Proceedings of the Fourth International Conference on Sentiment Analysis and Deep Learning (ICSADL-2025)

IEEE Xplore Part Number: CFP25UU5-ART; ISBN: 979-8-3315-2392-3

Cluster Head Selection Algorithm using

Machine Learning
[Link] Dr. E. Gurumoorthi [Link]
Associate Professor Associate professor Associate Professor
School of Computer Applications School of Engineering Dept of AI & DS
Dayananda Sagar University Malla Reddy University, Chennai Institute of Technology college,
Chennai
Bangalore egurumoorthiphd@[Link] babum@[Link]
2025 4th International Conference on Sentiment Analysis and Deep Learning (ICSADL) | 979-8-3315-2392-3/25/$31.00 ©2025 IEEE | DOI: 10.1109/ICSADL65848.2025.10933157

nidhyaphd@[Link]

A.C. Charumathi [Link] Naved Dr Balajee Maram,

Assistant Professor Associate Professor Professor & Dean(PhD Program),
Dept of Information Technology Dept of Business Analytics School of Computer Science and Artificial
Intelligence,
Sona College of technology Jaipuria Institute of SR University, Ananthasagar, Hasanparthy,
Management,Noida Warangal,
Charusps46@[Link] Mohdnaved@[Link] [Link]@[Link].

Abstract: Wireless Sensor Networks (WSNs) find extensive prolong network lifetime, and enhance data aggregation
applications in environmental monitoring, healthcare, and smart efficiency [1].
cities. Energy efficiency, however, continues to be a significant
challenge with the limited lifetime of sensor node batteries. The Conventionally, CH selection is based on heuristic methods that
conventional heuristic-based cluster head (CH) selection techniques select CHs using static parameters such as node energy or
tend not to adapt dynamically to network changes, resulting in distance from the base station. These methods can be inadequate,
inefficient energy utilization and decreased network lifetime. This as they may not account for the dynamic and complex
research discusses Machine Learning (ML)-oriented CH selection
characteristics of WSNs, resulting in poor performance. Recent
approaches to improve energy efficiency, network lifetime, and data
fusion. Using supervised and unsupervised learning methods, ML
studies have investigated Machine Learning (ML) methods for
algorithms are able to learn optimal CHs dynamically based on CH selection to alleviate these issues. ML may also be utilizing a
residual energy, network structure, and data traffic. The comparative number of properties of the node, such as the residual energy,
analysis show that ML-based CH selection enhances network centrality, data traffic volume, and distance between nodes and
stability by 95% and lowers energy consumption by 50% compared the base station in order to dynamically choose CH nodes [2]. By
to traditional methods. The research points to the promise of ML in utilizing supervised and unsupervised learning algorithms, the
WSN performance optimization and opens the doors to intelligent, ML-based CH selection algorithms would be capable of
adaptive clustering technologies. identifying the changes in a network environment and also
Keywords: Wireless Sensor Networks (WSN), Cluster Head Selection, enhance energy relocation while optimizing the load of the node
Machine Learning, Energy Efficiency, Network Lifetime, Supervised as well.
Learning, Unsupervised Learning, Adaptive Algorithm.
Classification models like decision trees and support vector
I. INTRODUCTION machines are able to predict CHs from labeled training data, as
per the research. These models are capable of learning the
Wireless Sensor Networks (WSNs) are an essential part of optimal selection of CH and take into account fluxing conditions
numerous contemporary applications, such as environmental in the network [3]. Other types of categories under the category
monitoring, healthcare, military applications, and smart cities. of unsupervised learning mechanisms are clustering that provides
The networks consist of spatially distributed sensor nodes that a chance for the sensor nodes to be clustered on the basis of the
exchange information wirelessly to obtain and forward data. One node's attributes without necessarily being allocated beforehand
of the most important challenges of WSNs is energy management to any labels. For instance, the K-means clustering is applied to
since the sensor nodes usually run on bounded energy using cluster the nodes and, simultaneously, choose the CHs on the
batteries and are deployed where battery replacement is not basis of the remaining energy and connectivity required [4].
feasible. Successful clustering and efficient cluster head (CH) There are some uncertain factors of ML application in CH
selection are key techniques to control energy consumption, selection such as energy saving, network longevity, and load
distribution. A novel algorithm for Machine Learning-based CH

979-8-3315-2392-3/25/$31.00 ©2025 IEEE 1452

Authorized licensed use limited to: Pusan National University Library. Downloaded on September 25,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fourth International Conference on Sentiment Analysis and Deep Learning (ICSADL-2025)
IEEE Xplore Part Number: CFP25UU5-ART; ISBN: 979-8-3315-2392-3

selection that leans towards making dynamic CH selection TABLE-1 COMPARISON OF ALGORITHMS FOR CLUSTER
relying on real-time network conditions and node characteristics SELECTION
is proposed in this paper. We demonstrate, through simulation,
that ML-based CH selection is 95% more stable and uses 50% Algorithm Application
Merits Demerits
less energy compared to traditional heuristic approaches. This Type s
work also establishes the capability of ML to provide intelligent, Simple;
Limited General
Randomize reduces
adaptive, and efficient WSN based on the design selected. adaptability; WSN
d direct
uneven applications
II. ENERGY-EFFICIENT CLUSTERING (Heuristic) communicati
energy ; periodic
ALGORITHMS [5] on with base
distribution. monitoring.
station.
Chaurasia et al. [5] introduced Mocraw, which is a meta- Balances Static
Static
heuristic-optimized CH selection-based routing technique. The Energy- energy selection
networks;
strategy improves network life by meta-heuristically optimally Based usage; criteria;
applications
selecting the CH in response to residual energy and network Selection improves limited real-
with stable
topological considerations. Likewise, Behera et al. [6] [6] network time
data rates.
conceived a residual energy-based CH selection method lifetime. adaptation.
suitable for IoT use cases. Their system aims at attaining Chain
Minimizes Energy-
formation
maximum energy efficiency through energy remaining-based Chain- transmission sensitive
overhead; less
selection of CH, thereby minimizing congestion in the network Based distance; WSNs;
effective for
and increasing its operational lifetime. Yadav and Poongodi [7] Clustering reduces environmen
dynamic
proposed a new chain-based clustering technique for green [7] energy tal
environments
communication in WSNs. Their scheme reduces energy usage. monitoring.
.
expenditure by facilitating effective data transmission among Balances Application
clusters. In addition, Li et al. [8] introduced an unequal energy- Increased
energy s with non-
efficient clustering mechanism, which balances energy usage Unequal overhead;
among CHs; uniform
Clustering complex
among sensor nodes through dynamic cluster size adjustment, mitigates node
[8] cluster
avoiding early energy exhaustion of CHs in high-traffic areas. "hot-spot" distribution
formation.
problem. .
Recent developments in machine learning (ML) have Adapts to
significantly enhanced WSN efficiency. Amutha et al. [9] have Requires Large-scale
changing
labeled data; WSNs;
examined the consistency of clustering protocols with Supervised network
higher predictive
supervised ML classification methods. Their solution advances ML [9] conditions;
computationa data
CH selection with learning from past network behaviors to accurate CH
l demand. gathering.
achieve enhanced cluster stability and power efficiency. selection.
Sivapriya et al. [12] have continued to examine hybrid cluster May suffer
Real-time
establishment within WSN with ML approaches that guarantee from Dynamic
adaptability;
optimized distribution of clusters as well as power-efficient CH convergence WSNs;
Unsupervis effective for
selection. Tadros et al. [10] proposed an unsupervised learning- issues; mixed-
ed ML-[10] diverse
increased activity
inspired clustering method for environmental pollution network
computationa monitoring.
monitoring. The model identifies optimal CHs independently topologies.
l overhead.
without pre-trained data, ensuring that the model is flexible to
Learns from
vary in WSN environments. Kaur and Aulakh [11] applied Computation WSNs with
network
reinforcement learning (RL) to optimize energy-aware CH Reinforcem ally intensive; variable
changes;
selection. Their RL-based method learns optimal clustering ent training may conditions;
long-term
patterns continuously by interacting with the network and Learning take adaptive
energy
(RL) [11] significant applications
results in long-term energy saving. Hybrid approaches that are optimization
time. .
the combination of several clustering approaches have been .
proposed for their resistance and flexibility. Saravana Kumar et Balances
High High-
al. [13] introduced a hybrid model based on machine learning efficiency
complexity; density
to provide energy-efficient and secure data transmission in Hybrid ML and
may require WSNs;
WSNs. Their approach merges ML algorithms and Approaches adaptability;
substantial real-time
cryptographic methods for improved data protection without [12, 13] suitable for
processing adaptive
compromising on energy efficiency. complex
resources. networks.
scenarios.

III. PERFORMANCE COMPARISON

979-8-3315-2392-3/25/$31.00 ©2025 IEEE 1453

Authorized licensed use limited to: Pusan National University Library. Downloaded on September 25,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fourth International Conference on Sentiment Analysis and Deep Learning (ICSADL-2025)
IEEE Xplore Part Number: CFP25UU5-ART; ISBN: 979-8-3315-2392-3

To analyze the efficiency of cluster head (CH) selection Figure-1 shows the efficiency score of the different algorithms
algorithms in WSNs, we can formulate an efficiency equation considered in Section II for the selection of cluster heads in
that considers key factors affecting the overall performance. Wireless Sensor Networks (WSNs). Measured scores consider
These factors often include energy consumption, latency, energy efficiency index, the latency of data arrival, the
network lifetime, adaptability, and computational complexity. durability of its network topology, adaptability and the
Each factor can be weighted based on the network's specific computational load. By considering energy consumption,
priorities. latency, network lifetime, adaptability, and computational
complexity parameters, algorithm such as supervised ML,
unsupervised ML and Hybrid ML algorithms are most efficient
Let: for cluster head selection than other algorithms. Figure-1
highlights the performance of each algorithm, helping in
E = Energy efficiency (higher values indicate better energy identifying the most efficient choice based on the given criteria.
conservation)
IV. CONCLUSION
L = Latency (higher values indicate lower latency)
This study has explored the application of Machine Learning
N = Network lifetime (higher values indicate longer network (ML) techniques to enhance Cluster Head Selection (CHS)
operation) algorithms in Wireless Sensor Networks (WSNs). Through the
analysis, it was evident that ML-based CHS algorithms offer
A = Adaptability (higher values indicate better adaptability to
significant advantages over traditional methods by adapting to
network changes)
dynamic network conditions. By incorporating factors like node
C = Computational complexity (lower values indicate lower residual energy, distance, and network topology, these ML-based
complexity, hence higher efficiency) approaches enable more balanced clusters and enhanced network
longevity. In future work, considering the real-time data and
• 𝑤𝐸 , 𝑤𝐿 , 𝑤𝑁 , 𝑤𝐴 , and 𝑤𝐶 represent the weights for each factor environmental factors could also contribute to the selection
based on network requirements, with 𝑤𝐸 + 𝑤𝐿 + 𝑤𝑁 + 𝑤𝐴 + 𝑤𝑐 process for other hierarchical and complex WSN applications.
= 1.
REFERENCES
Efficiency Equation
[1] Mody, Samkit, Sulalah Mirkar, Rutwik Ghag, and Priyanka Kotecha.
To overall efficiency score (Eff) for a CH selection algorithm "Cluster head selection algorithm for wireless sensor networks using
can be calculated as: machine learning." In 2021 International Conference on Computational
Performance Evaluation (ComPE), pp. 445-450. IEEE, 2021.
Eff = (𝑤𝐸 × 𝐸) + (𝑤𝐿 × 𝐿) + (𝑤𝑁 × 𝑁) + (𝑤𝐴 × [2] Sanhaji, Farah, Hassan Satori, and Khalis Satori. "Cluster head selection
based on neural networks in wireless sensor networks." In 2019
𝐴) − (𝑤𝐶 × 𝐶)-------(1)
International Conference on Wireless Technologies, Embedded and
Intelligent Systems (WITS), pp. 1-5. IEEE, 2019.
where: [3] Raj, Bryan, Ismail Ahmedy, Mohd Yamani Idna Idris, and Rafidah Md.
Noor. "A survey on cluster head selection and cluster formation methods
• 𝑤𝐸 × 𝐸 : Weighted energy efficiency. in wireless sensor networks." Wireless Communications and Mobile
• 𝑤𝐿 × 𝐿 : Weighted latency with a higher score Computing 2022, no. 1 (2022): 5322649.
representing lower latency. [4] Prajapati, Hardik, and Rutvij Joshi. "Cluster head selection algorithm based
on deep learning and LEACH for wireless sensor network." In AIP
• 𝑤𝑁 × 𝑁 : Weighted network lifetime. Conference Proceedings, vol. 2855, no. 1. AIP Publishing, 2023.
• 𝑤𝐴 × 𝐴 : Adaptation with ability factors or where the [5] Chaurasia, Soni, Kamal Kumar, and Neeraj Kumar. "Mocraw: A meta-
condition of the network or other environment differs. heuristic optimized cluster head selection based routing algorithm for
wsns." Ad Hoc Networks 141 (2023): 103079.
[6] Behera, Trupti Mayee, Sushanta Kumar Mohapatra, Umesh Chandra
Samal, Mohammad S. Khan, Mahmoud Daneshmand, and Amir H.
Gandomi. "Residual energy-based cluster-head selection in WSNs for IoT
application." IEEE Internet of Things Journal 6, no. 3 (2019): 5132-5139.
[7] Yadav, Suman Avdhesh, and T. Poongodi. "A novel chain‐based clustering
for green communication in wireless sensor network." International
Journal of Communication Systems 36, no. 13 (2023): e5523.
[8] Li, Chengfa, Mao Ye, Guihai Chen, and Jie Wu. "An energy-efficient
unequal clustering mechanism for wireless sensor networks." In IEEE
International Conference on Mobile Adhoc and Sensor Systems
Conference, 2005., pp. 8-pp. IEEE, 2005.
[9] Amutha, J., Sandeep Sharma, and Sanjay Kumar Sharma. "Reliability of
clustering algorithm in wireless sensor networks using supervised machine
learning classification approaches." In AIP Conference Proceedings, vol.
2853, no. 1. AIP Publishing, 2024.
Figure-1 Efficiency Score of Cluster head selection algorithm

979-8-3315-2392-3/25/$31.00 ©2025 IEEE 1454

Authorized licensed use limited to: Pusan National University Library. Downloaded on September 25,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fourth International Conference on Sentiment Analysis and Deep Learning (ICSADL-2025)
IEEE Xplore Part Number: CFP25UU5-ART; ISBN: 979-8-3315-2392-3

[10] Tadros, Catherine Nayer, Nader Shehata, and Bassem Mokhtar. [12] Sivapriya, N., R. Mohandas, and Mr P. Kiran Kumar. "Hybrid cluster
"Unsupervised Learning-Based WSN Clustering for Efficient formation in wireless sensor networks using machine learning
Environmental Pollution Monitoring." Sensors 23, no. 12 (2023): 5733. techniques." Computer Integrated Manufacturing Systems 28, no. 11
[11] Kaur, Navpreet, and Inderdeep Kaur Aulakh. "An Energy Efficient (2022): 1023-1045.
Reinforcement Learning Based Clustering Approach for Wireless Sensor [13] Saravana Kumar, N. M., E. Suryaprabha, and K. Hariprasath. "Machine
Network." EAI Endorsed Transactions on Scalable Information Systems 8, learning based hybrid model for energy efficient secured transmission in
no. 31 (2021). wireless sensor networks." Journal of Ambient Intelligence and
Humanized Computing (2022): 1-16.

979-8-3315-2392-3/25/$31.00 ©2025 IEEE 1455

Authorized licensed use limited to: Pusan National University Library. Downloaded on September 25,2025 at [Link] UTC from IEEE Xplore. Restrictions apply.