See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/350756075

Application of Machine Learning in Wireless Sensor Network

Article · March 2021

CITATIONS READS

2 899

3 authors:

Mahendra Prasad Nath Sachi Mohanty

Siksha O Anusandhan University KIIT University
17 PUBLICATIONS 122 CITATIONS 181 PUBLICATIONS 2,069 CITATIONS

SEE PROFILE SEE PROFILE

Sushree Bibhuprada B. Priyadarshini

Institute of Technical Education and Research
50 PUBLICATIONS 145 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mahendra Prasad Nath on 09 April 2021.

The user has requested enhancement of the downloaded file.

 Proceedings of the 15th INDIACom; INDIACom-2021; IEEE Conference ID: 51348
2021 8 International Conference on “Computing for Sustainable Global Development”, 17th - 19th March, 2021
th

Bharati Vidyapeeth's Institute of Computer Applications and Management (BVICAM), New Delhi (INDIA)

Application of Machine Learning in Wireless Sensor

Network
Mahendra Prasad Nath Sachi Nandan Mohanty Sushree Bibhuprada B. Priyadarshini
Dept. of Computer Science & Engineering Dept. of Computer Engineering Dept. of Computer Sc. & Info. Technology
Siksha „O‟ Anusandhan Deemed to be College of Engineering Pune Siksha „O‟ Anusandhan Deemed to be
University Pune, India University
Bhubaneswar, Odisha, sachinandan09@gmail.com Bhubaneswar, India
indiamahendranath272@gmail.com bimalabibhuprada@gmail.com

Abstract—Wireless Sensor Networks (WSNs) find broad signal checking in the social insurance industry, and so forth.
range of adoration and wide spread uses in various With the fast advancement in MEMS innovation, it is
applications. In such networks, the deployed sensors trap and conceivable to convey enormous scope sensors in the field
transfer the data intellectually to the base station that is the sooner rather than later [10, 12]. In any case, enormous
ultimate destination. This paper briefs about the various scope sensors arrange additionally unavoidably present huge
applications of machine learning in sensor networks. The measure of information in WSNs to be handled, transmitted,
Principal Component Analysis (PCA) along with k-means and gotten. Transmitting all information back to a base
clustering strategies, used in case of unsupervised learning, are
station for preparing and making derivations is only
discussed. A discussion on different functional challenges
outlandish because of the sensor's restricted vitality and data
occurring in case of sensor networks is also presented.
transfer capacity limitations. Along these lines, there is a
Keywords—Wireless Sensor Network, Machine Learning, requirement for applying AI techniques in WSN. This
Supervised Machine Learning, Unsupervised Machine Learning, system could fundamentally diminish the measure of
Re-inforcement Machine Learning. information correspondences and really use the distributive
quality of WSNs.
I. INTRODUCTION
II. WIRELESS SENSOR NETWORK
Machine learning (ML) is a very popular field of study,
with multiple applications in various areas ranging from The organizer of wireless sensors includes a number of
computer games [1], gaming hardware [2], Virtual Reality tiny hubs that capture sensed data and send it to end users or
[3], Medical Applications [4, 5], educational applications members. Sensor hubs kept detecting, occasion recognition,
[6], software engineering applications [7], Internet of Things area identification and a few attributes like adaptation to
applications (IoT) and wireless sensor networks (WSN) [8]. non-critical failure, fast arrangement, the self-association
In parallel, WSNs have pulled in a lot of consideration lately. makes them utilize numerous basic situations like Military
Research in the WSNs zone has concentrated on two reconnaissance, Home mechanization, Target following,
separate parts of such systems, in particular systems Health checking, Sales following, Fire discovery, Internet of
administration issues, for example, limit, delay, and directing Things, Seawater levels, Volcano perceptions. For making
techniques; and application issues [9-22]. This paper reviews choices these applications required great and complete
the AI strategies utilized in WSNs from the two viewpoints. information. In a vicious domain a huge number of endless
An assortment of different circulated sensors is broadly sensors are sent to regulate, collect detected data to sink hub.
known as Wireless Sensor Network (WSN). Broadly the Sensor hubs thickly sent single sensors, often connecting
WSN is utilized for fundamentally two applications radio to the interface with other sensor hubs via remote
−monitoring and following [19, 20]. Specifically, in WSN transmission medium. Sensor hubs have restricted
following applications the goal of sensor hub is to follow the computational force, vitality assets with their small size, and
creature, foe, human, traffic, vehicle/transport and so forth, another tremendous issue is the condition they send quickly
wherein WSN checking applications sensor hub is screen changes after some time. It is particularly necessary to
nature, creature/ understanding development, and security examine information obtained from sensors in a timely
discovery and so on. These days physical condition joins manner. Untreated data collected from the sensor hub is
huge and differed data choices, for example, seismic waves; inaccurate and is in culmination [17, 19].
movement, light, temperature, etc. It is important to get the Sending the invention of the WSN will take advantage of
data from various shifted alternatives for a more noteworthy a wide variety of utilizations starting from biological system
perception of the earth just as the remote sensor arrangement observation to battle zone observation. A portion of the
(WSN) is easy to utilize framework empowering getting this benefits include minimal effort, simple structure, strong
sort of inexhaustible data. identification of constancy, and self-association of WSNs,
WSNs innovation has a promising future in numerous among a few different benefits. Be that as it may be, the use
application spaces. They have been utilized as front line of WSN innovation accompanies incredible difficulties
reconnaissance instruments in military applications, building regardless of the numerous open doors the remote sensor
basic observing in the development industry, tolerant bio- systems give. As opposed to steering, to plan a remote

Copyright © INDIACom-2021; ISSN 0973-7529; ISBN 978-93-80544-41-0 403

 Proceedings of the 15th INDIACom; INDIACom-2021; IEEE Conference ID: 51348
2021 8 International Conference on “Computing for Sustainable Global Development”, 17th - 19th March, 2021
th

sensor organize originator should take care about difficulties hubs change or move starting with one spot then onto the
like confinement, arrangement, booking, security, next that will influence the outcome. So it's satisfying to
information total, Quality of administrations. Each senor hub create sensor arranges that can be controlled intensely in
present huge measure of information for transmitted, handled those situations.
and got to WSN. Handling and settling on choices is
outlandish because of the sensor restricted vitality and data IV. APPLICATIONS OF MACHINE LEARNING IN WSN
transfer capacity limitations. So there is require of relate Now a day‟s, Machine Learning is used in various
Machine Learning [11, 15]. applications around us to make applications more users
friendly. This section describes the various applications of
III. MACHINE LEARNING machine learning in Wireless Sensor Networks.
The latest developments in Machine Learning (ML) and
soft computing techniques allow for the construction of more A. Supervised Machine Learning
useful prediction frameworks defined as the set of The system model is developed with proven advance
measurements. The scholarly model may be just a inputs and output values in the supervised machine learning.
fundamental parametric capacity, gained from information, When designing supervised machine learning some steps has
two or three information factors generally customary to be recalled: (a) The intruder should know what kind of
measurements or perspective, permitting yield state or information is being used to plan the set. (b) Preparation
variable not out of the ordinary correctly. The WSN's can knowledge speaks to the work of this current truth. (c) One
incorporate heterogeneous, various self-ruling, and ideal structure should be represented by learning algorithm.
reasonable just as negligible force sensor hubs. AI lighting (d) Decide the key perspectives from preparing
up the data provides a solution for itself. AI is a worldview informational collection. (e) Finally, whether the range of
that alludes to gaining from antiquated occasions parameters and configuration determines if the precision of
understanding to improve future execution. The lone point of the learning estimate is feasible by calculating with the test
convergence of this field is programmed learning [11, 16]. set that is separated from the knowledge collection planning
the estimation mapping of the yield variable as we evaluate
ML ability is utilized immensely for a plentiful size of [11]. Fig.1 illustrates the flow of supervised machine
activities consolidates characterization, relapse further- learning process.
greater thickness estimation in the heterogeneity of usage
domains, for example, bioinformatics, discourse location,
spam acknowledgment, PC point of view or perception and
misrepresentation identification alongside communicating
systems. Machine learning is employed in these types of
situations. Several industries such as medicine, the military,
and so on apply machine learning to unsheathe usable data.
The principle advantage of utilizing AI is at one time a
calculation handle what to achieve with information; it can
achieve that work precisely. Rising treatment of AI advances Fig. 1. Supervised Machine Learning Process
in modernizing WSN activities makes them proficient and
profoundly talented [14, 18]. 1) Decision Trees
For the following reasons, machine learning is critical in It organizes the information by scientific categorization
WSN [19]: platform on their standards. This approach compares the
criteria with the state of preference that hits the particular
 WSNs built in raw areas where conventional level. Decision trees are used primarily for process
networks are not deployed, classification. Decision trees convey as a mathematical and
statistical procedure arrangement to help explain, classify a
 The number of assignments in WSN masterminded
given set of data. (a,B) = (a1, a2… ak. B). Here B is the
by some scientific reproduction like steering,
critical attribute to be represented, a1, a2 ... ak are the values
information collection, and so forth under related
on which we have to bear responsibility. Decision algorithms
circumstances AI important to gives low
resolve a number of unresolved issues in the design of the
unpredictability, the computational expense for the
wireless sensor systems, such as link loss, reliability,
framework, and
restoration, corruption, medium failure time [11, 18].
 WSNs, because of a tremendous measure of area Wireless network sensor capabilities have made them
sensors, gather a gigantic measure of information and accessible to all forms of applications. Wireless sensor
send it to the base station notwithstanding separating networks are often used in animal behavior, because they
significant data from immense information is call minimal criteria that make their application quick and
confounded at the base station just as a sensor hub. complicated. Each hub, which is foreshadowed by a
A frequently WSN application accompanies some level connection to the surface guide, remains on the regulated
of information including which lessens the force taking care sensor coordination across 3D space. Gathered knowledge
of with the assistance of AI models to greatest information from direct hubs sent to the focal unit and applies a
inclusion with prevalent sensor organization. Sensor preference tree order to figure out if there are any submerged
organizes for the most part introduced in unpleasant places. Option of trees used to select battery-dependent
situations which alter with respect to time for instance we group head, usability level as knowledge vectors for tree
introduced sensor hubs for constant applications when sensor construction.

Copyright © INDIACom-2021; ISSN 0973-7529; ISBN 978-93-80544-41-0 404

Application of Machine Learning in Wireless Sensor Network

2) K- Nearest Neighbour In Selective forward assault its decrease to sending messages

k-NN AI algorithm groups information is dependent on which it gets a few times its goes about as square gap
the estimations of close to information tests. Here K is a implies when the vindictive hub gets a course demand it will
client characterized worth, doubt or preliminary point is send a directing answer that shows the course to extortion
index through turning over the mark which is a decent goals. Here a Distributed methodology used to diminish the
number continuous in the midst of the k preparing models correspondence including sensor hubs alongside
bordering to the distrust point. It's for the most part abusing computational unpredictability is squat which lessens the
for grouping and relapse. Preparing knowledge is acquainted intensity of sensor hubs. Interruption is a lot of action that
with the student, and then we get to know the latest bargains the respectability classification of sensor organize.
information with the teachers, and the students look at the Sensor hub gathered information is first preprocessing
two information principles. K most related information is through the safe calculation later it will use for SVM for
taken from preparing informational collection alongside interruption identification. This calculation takes choice
larger pieces of k taken as a new class from new measurements like availability data, pointers, and quick
informational index. A K Nearest neighbor capacity with the exercises of sensor hubs for limitation. By utilizing SVM
closest focus so that the cost of calculation is low and strategies we will improve the effective learning process for
manages the more efficient operation of the sensor systems. learning calculations.
Preclusive sensing of the sensor hub can be determined by 4) The Bayesian Network
the use of a normal separation between the nearest neighbor
hubs covered by a clear cut-off. In WSN, the query handling The Bayesian system is probabilistic visual reproduction
subsystem work is set up by means of the nearest neighbor that implies arrange of self-assertive factors besides their
calculation [12, 16]. contingent habit. It is for the most part utilized for bunching
and order reasons. Arranging of Bayesian systems relies
The most deeply appropriate zone of the nearest neighbor upon the restrictive likelihood and produces trees dependent
in the wireless sensor network system is the matter of on their likelihood of an event. Contrast with other AI
handling, but the problem in the nearest neighbor calculations Bayesian systems require somewhat a number
measurement is that it is important to store a lot of of preparing tests which makes an advantage for remote
information from its state. The K-NN region is monitored at sensors which have low force assets however Bayesian use
a point where a hub is outside the boundary area and is present information colleague to refresh prior to ensuing
challenged by the application supervisor [20, 21, 22]. conviction. The reliability of events using incomplete data
sets can be calculated by using Bayesian Network in wireless
3) Neural Networks sensors by analysis of past experience [11].
The neural system is a subsidiary of the organic idea of
neurons. Every neuron association imparted a sign starting Basically, sensors are sent all over the patient's body and
with one neuron then onto the next, accepting neuron predict the behavior performed at each Secret Markov sensor
prepared on the sign and afterward signal neurons associated hub. Choose sensor hubs that capture the most movement
with another or got neuron. Neural systems can be suitable to data. To capture the last signal choice utilize the Bayesian
seven classes of issues: model recording, process control, system model to integrate the automatic hub forecasts.
work estimation, grouping, forecast, and improvement Bayesian systems additionally utilized for limitation utilizing
moreover recovery by content. In WSN utilizing neural just a couple of qualities of sensor hubs like radio recurrence
systems there is still no inescapable inferable from taking off got a signal point, and so on. Some Sensor hub gathered data
computational prerequisites for learning information has relative readings that are utilized to shape contingent
regarding set of associations. Neural systems in WSN for the likelihood holding between hubs to decide if any
most part used to discover the area of sensor hub by forthcoming anomalies in the amassed measurements [11].
Received sign quality, point of spread, and separation of B. Unsupervised Machine Learning
estimations [10].
Unsupervised learning calculations no need of preparing
WSN vivaciously intended for salvage and fire informational index. They take in scarcely any highlights
discovery. Hubs sense the general condition and assembled from the indicated informational collection and apply those
data sends to bunch head, group head process the highlights to the recently presented informational index.
information utilizing the neural system, and just significant Unaided calculations are widely abused in the clustering of
data is sent to leaders. The neural system utilizes Illuminated hubs just like information conglomeration [11, 16]. Figure2
Matrix to assess the evaluation of luminance inside the lit demonstrates the flow of unsupervised machine learning
quarter [16]. process.
Neural systems used to discover vigorous exercises as
well as their causes at advantageous hubs. DoS assault is
somewhat assaulted, focusing on floods with a gigantic
number of message traffic issues, which subsequently
prevents the transmission of productive information. The
focal unit disseminates the scheduling of the openings for Fig. 2. Unupervised Machine Learning Process
each sensor hub inside the system [9].
In WSN, steering is fundamental to move the information 1) Principal Component Analysis
from sensor hub to sink hub. While directing the data PCA is a measurement decrease and information pressure
busybodies make numerous assaults to take the information. movement that can lessen a lot of information variable to

Copyright © INDIACom-2021; ISSN 0973-7529; ISBN 978-93-80544-41-0 405

 Proceedings of the 15th INDIACom; INDIACom-2021; IEEE Conference ID: 51348
2021 8 International Conference on “Computing for Sustainable Global Development”, 17th - 19th March, 2021
th

little set information factors that as yet unmoving the greater Q principles are instated remain on a jump or bounce count
part of the data in huge informational index. It changes to each hub in the set-up. Jump or bounce figuring is
numeral related factors into a small measure of unassociated determined by hi messages between the sensor hub at the
factors entitled head segments. The primary part contains the hour of system organization. Multicast steering in remote
most elevated clash of the information alongside each sensors arranges every hub sends one centrality to various
ensuing segment variant contains as adequate as enduring collectors which expend power assets of the hub to stay
variability as feasible. This method is generally utilized in away from this issue. In the principal stage Join question
WSN hubs to decrease the measure of information forward find the ideal course and updates the Q esteems. In
transmitting rather than this PCA computes the uncorrelated Join's answer forward stage permits the ideal way to
segments which contain unique readings. In occasion multicast transmissions.
recognition, PCA makes more straightforward the critical
thinking by taking scarcely any highlights from the V. FUNCTIONAL CHALLENGES IN WIRELESS SENSOR
enormous sensor informational collection [12, 18]. NETWORK
This question information sent to the sensor hubs to For structuring WSN, there are a lot of practical and
remove the tactile data. At that point the first characteristics operational issues are there, for example, topology changes,
extricate from gathering data by turning around PCA. power, correspondence interface breakdowns in addition to
Besides PCA lessens the transmitted information by finding memory confinements of a hub by utilizing AI ideal models
a little arrangement of unique readings.PCA techniques take these sorts of operational issues are settled. AI calculations
care of critical thinking by taking a limited quantity of data are gainfully supported to address heterogeneous very much
from a lot of factors in remote sensor systems. structured difficulties in WSN like information accumulation
and grouping, directing, restriction, inquiry preparing just as
2) K-Means Algorithm Medium access control issues.
In K-Means algorithm, K implies calculation partitions A. Accumulation and Clustering of Information
information into divergent bunches. The things which
represent proportionate similar attributes are placed in lone Contingent upon various applications, sensor hubs can
bunch. It makes k number of group and doles out ascend to hundreds to thousands for the wide zone the board.
information focuses to every k bunch dependent on the To coordinate this overloaded sensor, sending the collected
trademark that gave. The mean qualities specifically group is information directly to the sink hub is unsystematic and
the focal point of that bunch. The k means calculation devours the energy of the sensor hubs. To solve these
broadly applying in WSN appropriate of its simple problems, information outperforms the closest aggregator,
accomplishment alongside straight intricacy. Haphazardly which collects all the information from its party territories
select k hubs as centroids for groups and guide every hub to and transfers it to the sink centre. Numerous calculations are
the closest centroid with separation estimations. Figure the introduced for the ideal choice of the bundle head, which is
centroids utilizing current hub participation by estimating the shown immediately [8].The AI techniques can build up the
edge esteem equivalent to the summation of interim among upsides of bunching and information total between hubs in
hubs in addition to their individual centroids [11]. an assortment of conventions in WSN, for example, (i) There
might be some faulty hubs in WSN's that might create
K means calculation utilized for object following and broken or incorrect readings that would hurt the accuracy of
gathering significant data by applying an extra learning the entire system activity. Applying ML, we are evacuating
calculation. K implies calculation is most effortless to damaged sensor hubs. (ii) AI strategy is locked in to
execute and low multifaceted nature and it is likewise creatively settle on the bunch head (CH) and appropriate CH
delicate to anomalies [18]. choice determination broadly shrivel power costs alongside
increment the life expectancy of the system. (iii) AI
C. Re-Inforcememt Machine Learning
calculations can adequately pack information locally at
The reinforcement calculation permits machines, hubs, or group heads by proficiently extricate significant connections
operator hub to find through helping out its environmental in detected crude information which will diminish the
factors. The machine or operator will locate the greatest measure of information to be sent to the sink hub. Methods
choice dependent on which moves to make with the goal that used in information accumulation are:
results will increasingly precise, so as to boost its
presentation. This cooperating learning procedure will  Implementing online Data Compression learning
accomplish for one time or prop up on as condition changes vector quantization.
[16].  Integrative storage of information through k-means
The most solid recognized reinforcement calculation is algorithm.
Q-realizing where specialist repetitively refreshes learning  Large scale network clustering using neural network.
respects bolstered on the activity got on the specific  Accumulation of information using the primary
specification. It will decide its situation in countless hubs by assessment of the modules.
utilizing the Q-learning process. At a specific time every
sensor hub tunes in to the area update message from the  Make a decision how to apply decision trees on a
portable reference point hub. In this convention sensor hub cluster head.
find steering relying on the estimation of messages assessed  Summarize a given data set latency computing.
redemption rate, power limits, message length. Bayesian  Clustering without Role.
systems used to discover the significance of the message and
locate the following bounce of the message in steering. The

Copyright © INDIACom-2021; ISSN 0973-7529; ISBN 978-93-80544-41-0 406

Application of Machine Learning in Wireless Sensor Network

B. Principle of Localization prepared from its former experience to make ideal course
Hub before sending gathered information to closest discovering activities and adjust to the powerful environment
neighbor hubs it must know its own area. Hubs are sent in in addition to steering data of sensor hubs trade with
the huge topographical territory and its makes base station different hubs in the system to locate the best ideal directing
hard to track down hub area to give an accurate area to the way. In this procedure AI assists with diminishing the
client so it's expected of every sensor hub to send the area trouble of steering calculations by taking into account
data alongside the data gathered. Participation among hubs is neighbor hubs data as it were. Every sensor hub self-
significantly significant with the goal that limitation can be governing executes the directing strategy to choose its way
capable by them excluding any individual association. In area. Following are the benefits: (i) Capable to find ideal
WSN such sort of collaboration should be possible between rotting which goes to draw out life expectancies of the
hubs that are situated in certain region codes. The majority system and make vitality sparing of sensor hub. (ii) Pass the
of the restriction calculations typical, signal and stay hubs specifications of computational burden with Routing
help out one another to ascertain the area data of the hub by Efficiency of Services using simple classifiers [10].Below
considering a few angles like constrained vitality utilization, are some WSN routing protocols employed in machine
number, and thickness of hubs and presence of hindrances. learning:
AI in limitation of WSN summarized as follows: In  Application mapping for the optimization of
reconnaissance and item target AI separates observed site Numerous WSN Sinks.
into various destinations to groups, each bunch speaks to a
particular area marker. Utilizing fixed grapple focuses we  Reinforcing information for improving routing.
can take out the scope of estimations. [10]. Some reasonable  System for Decentralized Reconstruction.
tricks for WSN limitation utilizing Machine learning are:  Using self-organizing map for route maintenance.
 Comprehensive reconnaissance of location-aware E. Inquiry Handling
activity.
Inquiry handling and Event discovery are estimated to be
 Using self-organizing map to locate. very much planned necessities of any immense degree sensor
 Regression vector support using localization. organize. In as of late numerous specialists takes
 Localization using Bayesian node. consideration regarding the structure of compelling question
handling and occasion location. The easiest method for
 Localization based on Decision tree. question handling and occasion discovery is making pivotal
 Gaussian processes sensor placements. severe limit esteem yet in most WSN applications inquiry
and occasion elements are troublesome in addition to involve
 Regression of Spatial Gaussian Process.
more than one edge esteem. AI procedures grow profoundly
 Using reinforcement learning, deciding the direction. created inquiry handling and occasion discovery
 Neural network enabled localization. arrangement techniques in WSN. Some WSN Query
handling's which utilized in AI are determined underneath:
C. Medium-jevel Access Control(MAC)
In WSNs, a gigantic measure of sensors teams up to  Discovery of emergency response incidents
capably transmit information. Subsequently, building MAC distributed using decision tree.
conventions for WSNs makes unique questions from  Recognition of an event through Bayesian algorithms.
unmistakable remote systems just as vitality depletion and  Detection of forest fires using neural network.
suspension. ML for the most part adventures to mindfully
choose the obligation pattern of a hub by means of the  Processing queries through neighbors nearest to k.
communicate history of the system implies rest meanwhile VI. NON-FUNCTIONAL CHALLENGES IN WIRELESS
just as wake up when the channel is assessed to be dormant. SENSOR NETWORK
Some of the relevant WSN MAC protocol stratagems using
ML are: In addition to the basic operating or functional behavior
of WSN systems, non-operational aspects are not related to
 Management of the duties cycle through the basic operation of the system.
reinforcement learning.
A. Detection of Privacy and Intervention
 Dynamic layer with MAC.
WSNs sent in an unpleasant domain where humans can't
 MAC based on neural network. appropriately work. In that condition customary networks are
D. Routing additionally neglected to send. Each sensor arranges gathers
the data and sends it to the alluring sink hub or base station.
Steering plays fundamental in conveying the information
Some data is confidential and only accessible via a secure
from the source hub to goal hub. In WSN planning steering,
channel. All information is not open to anyone. An
convention comprises of a scope of model deterrents like as
inspirational debate in WSN's on updating security methods
information inclusion, adaptation to non-critical failure,
is the imperatives of confined assets. Quite a few knowledge
vitality utilization, and versatility. Besides sensor hubs invest
approaches plan to grow unintended, mistaken information
with restricted force assets, minimized memory alongside
by acquainting the device with pretentious understanding [8].
knee-high data transfer capacity. The steering issue is
AI calculations have been locked in to find remote and
basically characterized as advancement the base cost way
pretentious measurements. AI in WSN's gives some income
beginning at the source vertex and arriving at all goal
those follows as: (i) E-learning and avoiding malicious
vertices. With the assistance of ML a sensor network is

Copyright © INDIACom-2021; ISSN 0973-7529; ISBN 978-93-80544-41-0 407

 Proceedings of the 15th INDIACom; INDIACom-2021; IEEE Conference ID: 51348
2021 8 International Conference on “Computing for Sustainable Global Development”, 17th - 19th March, 2021
th

attacks and security flaws. (ii) Forestalling misdirecting and Conference on Computing for Sustainable Global Development
transmission of anomaly spare hubs vitality and (INDIACom), New Delhi, 2016, pp. 1343-1346.
fundamentally increment the life expectancy of the system. [7] F.Q. Khan, S. Musa, G. Tsaramirsis and S.M. Buhari, “SPL features
quantification and selection based on multiple multi-level objectives.”
(3) Removing incorrect inaccurate readings increases Appl. Sci. 2019.
stability of the network. Some relevant WSN Safety [8] G. Tsaramirsis, I. Karamitsos and C. Apostolopoulos, “Smart parking:
stratagems using Machine Learning are: An IoT application for smart city,” in Proc. of the 3rd International
Conference on Computing for Sustainable Global Development
 Distinguish specific sending assaults with help vector (INDIACom), 2016, New Delhi, pp. 1412-1416.
machines. [9] M. P. Nath and S. Sagnika, “Capabilities of chatbots and its
 Approach of the Bayesian network of beliefs for performance enhancements in machine learning,”, in Proc. of the Intl.
Outlier identification. Conf. on Mach. Learn. and Inf. Process. (ICMLIP), 2020, Pune, India,
pp. 183-192.
 Isolated incident discovered by k-nearest neighbors. [10] M. Derbali et al., “Water desalination fault detection using machine
 Assessing intrusions with diagram of self- learning approaches: a comparative study,” IEEE Access, vol. 5, pp.
23266-23275, 2017.
organization.
[11] M. P. Nath, P. Pandey, K. Somu, and P. Amalraj, “Artificial
B. Information Accumulation and Quality of Service intelligence & machine learning: the emerging milestones in software
development”, International Journal of Research and Scientific
QoS places great emphasis on providing both real-time Innovation, vol. 5, issue. 9, pp. 36-44, 2018.
measures and results. Multi-hop data is typically transmitted [12] C. Apostolopoulos, G. Halikias, K. Maroukian, and G. Tsaramirsis,
to the end user as WSN. WSN struggles with the “Facilitating organisational decision making: a change risk assessment
requirements of transfer speed and force that can limit the model case study,” Journal of Modelling in Management(JMM), vol.
measurement of data that can be transferred from a source to 11 issue. 2, pp. 694-721, 2016.
a target hub. Furthermore, distribution of data and [13] M. P. Nath, K. Goyal, J. Prasad, and B. Kallur, “Chat Bot – An Edge
to Customer Insight”, International Journal of Research and Scientific
aggregation of data in WSN's can be inaccurate and Innovation, vol. 5, issue. 5, pp. 29-32, 2018.
ineffective. Machine Learning methods are capable of [14] K. Maroukian, C. Apostolopoulos, G Tsaramirsis, “Extending model
handling many of these while ensuring valuable resources driven engineering aspects to business engineering domain: a model
primarily electricity consumption plus bandwidth [8]. driven business engineering approach”, International Journal of
Information Technology, pp 49-58, 2017.
 Evaluating the veracity and steadiness of the sensor [15] M. P. Nath, S. B. B. Priyadarshini, D. Mishra, and S. Borah, “A
nodes using Gaussian multi-output processes. comprehensive study of contemporary IoT technologies and varied
machine learning (ML) schemes”, in Proc. of the international
 Using reinforcement learning to provide QoS. conference on computing and communication (IC32020), 2020,
Sikkim, India: pp. 623-634.
 Estimating QoS using neural network. [16] S. N. Mohanty, E. L. Lydia, M. Elhoseny, M. M. G. Al Otaibi, and K.
Shankare, “Deep learning with LSTM based distributed data mining
VII. CONCLUSION model for energy efficient wireless sensor networks”, Physical
In this paper, we gave literature overview of latest work Communication, 2020.
done in WSN. We examined different kinds of AI [17] M. P. Nath, S. B. B. Priyadarshini, and D. Mishra, “A comprehensive
study on security in IoT and resolving security threats using machine
approaches particularly utilized in WSN challenges learning (ML)”, in Proc. of the 3rd International Conference on
alongside operational and non-operational issues in WSN. It Intelligent Computing and Advances in Communication (ICAC-2020),
is suggested that more work should be done to extend 2020, Bhubanewar, India.
lightweight conveyed message transient systems, vitality [18] S. B. B. Priyadarshini and S. Panigrahi, “A quadrigeminal scheme
asset the executive‟s issue of WSN. based on event reporting scalar premier selection for camera actuation
in wireless multimedia sensor networks,” J. King Saud Univ. Eng. Sci.
vol. 31, no. 1, pp. 52-60, 2019.
REFERENCES
[19] K. S. Lakshmanaprabu, S. N. Mohanty, S. R. S, S. Krishnamoorthy,
[1] W. Alsaggaf, G. Tsaramirsis, N. Al-Malki, F.Q. Khan, M. Almasry, and J. Uthayakumar, “Online clinical decision support system using
M. Abdulhalim Serafi and A. Almarzuqi, “Association of game events optimal deep neural networks”, Applied Soft Computing Journal, vol.
with facial animations of computer-controlled virtual characters based 81, Issue. 4, pp. 105-116, 2019.
on probabilistic human reaction modeling,”, Appl. Sci. 2020, 10,
[20] S. B. B. Priyadarshini and S. Panigrahi, “Centralised cum Sub-
5636.
centralised Scheme for Multi-event Coverage and Optimum Camera
[2] S. Alraddadi, F. Alqurashi, G. Tsaramirsis, A. Al Luhaybi and S.M. Activation in Wireless Multimedia Sensor Networks”, IET Networks,
Buhari, “Aroma release of olfactory displays based on audio-visual vol. 4, no. 6, pp. 314-328, 2015.
content.”, Appl. Sci. 2019, 9, 4866.
[21] M. P. Nath, S. Sagnika, M. Das, and M. Pandey, “Object recognition
[3] G. Tsaramirsis, S.M. Buhari, M. Basheri and M. Stojmenovic, using cat swarm optimization”, International Journal of Research and
“Navigating virtual environments using leg poses and smartphone Scientific Innovation (International Journal of Research and Scientific
sensors.”, Sensors, 2019. Innovation (IJRSI)), vol. 4, issue. 7, pp. 47-52, 2017.
[4] G. Nagasubramanian, M. Sankayya, F. Al-Turjman and G. [22] M. Nath, J. Muralikrishnan, K. Sundarrajan, and M. Varadarajanna,
Tsaramirsis, “Parkinson data analysis and prediction system using “Continuous integration, delivery, and deployment: a revolutionary
multi-variant stacked auto encoder,” EEE Access, vol. 8, pp. 127004- approach in software development”, International Journal of Research
127013, 2020. and Scientific Innovation (IJRSI), vol. 5, issue. 7, pp. 185-190, 2018.
[5] K. Tsaramirsis, G. Tsaramirsis, F.Q. Khan, A. Ahmad, A.O. Khadidos, [23] M. P. Nath, R. Sridharan, A. Bhargava, and T. Mohammed, “Cloud
A. Khadidos, “More agility to semantic similarities algorithm computing: an overview, benefits, issues & research challenges”,
implementations.”, Int. J. Environ. Res. Public Health, 2020. International Journal of Research and Scientific Innovation (IJRSI),
[6] G. Tsaramirsis, S. M. Buhari, K. O. Al-Shammari, S. Ghazi, M. S. vol. 6, issue. 2, pp. 25-35, 2019.
Nazmudeen and K. Tsaramirsis, “Towards simulation of the classroom
learning experience: virtual reality approach,” 2016 3rd International

Copyright © INDIACom-2021; ISSN 0973-7529; ISBN 978-93-80544-41-0 408

View publication stats