2023 International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE) | 978-1-6654-9260-7/23/$31.

00 ©2023 IEEE | DOI: 10.1109/IITCEE57236.2023.10091031

ROCK / MINE CLASSIFICATION USING

SUPERVISED MACHINE LEARNING
ALGORITHMS
P Sathish Kumar
Sivachandra.K Kumudham.R
Department of Computer and
Department of ECE Department of ECE
Communication Engineering,
Vels Institute of Science, Vels Institute of Science,
Rajalakshmi Institute of Technology
Technology and Advanced Studies Technology and Advanced Studies
Chennai,India
Chennai,India Chennai,India
sathishmrl30@gmail.com
sivachandra487@gmail.com kumudham.se@velsuniv.ac.in

Rajendran.V G.R.Jothi lakshmi

Department of ECE Department of ECE
Vels Institute of Science, Technology and Advanced Studies Vels Institute of Science, Technology and Advanced Studies
Chennai,India Chennai,India
drvrajen@gmail.com jothi.se@velsuniv.ac.in

Abstract—Nowadays, Artificial Intelligence appears in the

domain of goetechnics, underwater acoustics, tunneling, geomor-
phology engineering and also in several fields too. This paper
focused on the prospectivefor machine learning approaches which
are sub field of artificial intelligence especially in underwater
acoustics domain. In this proposal, machine learning approaches
such as light gradient boosting, logistic regression, and random
forest classifier algorithms are used for categorizing rocks or
mines from collected sonar dataset.Based on performance met-
rics such as precision, F-score, recall, execution time, accuracy
and confusion matrix, evaluate overall performance of machine
learning models. Here, the experimental results shows that
among all classifier algorithms, light gradient boosting achieves
greater validation accuracy as 95% also training accuracy as
100* moreover, random forest classifier achieves 100% accuracy
duringtraining phase.
Index Terms—Light gradient boosting algorithm, Rock/mine Fig. 1. SONAR technology in underwater seafloor.
classification, Machine Learning algorithms (MLA), Accuracy,
Precision,Execution time.

I. INTRODUCTION zoomed imagery. Gradient Boosting classifier produced better

Sound Navigation and Ranging, or SONAR, is a type of accuracy and how SONAR technology works in underwater
unmanned vehicle which utilizes sound technology to detect sea is illustrated in figure 1.
submerged objects, navigate underwater, and communicate. Primary intention of this work is mentioned as below
The acoustics can range from infrasonic (low) to ultrasonic 1. To distinguish underwater seafloor metal like minerals
(high). Underwater acoustics, often known as hydro acoustics, into rock/ mine using sonar data sets.
is the study of underwater sound. Active sonar and passive 2. To develop the sub field of AI technique named as MLA
sonar are the two types of SONAR. The non-passive trans- especially supervised classifier algorithms appropriate for
ducers create some sound pulse into the water, which is rock/mine classification.
used to detect underwater objects. In case some of the things 3. To evaluate overall performance of machine learning
available in the way of sound pulse, the echo resiles off of models, measures like F-score, recall, precision, execu-
it and come back to the sonar device as a ”echo.” The signal tion time, accuracy also confusion matrix were evaluated.
strength is measured by the transducer if it is ready to receive 4.Based on several classes such as TP, FP, TN and FN,
signals.Several algorithms as decision tree, k nearest neighbor, how the underwater acoustics sonar data are predicted
support vectors and gradient boosting by [1] which isolates the and classified correctly and wrongly by the way finally
underwater sea objects specifically mines or rocks to obtain classification of rock/mine performs.

978-1-6654-9260-7/23/$31.00 2023
c IEEE 177

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on June 08,2023 at 09:53:48 UTC from IEEE Xplore. Restrictions apply.
 The contribution of my research work is described as follows
in terms of several advantages of rock or mine classification:

a. Make possible improved in both risk management as

well as optimization.
b. Provide novel prospective on underwater acoustics
sonar data for classifying mine like materials such as
rock, mine, mud, clay and gravels.
c. Provides high-quality sonar images in underwater
seabed appropriate for distinguishing rock/mine while
still maintaining accuracy metrics, by the way helping
to speed up decision making procedure in finding under-
water sea minerals.

II. RELATED WORK

Several researchers applied various machine learning tech- Fig. 2. Proposed architecture for rock/mine classification using ML classifiers.
niques in various domains especially rock/mine classification
in underwater seafloor data using sonar devices. Several ap-
StandardScalar for leveling as well as data conversion of
proaches such as artificial intelligence by Gholami et. al [19]
attributes in sonar.
and Lawal et. al [20], Machine learning algorithm by Revathy
iv. Fixing the threshold value as 0.5 named as alpha level
et. al [25] for finding malicious behaviour in network based
v. Splitting sonar dataset into training phase and testing
environment, Fast Region Convolutional Neural Network by
phase as 90%, and 10% correspondingly.
Abhishek et. al [13], SVM used by Harvinder singh et. al
vi. Apply the classification algorithm such as Light Gra-
[26] Ravi et. al [8], and Random Forest by Nikitha et. al
dient Boosting classifier, Random Forest classifier, and
[12] Feed Forward Neural Network utilized by choi et. al [3],
another regression technique named as logistic to distin-
introduced computer vision by Maitre et. al [17], introduced
guishing objects in submarine acoustics via resources.
back propagation Neural Network by Hafiz et. al [4]applied
vii. To estimate the performance of machine learning
deep learning based approach for mine/rock automatic classi-
model entirely, metrics namely recall, accuracy based on
fication by Alferez et. al [2] [7], introduced Recurrent CNN
confusion matrix, precision are evaluated.
algorithm by Han et. al [31] for detecting mine like materials
in underwater sea or ocean and finally distinguishing rock or The proposed architecture is shown in figure 2 which specifies
mine by Sidhartha et. al [27] and Qader et. al [10] using under- how the datasets are gathered, followed by pre-processing,
water acoustics data. Also some other researchers studied how selecting relevant features, apply machine learning classifier,
MLAs applied in developing mineralogical data by Jooshaki analysis of classifier, finally the output is evaluated as to
et. al [15] and exploration, exploitation and reclamation by distinguish rock or mine.
jung et. al [18]. Xie et. al [16] introduced computing models IV. STEPS USED FOR CLASSIFICATION (ROCK/MINE)
in calculating the size of rock present in underwater seafloor,
generic approach introduced by vitthal et. al [28] and described A. Import dataset
the framework for rock mechanics by Babaeian et al [29]. In this research work, initiallystage is to upload the Sonar
Deep learning based algorithms were utilized by Lee et. al [5] Dataset in Jupyter Notebook 6.0.1 along with python program-
and Abdesselam et. al [11] for underwater materials detection ming language version 3.7. The description of sonar dataset
Mukherjee et. al [9] and the underwater signal [14] detected comprises of 208 Rows, and 61 Columns
done by Fong et. al [6].Contreros et. al [30] focused on
distinguishing rock type in sandstone depends on formation B. Make in class column of Y axis, Rock is 0 and mine is 1
of wireline pressure data. Kumudham et al [33] utilised deep Here, actual Y-axis represents the output of Class Column,
learning approaches for classification of subbottom image. but it is in alphabet, hence the specification of minerals is
referred as 0 for Rockand 1 for mine.
III. PROPOSED METHODOLOGY
The workflow of our proposed method illustrates the C. Delete low correlations columns
categorization of minerals in underwater acoustics as mines, Here, deleting the features which are having less
rocks and mud via SONAR datasets. correlations (i. e) the input and output data have either
less relationship or else no relationship. So the irrelevant
i. Collecting dataset from the specified resource especially data appear in the column wise are deleted. If both input
SONAR dataset. and output of correlation matrix is zero, which indicates
ii. The sonar dataset are loaded for perform training. very less relationship leads to predict the rock or mine
iii. Import pre-processing for feature extraction and apply classification wrongly. The deleted attributes are shown below

178 International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE)

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on June 08,2023 at 09:53:48 UTC from IEEE Xplore. Restrictions apply.
 Fig. 3. Class balance for 208 instances using target analysis.

with specified features.

del a[’attribute23’], del a[’attribute24’]
del a[’attribute25’], del a[’attribute26’]
del a[’attribute27’], del a[’attribute28’]
del a[’attribute32’], del a[’attribute38’] Fig. 4. Selecting the significant features from sonar dataset.
del a[’attribute39’], del a[’attribute40’]
del a[’attribute41’], del a[’attribute42’]
del a[’attribute46’], del a[’attribute47’] column of the data into 0 to 1. This is utilized for converting
del a[’attribute48’], del a[’attribute50’] higher value into lower values of 0 to 1. And also this scalar
del a[’attribute54’], del a[’attribute57’] makes the calculation speed very fast. The StandardScaler
del a[’attribute58’], del a[’attribute59’] function produces a distribution with a standard deviation of
del a[’attribute60’] one. Because variance equals standard deviation squared, the
variance is also equal to one. And 1 divided by 2 equals 1.
The mean of the distribution is set to 0 by StandardScaler.
D. Target analysis
Approximately 68 percent of the readings will be in the range
The feature of underwater sonar dataset about which desire of -1 to 1. The standard scalar along with standard deviation
to learn more is known as the target variable of a dataset and mean are represented in equation (1)
or target analysis. A supervised machine learning algorithm
learns patterns and uncovers links between other features in džLŵĞĂŶ dž  
^ƚĂŶĚĂƌĚ^ĐĂůĂƌ  (1)
underwater sonar dataset and the target value with previous ^ƚĂŶĚĂƌĚĞǀŝĂƚŝŽŶ dž 
data. Here the target analysis is necessary for mentioning
H. Splitting data intotraining and testing phase
distinguishes of rock or mine (class 0 or class 1) along with
208 instances which are depicted in figure 3. Here split the features into Training stage as 90 percentage
and Testing data as 10 percentage. Since,during training stage,
E. Splitting of X and Y machine learning algorithm teach the features of sonar data
Next step is to split the data in to X and Y. Because, around 90 percentage. Also during testing stage, based on our
developing machine learning algorithm to isolate input data chosen machine learning classification algorithms validate the
as well as Output Data via training and testing phase. data nearly 10 percentage. In other words,during testing stage,
F. Feature importance the input of sonar data has only given but our machine learning
will produce the output automatically.
Can find which features are more important, with respect to
target specified in figure 4. V. MACHINE LEARNING ALGORITHMS
G. Standard scalar In this research work, supervised machine learning clas-
With eliminating average value and then leveling to entity sifiers are used which are appropriate for detecting the mine
variance, Standard Scaler standardizes a feature. Divide all of like materials and also classifying the data into mine or rock in
the numbers by the standard deviation to get unit variance. underwater seafloor from sonar datasets. The following are the
The formal concept of scale that introduced earlier does machine learning classifier which provides maximum accuracy
not apply to StandardScaler. Now, convert X axis of each in identification, classification of SONAR data too.

International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE) 179

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on June 08,2023 at 09:53:48 UTC from IEEE Xplore. Restrictions apply.
 Fig. 5. Logistic Regression based Sigmoid function.

A. Light Gradient Boosting classifier

Light Gradient Boosting classifier is one of the classifier Fig. 6. Random Forest Classifier .
approaches introduced by Guolin ke et. al [24] which specifies
the highly efficient gradient boosting decision tree for better
classification of dataset. This classifier is publicly accessible C. Random Forest classifier
toolkit which implements the gradient boosting technique in
Several decision tree classifiers together formed random
an efficient and effective manner. LightGBM improves on
forest method which is one of the SMLA utilized for both
the gradient boosting technique by incorporating automatic
regression classification issues shown in figure 6. Every de-
feature selection and concentrating with enhancing instances
cision tree grows as well with full development; it doesn’t
by superior ascents. In this situation, this method led into a
necessitate to engraving the processing. If the trees are in
significant increase in speed during learning and enhanced pre-
large quantity, then the outcome also will be high that are
diction accomplishment. As a result, while functioning through
not over-fitting. This Random Forest approach will perform
statistics tabulation via regression as well as categorization an-
estimation wholly and also has benefits especially selection of
alytical modeling responsibilities, this classifier developed into
features automatically. Lin et. al [22] and Manish kumar et. al
effective approach for learning based on supervised contests.
[21], Kumudham et.al. [32] utilized ensemble random forest
Consequently, along with Extreme Gradient Boosting, it shares
algorithm for large scale applications especially insurance
some responsibilities for growing esteems along with broader
business data analytics. Furthermore, xtreme machine learning
acceptance of GB approach in wide-ranging (XGBoost). For
based random forest method were applied for categorizing
this reason, several classifiers were applied for distinguishing
heart beat automatically attains an accuracy of 98%
minerals like objects in underwater seafloor or seabed via
When the input dataset is large, this approach performs
specific resources.
better. The forest is densely forested, and a significant amount
B. Logistic Regression of DTs are utilized in administrative process. The image
Logistic regression is one of the machine learning algo- collection is split into two halves with similar formation at
rithms represents some deviation of linear regression which first. The edge function using RF algorithm is mentioned in
can be utilized for classification function hence this algorithm equation (3)


appropriate for rock/mine distinguishing in underwater seabed
or seafloor. Logistic regression utilize the significant function
named as “squashing” that essentially convert any kind of Here I corresponds to Indication function h1 (X), h2(X). hn
factual value into value among 0 and 1. The representation (X) are classifiers with ensembling method where X and Y
of sigmoid function is shown in equation (2) and figure5. are random vectors. The function that denotes error which is
 defined as equation (4)
^ ƚ   (2)
Ğ−ƚ
Now the edge function of RF is rewrite as equation (5)
By defining the threshold value, detecting and classifica-
tion of rock/ mine performs on sonar dataset gathered from
underwater seafloor. Suppose for example if threshold value
defined is 0.5, also the output value generated from sigmoid
function exceeds 0.5 will be considered as class 1 specified as
Here represents the potency of classifiers which is repre-
mine, moreover the output value produced by sigmoid function
sented as (6)
reaches underneath or similar to threshold value (0.5) will be 
measured as class 0 indicated as rock.

180 International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE)

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on June 08,2023 at 09:53:48 UTC from IEEE Xplore. Restrictions apply.
 C. Precision
The precision is measured as the proportion among the
number of correctly identified Positive samples (rock) to the
total number of Positive samples (either suitably or impre-
cisely). This metric enhances increase while several MLA
shave many correct categorization and less significant amount
imprecise of categorization. Equation (9) specifies the formula
for calculating precision.

D. Recall
negative. The estimation of this metric has done using formula
Fig. 7. Confusion matrix on rock/mine classification along with classes (10)
Actually Positive.

The RF technique is a time-saving strategy that merges mul- 

tiple decision trees into a single tree for the most accurate E. Execution time
prediction. Execution time can be defined as the time period of every
VI. METRICS ESTIMATION classifier algorithm utilized for classification of mine/rock.
Three machine learning modelsare introduced hence it is
The entire machine learning approach performance are
necessary to calculate the execution time for every functional
analyzed based on specific metrics like recall, f-score, pre-
MLAs.
cision, accuracy, execution time and finally confusion matrix
in identifying minerals in underwater sea along with its F. F score
underwater object classification. By evaluating these metrics,
F score metric is also used to evaluate the binary classi-
the underwater seafloor materials are classified as true positive,
fication model. The F1 score, which is one of the accuracy
true negative, false positive and false negative on sonar data
measure metrics, defines the harmonic mean of both precision
collected from underwater seafloor via SONAR device.
and recall. This parameter is estimated here to separate the
A. Confusion matrix sonar data into rock and mine. Equation (11) can be used to
The two-label classifier confusion matrix is utilized to compute the value of f-score.
analyze the confusion matrix in order to identify the seafloor
minerals available in underwater as well as distinguishingThis
metric describes the proportion of truly predicted positively
from both truly predicted positively and falsely predicted

the data into rock/mine. This two-label classifier aid in the
VII. RESULTS ANALYSIS
classification of data into two classes namely class 0 (rock) and
class 1 (mine). Using confusion matrix, sonar data distinguish
into two classes is explained in figure given below. In this section, discussing abouthow classification and re-
B. Accuracy gression techniques appropriate in identifying materials like
rock, mine, clay, mud, gravels etc. Also, distinguishing
Accuracy describes as the percentage of number of appro- rock/mine using three MLAs are discussed as table. Descrip-
priatelysecret datas on sonar dataset to the whole quantity of tion of datasets, default attributes as well as how machine
samples taken. The most important thing need to know is based learning produces the results in terms of ROC (Receiver Op-
on this specific accuracy metric measure, the overall machine
erating Characteristic) curve, confusion matrix and some other
learning model performance is evaluated in identifying metal metrics. The quality of the evaluation of two classification and
mine detection along with two-level classification which may one regression model using sonar dataset for three classifiers
be either class 0 or class 1. Accuracy can be measured using is summarized in table 2.Also, the maximum accuracy value
the formula (7) in percentage measuring accuracy as equation for every machine learning classifier is highlighted.
(8).
 A. ROC under Curve
(8) ROC under curve, thehistogram that shows how a binary
classifier system’s diagnostic performance changes while the
discrimination threshold is changed. Starting in 1941, the ap-
proach was created for operative of martial sonar transceivers,
In percentage, the accuracy can be evaluated using formula hence the name. The below ROC curve represents the plot for
truly classified ratealigned with the rate of positively classified

International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE) 181

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on June 08,2023 at 09:53:48 UTC from IEEE Xplore. Restrictions apply.
 Fig. 9. ROC curve among FPR versus TPR using Logistic Regression.

Fig. 8. ROC curve for LGBM classifier.

inaccurately at dissimilar threshold levels. In machine learning,

the TPR is regarded asidentification of likelihood, recall and
sensitivity. Also, likelihood of fake alarm is recognized as the
False positive rate, and it can be computed as (1- specificity).
ROC is also known as a relative operating characteristic curve
since it compares two parameters (TPR and FPR) when the
criterion changes.The ROC curve for LGBM classifier in Fig. 10. ROC curve among FPR versus TPR using Random Forest classifier.
classifying mine/rock by plotting the graph as FPR versus TPR
is depicted in figure 7.
The below ROC curve represents the plot for truly classified classifier for classifying rock or mine by plotting the graph as
rate aligned with the rate of positively classified inaccurately FPR versus TPR is depicted in figure 9. The accuracy attained
at dissimilar threshold levels. In machine learning, the TPR is for distinguishing rock/mine as class 0 and class 1 is 100* for
regarded as identification of likelihood, recall and sensitivity. both ROC and micro-average ROC.
Also, likelihood of fake alarm is recognized as the False pos-
itive rate, and it can be computed as (subtract specificity from B. Discussion on confusion matrix of supervised ML tech-
one). ROC is also known as a relative operating characteristic niques
curve since it compares two parameters (TPR and FPR) when 1) Light Gradient boosting classifier (LGBC): The confu-
the criterion changes. The ROC curve for Logistic Regression sion matrix for LGBC along with two classes namely class 0
for classifying rock or mine by plotting the graph as FPR as rock, and class 1 as mine is illustrated in figure 10. The
versus TPR is depicted in figure 8. The accuracy attained for value of True Positive value is 12, False Positive value as 1,
distinguishing rock/mine as class 0 and class 1 around 94% False Negative as 0, and Truly Negative as 8.
for both ROC and micro-average ROC. 2) Logistic Regression: The confusion matrix for Logistic
The below ROC curve represents the plot for truly classified Regression along with two classes namely class 0 as rock, and
rate aligned with the rate of positively classified inaccurately class 1 as mine is illustrated in figure 11. The value of True
at dissimilar threshold levels. In machine learning, the TPR is Positive value is 9, False Positive value as 4, False Negative
regarded as identification of likelihood, recall and sensitivity. as 1, and Truly Negative as 7.
Also, likelihood of fake alarm is recognized as the False posi- 3) Random Forest Classifier (RFC): The confusion matrix
tive rate, and it can be computed as (Minus the specificity from for RFC along with two classes namely class 0 as rock, and
1). ROC is also known as a relative operating characteristic class 1 as mine is illustrated in figure 12. The value of True
curve since it compares two parameters (TPR and FPR) when Positive value is 10, False Positive value as 3, False Negative
the criterion changes. The ROC curve for Random Forest as 0, and Truly Negative as 8.

182 International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE)

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on June 08,2023 at 09:53:48 UTC from IEEE Xplore. Restrictions apply.
 Fig. 13. Evaluation of confusion matrix using Random Forest.

Fig. 11. Confusion matrix for LGB classifier.

minerals such as rock, mine, clay, mud, gravels etc from
sonar based dataset. Three MLAs such as Light Gradient
Boosting classifier, Random forest and logistic regression are
utilized for making better classification of rock or mine in
sonar dataset. Among three algorithms, light gradient boosting
achieves greater validation accuracy of 95%, logistic regres-
sion algorithm reaches less execution time of 0.015 seconds.
Several performance measures namelyf-score, recall and pre-
cision, confusion matrix, testing and training accuratenesswere
evaluated for estimating the overall performance of machine
learning classifiers in detecting and classifying rock or mine
in underwater seafloor on sonar dataset.

REFERENCES
Fig. 12. Performance of confusion matrix using logistic regression. [1] Padmaja, V., Rajendran, V. Vijayalakshmi, P. Study on metal mine
detection from underwater sonar images using data mining and machine
learning techniques. J Ambient Intell Human Comput 12, 5083–5092
(2021). https://doi.org/10.1007/s12652-020-01958-4
C. Comparison of several machine learning classifiers [2] Alférez, G. H., Vázquez, E. L., Martı́nez Ardila, A. M., Clausen,
Several metrics like training accuracy, testing accuracy, B. L. (2021). Automatic classification of plutonic rocks with
deep learning. Applied Computing and Geosciences, 10, 100061.
execution time, precision, recall and F1-score were evaluated doi:10.1016/j.acags.2021.100061
for three machine learning techniques are summarized in table [3] Choi, J., Choo, Y., Lee, K. (2019). Acoustic Classification
2. of Surface and Underwater Vessels in the Ocean Using Super- vised
Machine Learning. Sensors (Basel, Switzerland), 19(16), 3492.
https://doi.org/10.3390/s19163492
TABLE I [4] Leo, L.M. and Reddy, T.K., 2021. Learning compact and discriminative
COMPARISON OF SEVERAL METRICS AMONG MACHINE LEARNING hybrid neural network for dental caries classification. Microprocessors and
CLASSIFIERS Microsystems, 82, p.103836.
[5] Lee, S., Park, B., Kim, A. (2018). Deep Learning from Shallow Dives:
Metrics Light Logistic Random Sonar Image Generation and Training for Underwater Object Detection.
Gradient Regression Forest ArXiv, abs/1810.07990.
Boosting Classifier
[6] Leo, L.M. and Reddy, T.K., 2014. Digital Image Analysis in Dental Ra-
Training Accuracy 100% 88% 100% diography for Dental Implant Assessment: Survey. International Journal of
Testing Accuracy 95% 76% 86% Applied Engineering Research, 9(21), pp.10671-10680..
Execution Time (seconds) 0.10 0.015 0.14 [7] Ritwick Ghosh “Sonar Target Classification Problem: Machine Learning
Precision 0.94 0.77 0.86 Models”, International Journal of Science and Research, volume 9, Issue 1,
Recall 0.96 0.78 0.88 2020, pp. 247-248.
F1 Score 0.95 0.76 0.86 [8] Ravi K Jade, L.K. Verma, Kesari Verma “Classification using Neural
Network Support Vector Machine for Sonar dataset”, International Journal
of Computer Trends and Technology, volume 4, Issue 2, 2013, ISSN:
2231-2803, pp. 116-119.
VIII. CONCLUSIONS [9] Megalan Leo, L. and Kalpalatha Reddy, T., 2020. Dental caries clas-
sification system using deep learning based convolutional neural net- work.
This proposed research work focused on how machine Journal of Computational and Theoretical Nanoscience, 17(9-10), pp.4660-
learning classification algorithms helpful in finding underwater 4665..

International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE) 183

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on June 08,2023 at 09:53:48 UTC from IEEE Xplore. Restrictions apply.
 [10] S.Kannadhasan and R.Nagarajan, Design and Development of ing, Volume 100, Issue 1, 2018, Pages 102-116, ISSN 1674-7755,
Environmentally W-Shaped Structure Antenna for Wireless Applications, https://doi.org/10.1016/j.jrmge.2017.09.007.
International Web Conference on Smart Engineering [29] Babaeian,M., Ataei,M., Sereshki, F., Sotoudeh, F., Mohammadi, S.,
Technologies(IWCSET 2020), 26-27 June 2020, Ramco Institute of 2019. A new framework for evaluation of rock fragmentation in open
Technology, Rajapalayam, ISBN:978-93-5407-648-0, Published in Journal pit mines. J. Rock Mech. Geotech. Eng. Volume 11, Issue 2, 325–336.
of Green Engineering, Volume 10, Issue 9, September 2020. [30] Jose Victor Contreras “Supervised Learning Applied to Rock Type
[11] Abdesselam Bouzerdouma , Philip B. Chappleb , Mark Drasc , Yi Classification in Sandstone Based on Wireline Formation Pressure Data”,
Guod , Len Hameyc , Tahereh Hassanzadehc , Thanh Hoang Lea DOI:10.1306/42539Contreras2020.
, Omid Mohamad Nezamic , Mehmet Orgunc , Son Lam Phunga , [31] Fenglei Han, Jingzheng Yao , Haitao Zhu, and Chunhui Wang “Un-
Christian Ritza , Maryam Shahpasandc “improved deep-learning-based derwater Image Processing and Object Detection Based on Deep
classification of mine like contacts in sonar images from autonomous CNN Method”, Hindawi, Journal of Sensors, Volume 2020, Article ID
underwater vehicles”, UACE2019 Conference Proceedings, ISSN- 2408- 6707328, pp: 1-20, https://doi.org/10.1155/2020/6707328.
0195, pp.179-186. [32] R.Kumudham, Dr.V.Rajendran, “Classification Performance Assessment
[12] T.Nikitha, B.Hadhvika, Dr.K.Vaidehi “Prediction of Underwater Surface in Side Scan Sonar Image while Underwater Target Object Recognition
Target through SONAR by using Machine Learning Algorithms”, JETIR using Random Forest Classifier and Support Vector Machine”, Interna-
June 2019, Volume 6, Issue 6, ISSN-2349-5162, pp. 159-164. tional Journal of Engineering and Technology, IJET(UAE), Vol 7, No
[13] N Abhishek, Arjun R, Bharathesh R, Kavitha K S, Prof. Manonmani S, 2.21 (2018) pp 386-390, Special Issue-21, Special Issue on Advance
Dr. Shanta Rangaswamy “Underwater Mine Detection using Image Pro- Engineering and Science, ISSN-2227524X.
cessing”, International Research Journal of Engineering and Technology [33] Radhika Surampudi, Kumudham.R, , Ebenezer Abishek.B, , Rajen-
(IRJET), Volume: 07 Issue: 05, May 2020, pp. 7949-7952. dran.V, “Underwater Sediment Layers analysis using Convnet with
[14] Dhiraj Neupane, Jongwon Seok “A Review on Deep Learning-Based Adam Optimiser and mapping its reflection coefficient parameter with
Approaches for Automatic Sonar Target Recognition”, Electronics 2020, the Particle Size” International Journal of Engineering Trends and
volume 9, Issue 11, 1972; doi: 10.3390/electronics9111972. Technology, issn:2231 – 538, Dec 2021.
[15] Jooshaki, M.; Nad, A.; Michaux, S. A Systematic Review on the
Application of Machine Learning in Exploiting Mineralogical Data in
Mining and Mineral Industry. Minerals 2021, 11, 816. https://
doi.org/10.3390/min11080816
[16] Chengyu Xie, Hoang Nguyen, Xuan-Nam Bui, Yosoon Choi, Jian
Zhou, Thao Nguyen-Trang, “Predicting rock size distribution in mine
blasting using various novel soft computing models based on meta-
heuristics and machine learning algorithms”, Geoscience Frontiers,
Volume 12, Issue 3, 2021, 101108, ISSN 1674-9871,
https://doi.org/10.1016/j.gsf.2020.11.005.
[17] Maitre, J., Bouchard, K., Bédard, L. P. (2019). Mineral grains recog-
nition using computer vision and machine learning. Computers Geo-
sciences, 130 (2019),pp. 84-93, doi:10.1016/j.cageo.2019.05.009.
[18] Jung, D.; Choi, Y. Systematic Review of Machine Learning Applications
in Mining: Exploration, Exploitation, and Reclamation. Minerals 2021, 11,
148. https://doi.org/10.3390/min11020148.
[19] Gholami, R., Rasouli, V., Alimoradi, A. (2012). Improved RMR Rock
Mass Classification Using Artificial Intelligence Algorithms. Rock Me-
chanics and Rock Engineering, 46(5), 1199–1209. doi:10.1007/s00603-
012-0338-7.
[20] Lawal, A. I., Kwon, S. (2020). Application of artificial intelligence
to rock mechanics: An overview. Journal of Rock Mechanics and
Geotechnical Engineering. doi:10.1016/j.jrmge.2020.05.010.
[21] S.Kannadhasan, R.Nagarajan and R.Banupriya, Performance Improvement
of an ultra wide band antenna using textile material with a PIN diode,
Textile Research Journal, DOI: 10.1177/00405175221089690
journals.sagepub.com/home/trj
[22] WEIWEI LIN, ZIMING WU, LONGXIN LIN, ANGZHAN WEN, AND
JIN L, “An Ensemble Random Forest Algorithm for Insurance Big Data
Analysis”, Special Section On Recent Advances In Computational
Intelligence Paradigms For Security And Privacy For Fog And Mobile
Edge Computing, Volume 5, 2017, pp. 16568-16575, Digital Object
Identifier 10.1109/ACCESS.2017.2738069.
[23] Yang, P., Wang, D., Zhao, W.-B., Fu, L.-H., Du, J.-L., Su, H. (2021).
Ensemble of kernel extreme learning machine based random forest
classifiers for automatic heartbeat classification. Biomedical Signal
Processing and Control, 63, 102138. doi:10.1016/j.bspc.2020.102138
[24] Guolin Ke, “LightGBM: A highly efficient gradient Boosting Decision
Tree”, 2017.
[25] G. Revathy, V. Rajendran, P. Sathish Kumar, “Development of IDS using
mining and machine learning techniques to estimate DoS Malware” Int.
J. Computational Science and Engineering, Vol. 24, No. 3, 2021, pp.
259-275, DOI: 10.1504/IJCSE.2021.115646
[26] Harvinder Singh, Nishtha Hooda “Prediction of Underwater surface
target through SONAR: A Case-Study of Machine Learning”, Microser-
vices in Big Data Analytics, 2020. [
[27] Jetty Bangaru Siddhartha, T Jaya, V. Rajendran, “Distinguished Study
of Various Materials Like Metal Mine/ Rock from Underground Data of
Unmanned Vehicle Using Machine Learning”, International Journal of
Grid and Distributed Computing Vol. 13, No. 2, (2020), pp. 574–583 [
[28] Vitthal M. Khatik, Arup Kr. Nandi, A generic method for rock mass
classification, Journal of Rock Mechanics and Geotechnical Engineer-

184 International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE)

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on June 08,2023 at 09:53:48 UTC from IEEE Xplore. Restrictions apply.