International Journal of Scientific Engineering and Research (IJSER)

ISSN (Online): 2347-3878

Impact Factor (2020): 6.733

Exploring Clustering Algorithms for Internet of

Things: A Review
Deepak Jyoti
Assistant Professor Computer Science & IT, Shanti Devi Arya Mahila College Dinanagar Punjab, India
Email: erdjyoti346 [at]gmail.com

Abstract: In networking, the IOT is a major shift that helps to attach nearly all items on the world. Energy efficient routing will play a
crucial part in the effective implementation of these systems, given the limited existence of smart devices. Clustering techniques
organize the network nodes into groups or clusters, and the cluster head is accountable for the particular assigned node. In the sense of
Wireless Sensor Networks (WSN), clustering techniques have been especially proposed, but their implementation can also answer
related problems Internet of Things (IoT). By assigning big portions of coordination overhead to the cluster head, clustering could
promote energy-efficient routing and configuration management. This article describes a comparison of existing clustering protocols
depend on various parameters such as energy efficiency, scalability, etc of different IoT domains and proposes suitable cluster analysis
for each domain.

Keywords: Internet of things (IOT), Clustering, WSN, Networking.

1. Introduction Figure 1: Internet of Things (IOT) [1]

Table 1: Evolution of IOT Technologies
IoT is an evolving analysis field that integrates a variety of Technology Time Span Description
fields of study. The key IoT concept is to link all devices to Passive identification, wireless
RFID 1999
the Internet, such as home appliances, mobile phones, networks
WSN, Cloud computing, Web2.0, low
vehicles, houses, robots, machines, and so on. The concept WSN 2005
energy communication
is to provide a virtual equivalent for all applications in the Mobile computing, cooperating
real world that detects essential data from the environment Smart Things 2012 operation of objects, connecting
in order supply advanced end-user services [1]. Energy devices
efficiency is one of the major concerns when using these Advanced sensor fusion, faster
devices, as communication and computation on the IOT 2017 wireless connectivity, predictive
restricted device could rapidly discharge its battery capacity analysis
resources. In the scenario Nodes depend on self-organizing
multi-hoping networking methodologies of WSN, which Clustering [3] is the method of seeking a natural connection
could operate in the absence of a base station and between certain particular nodes or the gathering of related
comparable strategies could be implemented for IoT. A main objects. Three key forms of nodes are included in the
research problem is the creation of DR algorithms that could clustering network, including 1) cluster heads, 2) regular
effectively observe paths among mobile nodes. Due to the nodes, and 3) gateway nodes. Every cluster is made up of
obvious rise in computational burden on mobile nodes, one head of the cluster serving as a control unit. The node
dynamic networks do not use traditional algorithms. transmitted signal specifies the length of the cluster in the
Connect performance and network topology could differ single hop cluster and the amount of hops is calculated in the
while a message packet has been routed [2].The multi-hop cluster. The information is retrieved by the cluster
maintenance of a high quality connection therefore involves head from a regular node, which is transferred to the next
regular measurement and upgrading of routing paths. hop. A connection among two clusters with distinct cluster
Clustering is known to be the most efficient way to solve the heads is the action of the base stations. Together, they form
performance issues of ad hoc networks and inevitably the basis of the system, while there is no obligatory
opposes their usage in the sense of IoT due to similar existence of a gateway node [4]. In addition, the base station
difficulties. (BS) acts as a medium of contact among the user and the
sensor network. The volume of transmitted data is
minimized by selecting a cluster head and combining
neighboring nodes. Data aggregation is done to prevent
duplication and coordination burden provided by various
transmissions, Devices with improved computing abilities
such as smart phones may be equivalent to numerous
ubiquitous sensors such as temperature sensors, security
cameras, home appliances, and so on in the IoT paradigm. In
the Iot system t he transition in any node in the network
causes a change in the topology that occurs in overhead
topology improvement messages. Perfect alternatives for
solving certain situations are the clustering algorithms.
Various benefits are provided by clustering. Network

Volume 9 Issue 5, May 2021

www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: SE21527011641 17 of 21
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2020): 6.733
scalability is provided [5]. Routing choices within a cluster interference while high power is used for inter-cluster
and manages the corresponding head of the cluster thereby communication. HEED allows efficiently spread CHs
minimizing the complexity of routing. Clustering saves around the network and offers sound load balancing.
communication bandwidth by testing Cluster Heads (CH)
data, thereby preventing the repeated interchange of data  Energy efficiency Load-Balancing clustering
packets among nodes [4]. The overlap for topology (EELBC): EELBC [10] refers to a class of centralized
maintenance is significantly minimized by sustaining the networks in which sink executes an overall role by attaching
network at the cluster level. The end devices are only sensor nodes to various distance-based gateways and
concerned with the relation to their CHs without being constructing a min-heap on the set of sensors assigned to the
influenced by inter-CH tier changes [6]. A CH increases the gateway. The core of the minimal heap is created by the
network lifespan by using advanced control techniques to gateway and modification is done in a way that the lowest
conserve the power of the computer. In required to preserve packed gateway is at the root.
that nodes are operating in low-power mode when not active
to conserve battery resources, a CH could even applied on  Energy-efficient unequal clustering (EEUC):
different scheduling algorithms. CHs in close proximity to BS tend to remove quicker as they
transfer traffic from remote nodes in order to get data from
The article is organized as obeys: Section II provides the their own cluster is an issue that exists often in WSNs. This
various clustering algorithms. Section III show the related is referred to as the hot-spot issue that EEUC [11] is trying
work of clustering in IOT. In the last, it is concluded in to resolve between clusters by balancing resources. In order
Section IV. to save resources, clusters nearer to the sink are made lower
than distant clusters.
2. Clustering Algorithms
 EECS
Here, explains different clustering methods introduced in the In this scheme [12], nodes broadcast their residual energy to
context of WSN: neighboring nodes to appoint the node with highest residual
energy as the CH for a defined round. EECS expands the
 LEACH and LEACH-C: LEACH [7], which divides creation of clusters in LEACH, which selects CH on the
the network into clusters using a distributed approach. The basis of average distance; by dynamic cluster fitting depend
probability of being a CH is determined by every node and it on the distance from the BS. This tends to address the issue
broadcasts its judgment. In order to achieve the CH, a node that clusters far from the BS need high communication
decides its cluster so that it needs the minimum power than the neighboring clusters, leading to better use of
communication energy. The method produces for the limited storage resources and improved network life.
balance of energy between nodes by using random rotation Congestion in clusters close to BS can lead to early CH
between CHs. In order to fetch the information to the base death.
station, CH nodes serve as routers and also hire data fusion
and aggregation on the cluster sensed data. LEACH ensures  Algorithm for cluster establishment (ACE):The
energy conservation by allowing a node to select a cluster method comprises of 2 phases: new clusters are spawned
focused on minimum communication resources. In the and existing clusters are migrated. Initially, clusters are in an
choice of a cluster head, no overhead is also needed, while un-clustered state and wait for the next phase to adjust the
each node determines individually. Since CH has chosen is configuration and it decides to be a cluster member cluster
probabilistic in LEACH, there could be a circumstance head depending on the current state and surroundings [13].
where a very low energy node is identified as CH through
compromising the entire cluster. In addition, even though  GROUP
nodes may interact directly with CH, LEACH could result in A hybrid method that operates in four steps, i.e. creation of
one-hop intra- and inter-cluster configuration. In LEACH-C the CG, query forwarding, forwarding data as well as
[8], the setup and steady step is identical to LEACH, but the restoration of loss. One sink situated in the vicinity of the
choice of CH is focused on energy knowledge and load network is chosen as the primary sink and the sink elects one
balancing. In order to absorb more resources and minimize node as CH is focused on residual energy and distance
usability, the Base-Station is accountable for the overall value. Unicast and broadcast correspondence are used for
network configuration and cluster design. the forwarding of requests. This method finds its use in
large-scale WSN [14].
 HEED
It [9] is a multi-hop clustering method that selects CH for  PEACH
intra-cluster communication depend on residual battery and Due to the high workload of cluster creation, clustering
energy available. In a way close to other clustering algorithms mainly suffer. PEACH aims to resolve this issue
techniques, an initial version of CHs is probabilistically in way to minimize the power cost as well as to enhance the
selected depend on the residual energy. The value of lifespan of the network., PEACH decreases over-head
communication in the intra-cluster serves as an implicit clustering, by using wireless channel overhearing features to
variable to provide an indication of the level or proximity of allow adaptive multilevel clustering.
a node to neighboring nodes. This makes it possible for a
node to determine why not to join a specific cluster. A  Passive clustering (PC): PC [15] is a clustering
cluster that uses low power rates for communication enables technique that only operates when the clusters are actively
spatial reuse because of its long range because of reduced involved in communication in an on-demand manner,
Volume 9 Issue 5, May 2021
www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: SE21527011641 18 of 21
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2020): 6.733
forming clusters. The conditions of various nodes in the of these protocols needs to be carried out in order to use
cluster between the communicating nodes are piggybacked these methods in IoT systems. A comparison study of the
and ex-changed. When the nodes receive a packet from above mentioned methods is provided in Table 2 on the
them, they refresh the database of CH and Gateway (GW) foundation of variables such as energy efficiency, data
nodes. In a distributed manner, the PC uses Gateway aggregation, mobility, form of network, etc. Such research
Selection Heuristics (GSH) to decide how node states adjust will help determine which strategies for a specific IoT
in just such a way that the minimum level of GW nodes is specification are most likely possible. Since IoT faces WSN-
used to maintain connectivity. The method used is easy to like problems such as energy efficiency, usability and data
applied and reduces overhead included in the creation of aggregation, suitable evolutionary techniques could be used
clusters. in IoT to solve significant difficulties. Even so, concerns
such as accessibility and the form of network, which may
To fix specific concerns facing WSN, the formulas not be very important in the sense of WSN, reflect important
mentioned above have been suggested. A comparison study issues for different IoT applications.

Table 2: Comparsion of clustering algorithms [7-15]

Data
Algorithm Control Manner Energy Efficiency Scalability CH Selection CH Rotation Mobility Network Type
Aggregation
LEACH Distributed Less Never Never Probabilistic Agreed Agreed Homogeneous
LEACH-C Centralized Less Never Never Probabilistic Agreed Never Homogeneous
EELBC Centralized Average Never Never Probabilistic Agreed Never Homogeneous
EEDC Centralized High Never Never Probabilistic Agreed Never Homogeneous
LCA Distributed Less Never Never Id-Based Agreed Agreed Homogeneous
PEACH Distributed Medium Agreed Agreed Probabilistic Agreed Agreed Homogeneous
GROUP Mixed High Never Never Weight Based Never Never Homogeneous
HEED Distributed Medium Agreed Agreed Probabilistic Agreed Agreed Homogeneous
EEHCA Distributed High Agreed Agreed Probabilistic Agreed Never Heterogeneous
EACLE Distributed Average Agreed Never Connectivity Agreed Never Heterogeneous
EECS Distributed Medium Never Agreed Weight Based Agreed Never Homogeneous
EEUC Distributed High Agreed Agreed Probabilistic Agreed Never Homogeneous
S-WEB Distributed High Never Agreed Probabilistic Never Agreed Homogeneous
MCEEC Centralized Agreed Yes Never Probabilistic Agreed Never Heterogeneous

3. Literature Survey that the implementation of a clustering method minimize

power usage [18].
Liu et al., (2019) Introduces a DRL method for DNC with
edge servers in IoT systems. The goal is to satisfy the Aboalnaser et al., (2019) It examines energy usage in IoT
specifications of either the IoT system and EC via improving networks as well as suggests a latest TA method depend on
the group of data clustering. The outcomes demonstrate that clustering solutions to manage communication energy
the suggested DQN design could produce strong among nodes. The engineered system tries to optimize the
performance as contrasted to the static benchmark solution energy between all SD or continue the lifespan of the
[16]. network. In means of energy efficiency and lifespan of
nodes, the efficacy of the suggested method is assessed or
Bensaid et al., (2020)A new clustering method depend on the outcomes suggest better productivity of the suggested
FCM is being proposed for WSN-focused IoT apps. The method [19].
algorithm considers an FCM approach to shape the clusters
and minimize the total consumed energy in each cluster to Behera et al., (2019) The emphasis is on an effective CH
find the optimal Cluster Head (CH) for every transmission election approach that vibrates the location of the cluster
round. A comparative with the LEACH algorithm is head between the greater energy nodes related to others.
provided to evaluate the efficiency as well as to certify the Initial energy, residual energy and the optimum conditions
suggested approach, Findings demonstrate the suggested of cluster heads are considered in the method to pick the
FCM approach increases the lifespan of system that via next category of system cluster heads appropriate for IoT
improving the remaining energy including 50 percent [17]. applications, like weather forecasting, smart buildings and
networks. The findings demonstrates that the improved
Ganeshan et al., (2019)To efficiently make use of the design plays best as compared to the Leach algorithm
available energy, a clustering scheme is introduced. The through raising performance including 60%, lifespan by
suggested approach would position an additional IoT system 66%, residual energy including 64% [20].
that functions as CH in the user’s structure with strong
loading power. The data would be distributed to CH by the Fan et al., (2019) For the complex strategic situation of NB
cluster participants and the CH passes the aggregated IoT, in which the system keyboards are grouped depending
information to the CP. The 2 efficiency metrics, such as the on the time from the BS and TTI, as well as assigned recycle
amount of data transmissions and the maximum set of preamble software to increase the ability of network access,
information transmissions to the cloud network, are utilized an improvement access approach is applied to get high QoS.
for experimentation. The findings outcomes have confirmed To test the modified entrance approach depend on CRPD,
the capability is modeled and evaluated (ERACRPA). The

Volume 9 Issue 5, May 2021

www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: SE21527011641 19 of 21
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2020): 6.733
test results reveal that the ERA-CRPA could efficiently systems in future networks”, Sensors, vol. 17, no. 7,
minimize the risk of collision with the random access pp. 1574, 2017.
preamble, thus enhance access productivity and enhance the [4] S. Bhandari, S. K. Sharma and X. Wang, “Cloud-
NB-IoT network's equivalent capacity [21]. assisted device clustering for lifetime prolongation in
wireless IoT networks, ” in IEEE 30th Canadian
Xiu et al., (2018) Suggest a PADC for confidentiality and Conference on Electrical and Computer Engineering
accessibility focused on the k-means method and original (CCECE), pp. 1-4, 2017.
centre values as well as the process of distance measurement [5] A. Abbasi, M. Younis, “A survey on clustering
from another locations to the centre value. In addition, algorithms for wireless sensor networks, ” Computer
PADC efforts to minimize the deviation impact by Communications, Vol. 30, pp. 2826-2841, 2007.
identifying deviations through the clustering phase. Security [6] D. Bandyopadhyay and J. Sen, “Internet of things:
review shows that device fulfils the aim of differential Applications and challenges in technology and
security as well as avoids leakage of confidential data. In the standardization”, Wirel. Pers. Commun., vol. 58, no. 1,
meantime, performance evaluation indicates that our method pp. 49–69, 2011.
increases the accessibility of clustering outcomes at the [7] J. Duan, D. Gao, D. Yang, C. H. Foh, and H.-H. Chen,
similar phase of privacy contrasted to the current differential “An energyaware trust derivation scheme with game
privacy approaches, indicating that the suggested PADC theoretic approach in wireless sensor networks for IoT
system outperforms for smart power facility in IoT [22]. applications”, IEEE Internet Things J., vol. 1, no. 1,
pp. 58–69, 2014.
Miead et al., (2019) It focused on directive transmission in [8] H. Junping, J. Yuhui, and D. Liang, “A time-based
urban outdoor situations and illustrates the value of correctly cluster-head selection algorithm for LEACH, ” in
estimating the set of mm Wave radio channel clusters Computers and Communications ISCC 2008, IEEE
modeled with a ray-tracer software package. Studies show Symposium on, pp. 1172–1176, 2008.
that clustering is a tough job, since one or a mixture of more [9] K. Maraiya, K. Kant, and N. Gupta, “Efficient cluster
CVIs does not always offer the optimum set of clusters. Use head selection scheme for data aggregation in wireless
score fusion techniques to observe the optimum partitioning sensor network, ” Int. J. Comput. Appl., vol. 23, no. 9,
for the k-means method depend on the multipath rays' power pp. 10–18, 2011.
and time of arrival or based on their angle of incidence. [10] B. Singh and D. K. Lobiyal, “A novel energy-aware
When the k-power-means technique is utilized, the intensity cluster head selection based on particle swarm
of each entering ray is the most critical clustering element, optimization for wireless sensor networks, ” Human-
enabling the prevailing routes collected to drag the others Centric Comput. Inf. Sci., vol. 2, no. 1, p. 13, 2012.
into them to build a cluster. The collection of clusters is [11] P. C. S. Rao, P. K. Jana, and H. Banka, “A particle
lower and it is quicker to carry a conclusion on CVIs or swarm optimization based energy efficient cluster head
score fusion factors [23]. selection algorithm for wireless sensor networks, ”
Wirel. networks, vol. 23, no. 7, 2017.
4. Conclusion [12] H. Farman et al., “Multi-criteria based zone head
selection in Internet of Things based wireless sensor
In this article, which examined numerous clustering networks, ” Futur. Gener. Comput. Syst., 2018.
techniques commonly utilized in the context of WSN, [13] S. Kallam, R. B. Madda, C.-Y. Chen, R. Patan, and D.
Techniques were categorized on the foundation of variables Cheelu, “Low energy-aware communication process in
including energy efficiency, mobility, data aggregation as IoT using the green computing approach, ” IET
well as network form. Clustering algorithms provide a Networks, vol. 7, no. 4, pp. 258–264, 2017.
promising solution to topology management, energy needs, [14] J. Jin, J. Gubbi, S. Marusic, and M. Palaniswami, “An
and data aggregation problems of IoT. Even so, problems information framework for creating a smart city
such as fault tolerance, service efficiency and safe through internet of things, ” IEEE Internet Things J.,
communication have to be addressed. With the aid of this vol. 1, no. 2, pp. 112–121, 2014.
document, different parameters will enable the best [15] T. M. Behera, U. C. Samal, and S. K. Mohapatra,
clustering method to be found. “Energy Efficient Modified LEACH Protocol for IoT
Application, ” IET Wirel. Sens. Syst., vol. 8, no. 5, pp.
223 – 228, 2018.
References [16] Liu, Q., Cheng, L., Ozcelebi, T., Murphy, J., &
Lukkien, J., “ Deep Reinforcement Learning for IoT
[1] Al-Fuqaha, A., Guizani, M., Mohammadi, M.,
Network Dynamic Clustering in Edge Computing”,
Aledhari, M.and Ayyash, M., “Internet of Things: A
19th IEEE/ACM International Symposium on Cluster,
Survey on Enabling Technologies, Protocols, and
Cloud and Grid Computing (CCGRID), 2019.
Applications, ” IEEE Communications Surveys &
[17] Bensaid, R., Ben Said, M., & Boujemaa, H., “Fuzzy C-
Tutorials, Vol. 17 No. 4, pp. 2347-2376, 2015.
Means based Clustering Algorithm in WSNs for IoT
[2] Yao, X., Wang, J., Shen, M., Kong, H., & Ning, H.,
Applications”, International Wireless Communications
“An Improved Clustering Algorithm and Its
and Mobile Computing (IWCMC), 2020.
Application in IoT Data Analysis” Computer
[18] M.Ganesan, Dr.N.Sivakumar, “An energy efficient IoT
Networks, 2019.
based Healthcare System based on clustering
[3] Xu, Lina, Gregory MP OHare, and Rem Collier, “A
technique”, Third International Conference on
smart and balanced energy efficient multihop
clustering algorithm(smart-beem) for mimo iot
Volume 9 Issue 5, May 2021
www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: SE21527011641 20 of 21
 International Journal of Scientific Engineering and Research (IJSER)
ISSN (Online): 2347-3878
Impact Factor (2020): 6.733
Electronics Communication and Aerospace
Technology ICECA, 2019.
[19] Aboalnaser, S. A., “ Energy–Aware Task Allocation
Algorithm Based on Transitive Cluster-Head Selection
for IoT Networks”, 12th International Conference on
Developments in eSystems Engineering (DeSE), 2019.
[20] Behera, T. M., Mohapatra, S. K., Samal, U. C., Khan,
M. S., Daneshmand, M., & Gandomi, A. H., “Residual
Energy Based Cluster-head Selection in WSNs for IoT
Application”, IEEE Internet of Things Journal, 2019.
[21] Fan Wu, Sijia Huang, “An Enhanced Random Access
Algorithm Based on the Clustering-Reuse Preamble
Allocation in NB-IoT System”, IEEE, Vol. 7, 2019.
[22] Xiong, J., Ren, J., Chen, L., Yao, Z., Lin, M., Wu, D.,
& Niu, B., “ Enhancing privacy and availability for
data clustering in intelligent electrical service of IoT”,
IEEE Internet of Things Journal, 2018.
[23] Miead Tehrani Moayyed, Stefano Basagni, “Clustering
Algorithms and Validation Indices for mmWave Radio
Multipath Propagation”, IEEE, 2019.

Volume 9 Issue 5, May 2021

www.ijser.in
Licensed Under Creative Commons Attribution CC BY
Paper ID: SE21527011641 21 of 21