2022.20.9.NQ44475
Dr. K. Kishore Anthuvan Sahayaraj, Dr. R. Chithra, Sankar R, Satishkumar J, Dr.Vilas Ramrao Joshi, Dr.R.Thiagarajan/ Virtual Storage Failure
Prediction model using supervised machine learning

The above graph illustrates that the cross- The AI is used to predict the failure over the
validation using the different algorithms to hard drive model for identifying the exact
identify the hard disk drive is good or bad. Based precision. Proactive failure in the hard disk
upon the criteria, the HDD indicates whether it is occurs due to the improper way of handling of
a good drive or a bad drive. From the drive the system. To attain the outline over the data
deployment, the user can get whether it is time values with enhanced accuracy performance, the
to replace the drive. After the comparative supervised and unsupervised learning 4153
analysis of the drive, the drive can be replaced as techniques are been followed. Hard disk drive
it prior to indicate the condition state of the failure reduces the speed, increases the
drive’s condition. Using the smart monitoring complexity of the runtime and threshold range.
methodology, such as self-monitoring, analyzing Self-monitoring of the hard disk can able to
and reporting the technology in the hard disk analyze the failure and it is considered as the
drive, the following steps algorithms are carried early way of identifying the failure.
out. Use the machine learning algorithm, where
threshold range D indicates the ratio of the total
samples and multiple decision trees are followed References
- Trees. Wh indicates the size where WH. (1) [1] Chhetri, T. R., Kurteva, A., Adigun, J. G., &
Initiate the sample using the sampling algorithm. Fensel, A. (2022). Knowledge Graph Based
(2) After sampling process, use each tree T. (3) Hard Drive Failure Prediction. Sensors,
calculate the classification steps using the tree 22(3), 985.
classification and sample S (4) Check whether or
[2] Yu, J. (2019, November). Hard disk drive
not if any abnormal samples are comparatively
failure prediction challenges in machine
greater. (5) Check whether the HDD is good or
learning for multi-variate time series. In
fails for a defective piece. (6) The predicted
Proceedings of the 2019 3rd International
results are generated. (7) End the process. Self-
Conference on Advances in Image
monitoring of the hard disk can be able to
Processing (pp. 144-148).
analyze the failure, and it is considered an early
way of identifying the failure. These allow the [3] Mudiyanselage, A. R. (2020). Data
user to backup the data automatically by Engineering and Failure Prediction for Hard
detecting the data. Using the self-monitoring, Drive SMART Data. Bowling Green State
analyzing and report method, the HDD can be University.
monitored. It checks and monitors various [4] Pinciroli, R., Yang, L., Alter, J., & Smirni, E.
attributes in the hard disk drive. The smart way (2020). The life and death of SSDs and
of using the Artificial Intelligence can check the HDDs: Similarities, differences, and
drive reliability and determines the error rate. It prediction models. arXiv preprint
checks the different attributes and identifies the arXiv:2012.12373.
error rate.
[5] Li, W. (2020). Hard Disk Drive Failure
Detection with Recurrence Quantification
Conclusion Analysis (Doctoral dissertation,
Northeastern University).
The Prediction of failure test can improve the
reliability in the storage system. By using the [6] Li, J., Stones, R. J., Wang, G., Liu, X., Li, Z., & Xu,
artificial intelligence, the smart way to detect the M. (2017). Hard drive failure prediction
malfunction in HDD reports various attributes using decision trees. Reliability Engineering
along with the viruses. Detection of the hard disk & System Safety, 164, 55-65.
features is by determining the attributes with [7] Yang, W., Hu, D., Liu, Y., Wang, S., & Jiang, T.
the statistical tests of compatibility. Continuous (2015, September). Hard drive failure
monitoring of the computer can reduce the prediction using big data. In 2015 IEEE 34th
efficiency of the lack over the security aspects Symposium on Reliable Distributed Systems
and time complexity. In the proposed work, the Workshop (SRDSW) (pp. 13-18). IEEE.
artificial intelligence with machine learning is
used to predict the potential features in the disk. [8] Yang, Q., Jia, X., Li, X., Feng, J., Li, W., & Lee, J.
(2020). Evaluating feature selection and
Neuro Quantology | September 2022 | Volume 20 | Issue 9 | Page 4146-4154 | doi: 10.14704/nq.2022.20.9.NQ44475
Dr. K. Kishore Anthuvan Sahayaraj, Dr. R. Chithra, Sankar R, Satishkumar J, Dr.Vilas Ramrao Joshi, Dr.R.Thiagarajan/ Virtual Storage Failure
Prediction model using supervised machine learning

anomaly detection methods of hard drive [10] Chaves, I. C., De Paula, M. R. P., Leite, L. G.,
failure prediction. IEEE Transactions on Gomes, J. P. P., & Machado, J. C. (2018, July).
Reliability, 70(2), 749-760. Hard disk drive failure prediction method
[9] Huang, X. (2017). Hard drive failure based on a Bayesian network. In 2018
prediction for large scale storage system International Joint Conference on Neural
(Doctoral dissertation, UCLA). Networks (IJCNN) (pp. 1-7). IEEE.
4154