International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.

1/2, April 2024

IMPLEMENTING MACHINE LEARNING

ALGORITHMS FOR PREDICTIVE NETWORK
MAINTENANCE IN 5G AND BEYOND NETWORKS
Yamini Kannan and Dharika Kapil

Department of Computer Science, New York University, New York, USA

Software Engineer, MarketAxess, New York, USA

ABSTRACT
With the evolution of fifth generation (5G) network technologies, network maintenance strategies have
become increasingly complex, necessitating the use of predictive analysis enabled by Machine Learning
(ML) algorithms. This paper emphasizes exploring how ML algorithms can further enhance predictive
maintenance in 5G and future networks. It reviews the current literature on this interdisciplinary topic,
identifying key ML models such as Decision Trees, Neural Networks, and Support Vector Machines, and
discussing their benefits and limitations. Special attention is given to the methodologies in applying these
models, handling of data stages, and the training process. Major challenges in implementing ML in the
context of network maintenance, such as data privacy, data gathering, model training, and generalizability,
are discussed. Furthermore, the research aims to go beyond predicting maintenance needs to introduce a
proactive approach in improving overall network performance and pre-empting potential issues based on
ML predictions. The paper also discusses possible future trends including advancements in ML algorithms,
Automated Machine Learning (AutoML), Explainable AI, and others. The objective is to provide a
comprehensive understanding of the current ML-based predictive maintenance field and outline
possibilities for future research. The study finds that the application of ML algorithms continues to show
promise in transforming the landscape of network management by improving predictive maintenance and
proactive performance enhancement strategies. It remains a challenging yet important area in the context
of 5G networks.

KEYWORDS
5G, Predictive Maintenance, Machine Learning, Decision Trees, Neural Networks, Support Vector
Machines, Traffic Classification, Data Pre-processing, Model Training, Proactive Performance, IoT
Networks, SDN, AutoML, Explainable AI, Domain-Specific Learning

1. INTRODUCTION
With the advent of the fifth-generation (5G) network technology, communication systems
worldwide are undergoing a significant transformation. As the architecture of these networks
becomes more complex, conventional maintenance strategies are increasingly less effective,
leading to serious reliability concerns and potential service disruptions. This raises the pressing
need to develop advanced maintenance strategies that can proactively address potential issues
before leading to network downtime. To tackle this challenge, some have turned to predictive
maintenance strategies underpinned by Machine Learning (ML) algorithms.

Predictive maintenance refers to the use of data-driven methodologies to predict the state of
network components, intending to anticipate and prevent failures. This is a notable shift from the
traditional reactive maintenance strategies, where issues are tackled as they arise. Predictive

DOI:10.5121/ijwmn.2024.16201 1
 International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.1/2, April 2024
maintenance offers a proactive strategy, aiming to prevent potential problems before they
significantly impact the network performance.

Machine learning, a subset of artificial intelligence, provides a promising solution to this issue.
Machine learning algorithms can process vast amounts of data and identify patterns too subtle or
complex for human operators or traditional automation tools. By applying these algorithms to
network data, it becomes possible to predict potential faults or areas of inefficiency in the
network infrastructure, thereby enabling more accurate forecasting, efficient troubleshooting and
improvements in network reliability and resilience.

The use of machine learning algorithms in creating predictive models for network maintenance is
a growing field that requires comprehensive evaluation. This research paper, therefore, aims to
explore how machine learning algorithms can be leveraged for predictive maintenance in 5G
networks and provide an assessment of their potential and performance. We pose the main
research question as: "How can Machine Learning algorithms be utilized to improve the
Predictive Maintenance strategies in 5G and beyond networks?"

This study examines the existing machine learning algorithms currently utilized, evaluates their
efficacy in the sphere of predictive maintenance, and explores potential areas of advancement or
improvement. We aim to contribute to the existing body of knowledge by providing a
comprehensive analysis of ML-based predictive maintenance techniques in the context of 5G
networks.

2. BACKGROUND AND LITERATURE REVIEW

The development and expansion of 5G networks have spurred a wave of advancements in
network maintenance strategies. As these networks grow more complex and the demand for
uninterrupted services increases, predictive maintenance supported by machine learning
algorithms is receiving heightened attention in the research community.

One pivotal example of machine learning application in network management was proposed by
Kim et al. (2018) [1]. In their study, they introduced an autonomous link-based cell scheduling
scheme known as ALICE for Time-Slotted Channel Hopping (TSCH) in low-power lossy
networks (LLNs) [1]. ALICE is an interesting contribution since it leverages local information
and interacts closely with the routing layer to effectively manage and schedule communication
cells, thereby reducing communication overhead and enhancing network reliability.

Interestingly, Kim's work builds on previous research, namely Orchestra, introduced by

Duquennoy et al. (2015) [2]. Orchestra proposed an autonomous scheduling mechanism born out
of experiences from the emergence of IPv6 Routing Protocol for Low-Power and Lossy
Networks (RPL) and its shortcomings [2]. Their focus was on creating a robust mesh network
using TSCH showcasing its strengths in achieving reliable low-power networking through
scheduling and channel hopping.

Jung et al. (2017) also focused on network maintenance with a specific interest in applying
machine learning algorithms for analyzing vibrations in the Internet of Things (IoT) enabled
predictive maintenance [3]. The research is particularly relevant as it demonstrates the real-world
application and efficiency of machine learning in predictive maintenance.

As per Palattella et al. (2012), they explored Traffic-Aware Scheduling Algorithm (TASA) for
reliable low-power multi-hop IEEE 802.15.4e networks [4]. Their work laid the foundation for

2
 International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.1/2, April 2024
the implementation of machine learning algorithms in network traffic maintenance, which
significantly contributes to the realization of predictive maintenance in 5G networks.

However, the application of machine learning in predictive maintenance is still a growing field
with novel approaches and improvements to be discovered. The existing studies have
predominantly focused on schematic improvements and algorithmic enhancements, leaving room
for exploring how machine learning algorithms can optimize the predictability, reliability, and
overall functional efficiency of 5G maintenance strategies. This research gap indicates the
potential for our research and the relevance of the field. This work aims to enrich the body of
knowledge by providing an in-depth evaluation of utilizing machine learning algorithms in 5G
predictive network maintenance.

3. SURVEY OF ML ALGORITHM APPLICATIONS

During our research, we conducted an extensive literature survey to understand how machine
learning models, are being applied in the domain of predictive maintenance in networking. Our
aim was to glean insights into the methodologies adopted by other researchers in applying these
ML models, understand the kind of data they utilized, comprehend how the data was processed
and prepared for the model, ascertain the process of training the models, and, most crucially,
observe the results they yielded in predictive maintenance accuracy and efficiency.

The research conducted by Shafiq et al. (2016) [5] signifies the effective role of machine learning
in achieving successful network traffic classification. The researchers were able to develop a real-
time internet dataset through a network traffic capture tool, followed by the extraction of features
from this captured traffic. They applied four machine learning classifiers and noted the
exceptional performance of the C4.5 classifier, demonstrating its potential in managing complex
networking tasks. This finding is crucial in understanding the proficiency of decision tree models
in predictive maintenance within networks [5].

In the research conducted by Pacheco et al. (2014) [6], a comprehensive systematic review of
traffic classification is introduced which rests upon the use of machine learning techniques [6].
They identified a set of trends and expected future directions for ML-based traffic classification,
thereby giving us new insights and aiding us in understanding the potential applicability and
scalability of machine learning mechanisms in traffic classification and, more broadly, in the
realm of predictive maintenance.

The study by Soysal er al. (2010) leveraged various supervised machine learning algorithms and
reinforced the robustness and applicability of these tools in even high-pace dynamic network
environments. This contributes to our understanding of the adaptability and suitability of machine
learning algorithms like Bayesian Networks and Decision Trees in managing and predicting data
flows across networks with dynamic ports [7].

The fourth research conducted by Fan et al. (2017) underlines the effectiveness of Support Vector
Machine and unsupervised K-means clustering in traffic classification tasks within Software
Defined Networks (SDN) [8]. The researchers were able to achieve an overall accuracy of over
95% with these ML models, thus emphasizing the powerful capabilities of these algorithms in
accurately classifying network traffic and predicting network performance [8].

In the research conducted by Knapińska et al. (2021), researchers predicted the network data of
various frame sizes utilizing machine learning algorithms such as linear regression, k nearest
neighbors, and random forest, based on historical data [9]. The focus here on the importance and
effectiveness of 'on-line multiple time series prediction' is particularly interesting for our study, as
3
 International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.1/2, April 2024
it expands the scope and applicability of machine learning in predicting dynamic network data for
predictive maintenance [9].

Collectively, these studies illustrated the evolving landscape of network management,

empowered by machine learning. They inform our exploration of machine learning models in the
context of predictive maintenance in 5G networks.

3.1. Methodologies Adopted in Applying ML Models

Analyzing the implementation of machine learning algorithms in the literature provides a more
profound understanding of the different approaches and how they specifically influence network
traffic management tasks.

For instance, in the work of Shafiq et al. [5], a two-step process is utilized for data capture and
feature extraction. This information-rich dataset provides a more complex and nuanced
environment for the ML algorithms to train on. This level of intricate process implementation and
data handling are key aspects to consider when dealing with large network data, specifically
within the evolving 5G networks, known for their vast volume of data traffic.

Many researchers have set unique methodological stages for their studies. As seen in the
systematic review carried out by Pacheco et al. [6], the ML-based traffic classification study is
broked down into a well-structured pathway, establishing a well-defined framework which can be
valuable for newcomers in the field. Similar structured approaches might prove beneficial in
understanding the stepwise progression needed for ML enhanced predictive maintenance tasks in
the 5G networks.

The ability to modify and adapt methodologies to suit specific challenges of a network
environment is a crucial finding from our literature survey. As illustrated by Soysal et al. [7], the
implementation of ML algorithms, even in high-speed dynamic network settings, highlights the
potential of ML methodology for customizability, meeting the needs of various network
configurations.

The use of hybrid predictive models is also a recurring theme in the literature. The research by
Fan et al. [8] identifies how Support Vector Machine and unsupervised K-means clustering can
be combined for more robust and accurate traffic classification tasks. This approach of using
multiple ML algorithms can potentially provide improved results, and this might be especially
valuable considering the diverse data types present in 5G networks.

Finally, the importance of 'on-line multiple time series prediction' emphasized by Knapińska et
al. [9], significantly broadens the scope and applicability of machine learning in predicting
dynamic network data for predictive maintenance. These real-time prediction models might be
increasingly relevant in the complex, continuously changing environments of 5G networks, and
need to be a key consideration when designing predictive maintenance models.

Studies like these, featuring diverse and effective approaches, offer valuable insights into
potential methodologies that could be leveraged for implementing machine learning algorithms as
part of predictive maintenance strategies, considering the unique operational demands of 5G
networks.

4
 International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.1/2, April 2024
3.2. Data Collection and Pre-processing

Data forms the bedrock of predictive maintenance in network operations. The form and quality of
data determine the efficacy of the machine learning models. Therefore, detailed attention is
required in the collection and preprocessing phases.

In the research undertaken by Shafiq et al. [5], for instance, the primary data was sourced from
real-time Internet data that the researchers developed using a network traffic capture tool. Such
collection methods, which direct capture data from the network operations, tend to provide rich
and nuanced information which can enhance the training process for the ML algorithms.

On the other hand, the study by Knapińska et al. [9] is an excellent example of how to leverage
historical network data for predictive purposes. In their research, historical data of various frame
sizes were utilized to train their ML models for future predictions. Importance of past patterns
and trends in the functioning of networks can greatly enhance the ability of ML models to
forecast future network behavior.

Preprocessing of data is as critical, if not more, as data collection. It can majorly impact the
effectiveness of the subsequent ML model training. Several preprocessing steps were observed,
like data cleaning, standardization or normalization of numerical data, and transforming
categorical data into numerical labels. These steps can significantly enhance the quality and
usefulness of the data collected, thereby increasing the performance and robustness of the ML
models trained on it.

Understanding the data collection and preprocessing steps associated with the design of the ML
models in these studies has been beneficial in determining the best practices for our research.

3.3. Training of ML Models

The training process of the ML models was largely guided by the split of data into train-test sets,
the choice of loss function, and the optimization method used. A common approach is to allocate
a percentage of the data, usually around 70-80%, for training the model and then using the
remaining data for testing the model’s performance. The choice of loss function, which measures
the discrepancy between the model’s predictions and the actual results, and the optimization
method, that adjusts the model parameters to minimize the loss, were specifically tailored to align
with the predictive maintenance task in each study.

For instance, in the study by Fan et al. [8], they discussed the importance of selecting appropriate
loss functions and optimization methods during the training process. This decision greatly
influenced the accuracy and robustness of the ML models in classifying network traffic and
predicting network performance.

Soysal et al. [7], on the other hand, leveraged supervised learning, a machine learning paradigm
where the models are trained using labelled data. This approach allowed the models to 'learn' and
create an internal model for making predictions. The study emphasizes the importance of
supervised learning in traffic classification tasks, informing us of the potential benefits and
challenges of such an approach for predictive maintenance in 5G networks.

Recognizing and understanding these diverse strategies in training ML models provide valuable
insights when designing a predictive maintenance model for 5G networks.

5
 International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.1/2, April 2024
3.4. Results Obtained

We took special interest in the outcomes of each study, particularly in how effectively the
implemented ML models could conduct predictive maintenance tasks. A common metric in these
studies was prediction accuracy, which measures the proportion of total predictions made by the
model that were correct. Another metric that was often reported was confusion matrix elements
including precision (the proportion of positive identifications that were correct), recall (the
proportion of actual positives that were correctly identified), and F1 score (a balanced measure of
precision and recall). Some studies also reported computational efficiency, model robustness, and
how their solutions performed under different network conditions.

4. COMPARISON OF MACHINE LEARNING MODELS

In this section, we explore a theoretical comparison of different machine learning (ML) models
based on their inherent characteristics and capabilities, with a focus on how these models might
align with 5G network requirements for predictive maintenance.

Decision Trees (DTs): DTs are a subset of ML models primarily used for classification and
regression tasks. Their tree-like model maps decisions, resulting in an easily interpretable
structure that scales with the complexity of the decision-making process. For large networks like
5G, where multiple parameters may impact network performance, DTs can be beneficial.
However, DTs are prone to overfitting, tending to create complex models that don't generalize
well to new data [10].

Neural Networks (NNs): Inspired by the human brain, NNs can recognize patterns in large and
complex datasets. They consist of interconnected layers of nodes, or "neurons," which can
process input data and adjust their internal model parameters during training. NNs can be highly
effective in handling complex 5G network data, but their "black-box" nature can make them
difficult to interpret. Also, training NNs typically requires substantial computational resources
and time [11].

Support Vector Machines (SVMs): SVMs are renowned for their ability to handle high-
dimensional spaces, which would be useful in the complex environment of a 5G network. SVM
classifiers work by dividing data into separate classes with the maximum margin. However, they
can be computationally intensive and require careful tuning to avoid overfitting or underfitting
[12].

In summary, while each of these ML models holds promising potential for predictive
maintenance tasks in 5G networks, they come with their own nuances and challenges. The choice
between them should consider factors such as interpretability, computational complexity, data
requirements, and the specific characteristics of the 5G networks in question.

5. APPLICATION OF ML IN DIFFERENT NETWORK SETTINGS

Machine learning algorithms have been explored in various network settings, each with its unique
challenges and requirements.

5.1. Software-Defined Networks (SDNs)

SDNs are characterized by their flexibility and programmability. These networks separate their
control plane from the data plane, making the network more easily adjustable to changing needs.

6
 International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.1/2, April 2024
The use of ML in SDN, as evidenced by Fan et al. [8], can help in better traffic classification,
enabling more accurate network resource allocation and improving network performance.
Learning from this, the application of ML in 5G networks can be tailored for optimizing network
resources responsive to dynamic requirements.

5.2. Internet of Things (IoT) Networks

IoT networks consist of a vast number of devices generating a huge amount of data. The work by
Shafiq et al. [5] signifies the effective role of machine learning in achieving successful network
traffic classification in such a high-connectivity and high-data environment. Given the scale and
complexity of 5G networks, these considerations concerning high volumes of data traffic and
their effective management are very relevant.

5.3. Traditional Networks

In conventional network settings, machine learning can provide valuable insights for traffic
management, security, and quality of service. The work done by Soysal et al. [7] demonstrates
the applicability of machine learning algorithms in even high-pace dynamic network
environments. These experiences can be valuable when designing ML models for 5G networks,
which are expected to handle high-speed data with low latency.

5.4. Mobile Networks

Mobile networks present unique challenges due to their inherent mobility and variations in
network quality. Understanding how to handle these elements while implementing machine
learning algorithms for predictive maintenance can be quite beneficial for 5G networks, which
have a major focus on enhancing mobile connectivity.

In conclusion, machine learning applications in varied network settings provide important

learnings that can inform practices for implementing ML in 5G networks. Leveraging ML
effectively can lead to significant improvements in predictive maintenance strategies, enhancing
reliability and performance in 5G networks.

6. CHALLENGES IN THE IMPLEMENTATION OF ML

Implementing machine learning models for predictive maintenance in 5G networks brings its
own set of challenges, ranging from issues related to data management to model training and
generalizability.

6.1. Data Privacy and Security

With the wealth of data required to train and test ML models, data privacy and security become
central concerns. Ensuring the confidential and sensitive information contained within network
data is protected while efficiently utilizing it for predictive maintenance poses a significant
challenge. Techniques such as anonymization and pseudonymization can be used to protect
privacy, but they need to be designed and implemented carefully to avoid compromising the
utility of the data.

7
 International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.1/2, April 2024
6.2. Data Gathering and Quality

The effectiveness of ML models largely depends on the quality and relevancy of the data they're
trained on. Gathering high-quality, relevant data can be challenging, particularly in the diverse
and dynamic environment of a 5G network. Additionally, pre-processing of data for cleaning,
normalization, and encoding, while crucial, can be time-consuming and require substantial
computational resources.

6.3. Model Training

Training ML models requires computational resources and time. Further, the choice of model
parameters, loss function, and optimization method require expertise to avoid underfitting or
overfitting. Resource allocation for model training needs to be optimized for efficient and
effective ML implementation.

6.4. Generalizability

ML models need to be able to generalize well to new data. That is, a model trained on a specific
set of data should still perform well when exposed to new, unseen data. Ensuring this
generalizability, particularly in the rapidly evolving area of 5G networks, can be challenging.
Addressing these challenges requires a carefully planned approach that incorporates robust data
management strategies, appropriate resource allocation, and continuous monitoring and
adjustment of ML models to ensure they remain effective as network conditions change.

7. BEYOND PREDICTION: PROACTIVE NETWORK MAINTENANCE

STRATEGIES
While predicting maintenance needs is a cornerstone of network management, it can be just as
valuable to leverage ML predictions to proactively improve network performance. Shifting from
a reactive strategy to a proactive one can minimize network downtime, improve resource
allocation, and ultimately provide a better user experience.

7.1. Real-time Adjustments

With the high-speed, low-latency requirements of 5G networks, real-time adjustments based on

ML predictions could be critical in maintaining optimal network performance. For example, ML
models could predict high-traffic periods in various sections of the network and dynamically
allocate resources to prevent congestion, providing a consistent user experience even during peak
traffic periods.

7.2. Pre-emptive Resource Allocation

By accurately predicting future network demands, machine learning can enable more efficient
resource allocation. For instance, if ML models predict an increase in demand in a specific part of
the network, resources can be pre-emptively redirected to that area to ensure seamless service.

7.3. Proactive Security Measures

Machine learning predictions can also be utilized for network security. If an ML model predicts
potential security loopholes or threats, preventative measures can be taken proactively to tighten
security protocols and prevent potential intrusions or attacks.
8
 International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.1/2, April 2024
7.4. Improving Network Topologies

ML predictions can inform decisions about network topologies. By predicting the future usage
patterns and requirements of different nodes in the network, decisions about network topology,
including the placement of new nodes or rerouting connections, can be made proactively to
improve overall network performance.

Proactive network maintenance strategies offer a promising avenue for improving 5G network
management. These strategies leverage the full potential of machine learning, transforming
network maintenance from a reactive process to a proactive one, bringing significant
improvements in network reliability and performance.

8. FUTURE TRENDS IN ML AND PREDICTIVE MAINTENANCE

As we look forward, the intersection of machine learning and predictive maintenance is set to
become an increasingly active area of research and development. Several potential trends could
shape the future of this field, particularly in the context of 5G networks.

8.1. Advanced ML Algorithms

While traditional machine learning models like Decision Trees and Neural Networks have
demonstrated significant value, more advanced algorithms are on the horizon. For instance, Deep
Learning models, a subset of machine learning that mimics the human brain's neural networks on
a much larger scale, could deal with the complex, multi-dimensional scenarios inherent in 5G
networks. Reinforcement Learning, where models learn optimal actions based on reward
feedback, could dynamically optimize network performance in real-time.

8.2. Automated Machine Learning (AutoML)

With the increasing complexity and volume of network data, the manual process of designing and
tuning ML models could become unwieldy. AutoML, which automates the process of applying
machine learning, could help in scaling ML applications in predictive maintenance.

8.3. Explainable AI

With the complex decision-making process and resulting 'black box' phenomenon in some ML
models, there's growing emphasis on Explainable AI - AI systems that can provide
understandable explanations for their decisions. This increased transparency could boost trust in
ML-based predictive maintenance systems and aid human operators in decision-making.

8.4. Advanced Data Collection and Pre-processing

Future advancements in data collection could involve more sophisticated IoT sensors, on-device
data processing, and real-time data streaming [13]. These improvements could enhance the
accuracy and efficiency of the data used for predictive maintenance, thereby making ML
predictions more reliable.

8.5. Incorporation of Domain Knowledge

Integrating domain-specific knowledge into ML models, known as Domain-Specific Learning,

could potentially further boost the accuracy of predictive maintenance algorithms in the future.

9
 International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.1/2, April 2024
For instance, certain principles or patterns known about network behaviours could be integrated
into the models, aiding them in processing data and making predictions.

The future trends of machine learning models and predictive maintenance promise new
opportunities and challenges. It is important to continue this exploration and keep pace with these
developments, to fully leverage the potential advantages they might hold for 5G networks.

9. CONCLUSIONS
This paper explored the use of machine learning (ML) algorithms for predictive maintenance in
5G networks, addressing the central research question concerning the utilization of ML to
improve maintenance strategies. The insights derived from the comprehensive literature review,
methodical surveys, and theoretical comparisons established that ML algorithms can significantly
enhance predictive maintenance, leading to more reliable and efficient network operations.
Our findings demonstrate that decision trees, neural networks, support vector machines, and other
ML models each present unique benefits and limitations when applied to predictive maintenance
tasks in 5G settings. The adaptability of these models to the complex and dynamic nature of 5G
network traffic underscores their potential to improve upon traditional maintenance approaches.
The application of ML in varied network environments, such as software-defined networks, IoT
networks, traditional, and mobile networks, revealed the extensive applicability and the capacity
of ML algorithms to handle vast amounts of data and diverse network behaviors. However,
challenges such as data privacy, model generalizability, computational demands, and resource
optimization emerged as critical considerations for effective implementation.

Moving beyond prediction, we underscore the importance of transitioning to proactive network

maintenance strategies. ML predictions enable real-time adjustments, preemptive resource
allocation, proactive security measures, and improvements in network topologies—transforming
reactive maintenance into a proactive, performance-enhancing process.

Looking ahead, we anticipate that advancements in ML algorithms, AutoML, explainable AI,

enhanced data collection methods, and the incorporation of domain knowledge will drive future
trends in ML and predictive maintenance. These developments hold the promise of refining
predictive maintenance strategies and optimizing network performance in 5G and beyond.

Our contribution to the body of knowledge in ML-based predictive maintenance for 5G networks
underscores the transformative impact of ML algorithms on network management. As the
landscape of 5G technology evolves, so too must the strategies for maintaining its infrastructure.
Through continuous research and innovation in ML applications, predictive maintenance will
play a pivotal role in sustaining the growth and resilience of next-generation networks.

ACKNOWLEDGEMENTS
The author would like to extend sincere thanks to New York University for graciously providing
the resources to conduct the research.

10
 International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.1/2, April 2024
REFERENCES

[1] Kim, S., Kim, H.S. and Kim, C., 2019, April. ALICE: Autonomous link-based cell scheduling for
TSCH. In Proceedings of the 18th International Conference on Information Processing in Sensor
Networks (pp. 121-132).
[2] Duquennoy, S., Al Nahas, B., Landsiedel, O. and Watteyne, T., 2015, November. Orchestra: Robust
mesh networks through autonomously scheduled TSCH. In Proceedings of the 13th ACM
conference on embedded networked sensor systems (pp. 337-350).
[3] Jung, D., Zhang, Z. and Winslett, M., 2017, April. Vibration analysis for IoT enabled predictive
maintenance. In 2017 ieee 33rd international conference on data engineering (icde) (pp. 1271-1282).
IEEE.
[4] Palattella, M.R., Accettura, N., Dohler, M., Grieco, L.A. and Boggia, G., 2012, September. Traffic
aware scheduling algorithm for reliable low-power multi-hop IEEE 802.15. 4e networks. In 2012
IEEE 23rd International Symposium on Personal, Indoor and Mobile Radio Communications-
(PIMRC) (pp. 327-332). IEEE.
[5] Shafiq, M., Yu, X., Laghari, A.A., Yao, L., Karn, N.K. and Abdessamia, F., 2016, October.
Network traffic classification techniques and comparative analysis using machine learning
algorithms. In 2016 2nd IEEE International Conference on Computer and Communications (ICCC)
(pp. 2451-2455). IEEE.
[6] Pacheco, F., Exposito, E., Gineste, M., Baudoin, C. and Aguilar, J., 2018. Towards the deployment
of machine learning solutions in network traffic classification: A systematic survey. IEEE
Communications Surveys & Tutorials, 21(2), pp.1988-2014.
[7] Soysal, M. and Schmidt, E.G., 2010. Machine learning algorithms for accurate flow-based network
traffic classification: Evaluation and comparison. Performance Evaluation, 67(6), pp.451-467.
[8] Fan, Z. and Liu, R., 2017, August. Investigation of machine learning based network traffic
classification. In 2017 International Symposium on Wireless Communication Systems (ISWCS)
(pp. 1-6). IEEE.
[9] Knapińska, A., Lechowicz, P. and Walkowiak, K., 2021, June. Machine-learning based prediction
of multiple types of network traffic. In International Conference on Computational Science (pp.
122-136). Cham: Springer International Publishing.
[10] Quinlan, J.R., 1986. Induction of decision trees. Machine learning, 1, pp.81-106.
[11] LeCun, Y., Bengio, Y. and Hinton, G., 2015. Deep learning, nature, 521, 436–444. DOI: https://doi.
org/10.1038/nature14539.
[12] Cortes, C. and Vapnik, V., 1995. Support-vector networks. Machine learning, 20, pp.273-297..
[13] Zhou, D., Yan, Z., Fu, Y. and Yao, Z., 2018. A survey on network.

AUTHORS
Yamini Kannan, an NYU graduate with a Master's in Information Systems, has swiftly
advanced from a data analyst and research assistant role to a network engineer at
prestigious firms like Meta and now, to a top-tier fintech firm. With a proven track record
of automating processes and boosting network efficiency, her innovative solutions at
Meta garnered senior leadership recognition. Yamini’s academic achievements include a
summer at the London School of Economics and a Bachelor's in Electronics from PES
University, underscoring her robust foundation in tech and finance.

Dharika Kapil, a master’s degree holder from Stevens Institute of Technology, embodies
a fusion of inquisitive spirit and innovative prowess in Full Stack Development within
FinTech. Her career has spanned significant positions at MarketAxess, Instinet, and Tata
Consultancy Services, where she has excelled in enhancing trading systems and user
experiences. Dharika's journey, marked by her scientist father's influence, has been
dedicated to technological advancements and breakthroughs.

11