LEVERAGING MACHINE LEARNING FOR RESOURCE OPTIMIZATION IN USA DATA CENTERS: A FOCUS ON INCOMPLETE DATA AND BUSINESS DEVELOPMENT

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

119

https://www.theamericanjournals.com/index.php/tajet

PUBLISHED DATE: - 19-12-2024

DOI: -

https://doi.org/10.37547/tajet/Volume06Issue12-12

PAGE NO.: - 119-140

LEVERAGING MACHINE LEARNING FOR
RESOURCE OPTIMIZATION IN USA DATA
CENTERS: A FOCUS ON INCOMPLETE DATA
AND BUSINESS DEVELOPMENT

Md Sumsuzoha

Master of Science in Business Analytics, Trine University

MD Sohel Rana

Executive Ph.D. in Business Analyst, University of Cumberlands

Md Shahidul Islam

MBA in business analytics, International American University

Md Khalilor Rahman

MBA, Business analytics, Gannon University, Erie, PA, USA

Mitu Karmakar

School of Business, International American University, Los Angeles,
California, USA

Md Sazzad Hossain

MBA, business analytics, Gannon University, Erie, PA, USA

Reza E Rabbi Shawon

MBA Business Analytics, Gannon University, Erie, PA

Corresponding Author: Md Sumsuzoha

RESEARCH ARTICLE

Open Access

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

120

https://www.theamericanjournals.com/index.php/tajet

INTRODUCTION

Background and Motivation

As per Rahman et al. (2023), data centers form the
linchpin of modern digital infrastructure, enabling
everything from e-commerce and streaming to
cloud computing and artificial intelligence. The
United States, at the forefront of technology,
houses some of the world's most extensive and
technologically advanced data centers as a part of
its economic and technological framework. Recent
studies have indicated that U.S. data centers make
up about 2% of the nation's total electric use, and
this trend is probably continuing to increase with
the growth in the number of computational
resources the modern world requires. Shawon et
al.(2023b), asserted that the scale and power
involved in such facilities outline the importance of
efficiently

using

resources

to

achieve

sustainability. To that effect, efficient resource
utilization not only cuts operation costs but also

meets global efforts to ensure carbon emissions
are reduced. An area where inefficiency reins in
include cooling systems, hardware utilization, and
power distribution. It is on this basis that
addressing such inefficiencies becomes quite
important in maintaining the operational viability
of data centers to support their expansion in the
digital economy (Alam et al., 2023; Abdulazeez et
al. 2024).

Problem Statement

Notwithstanding, most data centers still suffer
from ineffective use of their resources due to
factors such as imbalance in workload, inefficient
cooling methods, and underutilization of
hardware. Incomplete data is one of the most
critical barriers to optimization (Al Mukaddim et
al., 2023a; Feng et al., 2024). Missing sensor
readings, missing hardware logs, and incomplete
environmental data pose a problem to the

Abstract

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

121

https://www.theamericanjournals.com/index.php/tajet

traditional resource management system, leading
to less-than-optimal decisions. The inability to
handle incomplete data weakens the reliability and
robustness of resource optimization. The specific
value of machine learning lies in providing
transformative opportunities regarding the
aforementioned challenges exacerbated by
incomplete data that plagues traditional
optimization approaches so often (Buiya et al.
2023; Ezeigweneme et al., 2024). This study aims
to explore how different ML techniques can be
used for optimizing resource utilization by data
centers in the US, focusing on strategies to handle
incomplete data and implications for business
development. Machine Learning potentially
improves efficiency, reduces cost, and unleashes
paths toward sustainable growth by applying
predictive analytics, anomaly detection, and
reinforcement learning. Examples of incomplete
data are being treated with imputation techniques
or robust model design to meet the promise of ML
in these most critical environments.

Research Questions

RQ1: How can machine learning techniques be
applied to optimize resource usage in USA data
centers?

This research question aimed at understanding
how machine learning algorithms can help in
improving efficiency in data center operations.
Analyzing the trend of past data on energy
consumption, server utilization, and several other
factors, machine learning models can identify
patterns and thereby predict the future need for
resources. This information can thus be used to
optimize resource allocation, reduce energy
consumption, and minimize operational costs.

RQ2: How can incomplete data be addressed in
machine learning models for data center resource
optimization?

This research question aims to explore imputation,
data augmentation, or robust learning algorithms
that can help deal with missing data for the
accuracy and reliability of machine learning
models. Addressing incomplete data challenges
will improve the performance of machine learning
models and enhance data center resource
optimization. Data centers often generate large
volumes of data; however, due to several reasons
such as sensor failures, network outages, or human
error, this data can be incomplete.

RQ3: What are the business development
implications of optimizing resource usage in
data centers?

The question deals with the higher-order business
value linked to data center resource optimization.
The

improvement

in

profitability

and

sustainability for a data center will result from
reductions in energy consumption and operational
costs. Besides, optimized resource usage will
enable data centers to scale their operations more
efficiently and respond to changing business
needs. Moreover, data-driven insights gained from
machine learning models can help identify new
business opportunities and improve service
delivery.

Significance of the Study

The significance of this research lies in its
capability to resolve the pressing challenges
confronted by U.S. data centers, which are
instrumental to the digital economy yet burdened
by inefficiencies in resource utilization. This work
will advance not only the technical capability of
data centers but also their alignment with goals of
sustainability and cost reduction through machine
learning for optimization. Addressing incomplete
data makes this optimization robust and reliable in
real-world conditions, considering that data
centers must function under such conditions. From
the perspective of business development, it has
far-reaching implications, as this may mean

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

122

https://www.theamericanjournals.com/index.php/tajet

reduced operational costs, improved reliability in
service delivery, and increased competitiveness in
the marketplace. This has contributed to wider
goals of environmental stewardship and economic
development by assuring that data centers
continue to serve the growing digital
infrastructure safely and efficiently.

LITERATURE REVIEW

Overview of Data Center Operations and
Resource Usage in the USA

Islam et al. (2023), posited that data centers have
emerged as the foundational building block of the
digital economy, supporting a diverse array of
applications and services, ranging from cloud
computing and e-commerce to AI and critical
infrastructure systems. Among all types of facilities
in the United States, data centers rank as some of
the most power-intensive, consuming about 2% of
the total electricity consumption in the nation. The
general

operations

will

include

server

management,

cooling

systems,

storage

optimization, and networking infrastructure
(Hasanuzzaman et al.; Gong et al., 2024). These
centers rely on a delicate balance between
computation requirements, power consumption,
and physical space, all at adequate uptime and
availability.

According to Karmakar et al. (2023), the most
critical concerns for resource management facing
US data centers are related to energy efficiency.
The cooling systems themselves, required to
maintain

hardware

at

ideal

operating

temperatures, consume enormous amounts of
energy. Inefficiently utilized servers that are often
left running for redundancy further exacerbate the
problem. As per Gebreyesus et al. (2023), other
challenges include the integration of renewable
energy

sources

and

real-time

workload

distribution across servers. All of these challenges
require highly sophisticated approaches, where
optimization of resources should guarantee

reliability and scalability of operation. ML can
outline some very promising ways to solve these
complexities through intelligent decision-making
in resource management automation (Khan et al.,
2023).

Machine Learning for Resource Optimization

Predictive Modeling

Nasiruddin et al. (2023), articulated that Predictive
modeling is among the cornerstones of machine
learning applications in data centers that involve
the forecasting of requirements well in advance
from historical and real-time data to enable
proactive resource allocation. Examples of some of
the applied techniques include regression analysis,
neural networks, and time series modeling. For
instance, ML-driven predictive models may
forecast peak workload periods and take actions
that involve the pre-emptive allocating of extra
resources to avoid latencies and system overload.
These models shine in situations such as seasonal
e-commerce traffic or periodic data processing
spikes (He et al., 2024; Kumari et al. 2024).

Clustering

Clustering algorithms, like k-means or hierarchical
clustering, play a significant role in workload
analysis concerning workload patterns and
resource optimization. The group data points of
similar configuration into one group (Shawon et al.
2024; Li et al., 2024). For example, by applying
different clustering methodologies, workloads can
be segmented based on the requirements
associated with workload processing and then
dispatched onto servers that are capable of
matching processing needs. This ensures that
resources are neither over-provisioned nor
underutilized, hence contributing to energy
savings and cost reductions (Sumon et al. 2024).

Decision Trees and Ensemble Methods

As per Shil et al. (2024), decision trees and

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

123

https://www.theamericanjournals.com/index.php/tajet

ensemble methods, like random forests and
gradient boosting machines, are of particular
importance to resource allocation decisions. The
algorithms offer interpretable models that forecast
optimal settings related to server power
distribution, cooling, and workloads. Decision-
making is a big positive attribute in generally
dynamic

environments

where

operational

parameters often change. For example, the
decision tree model could suggest operating a
cooling power reduction in a zone for low
workloads while sustaining efficiency without
losing reliability (Krothapalli et al., 2023; Ma et al.,
2024).

Reinforcement Learning

Wang et al. (2024), stated that Reinforcement
learning has recently become one of the state-of-
the-art machine learning approaches for data
center optimization. The nature of the RL models,
which learn through trial and error to maximize a
reward signal, makes them suitable for managing
dynamic resources. For instance, using an RL
algorithm, cooling systems can autonomously read
real-time environmental conditions and optimize
energy use independently of human adjustments
(Shawon, 2023c). Notable applications include
Google's collaboration with DeepMind, in which
reinforcement learning reduced cooling energy
consumption by 40% within their data centers.

Handling Incomplete Data in Machine Learning

According to Zhang et al. (2024), missing data in a
data center operation is a pervasive challenge that
might

result

from

hardware

failures,

malfunctioned sensors, or network failures. Most
ML models require large-scale, high-quality
datasets for good training and deployment, and
hence the treatment of missing data becomes
crucial.

Imputation Techniques

Imputation is the process of estimating the values

of missing data for a complete dataset. The simple
forms mean or median imputation, only take the
average of available data for the missing ones.
These methods, though simple, tend to
oversimplify many complex patterns in data.
Richer imputation methods such as KNN or matrix
factorization substantially improve accuracy by
considering the relationships between variables
(Rahman et al., 2023). For instance, KNN uses the
correlation between server workloads and cooling
demands for missing reading imputation when a
dataset of power consumption for a data center is
provided.

Interpolation Techniques

Interpolation uses mathematical functions to
estimate missing data points within a given range.
Linear interpolation, spline interpolation, and
polynomial interpolation are common, depending
on the complexity of the dataset. For example, in
sensor networks that monitor temperature across
a data center, missing values will be reconstructed
by using spline interpolation to keep the
monitoring continuous.

Data Augmentation

Debnath et al. (2023), stated that Data
augmentation is a type of technique involved in
adding to the size of available datasets with
artificially created new records. It would include
techniques of bootstrapping or even the
generation of synthetic data using variants of GANs
for generating realistically distributive samples.
This should particularly be the case when missing
data occurs in complex patterns, or where not
much was initially available from training. Instead,
GANs could fill historical lacunas by simulating
probable patterns in workload distributions.
Robust ML Architectures Equally important is the
art of designing ML models that are insensitive to
incomplete data. To handle the uncertainty and
variability presented by missing data, ensemble
models, dropout layers in neural networks, and

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

124

https://www.theamericanjournals.com/index.php/tajet

Bayesian frameworks are particularly adept. For
instance, ensemble models, such as random
forests, aggregate predictions from a set of
decision trees and therefore reduce dependence
on any single tree where data may be missing
(Buiya et al. 2024).

Business

Development

Implications

of

Resource Optimization

The optimization of resource utilization in data
centers has far-reaching ramifications for business
development, particularly within the U.S. context,
where data centers play a crucial role in both
technological and economic growth as described
below:

Operational Cost Reduction. As stated earlier,
resource optimization can positively affect
operational costs, being some of the highest
financial burdens for data centers. Consequently, it
leads to great savings for the data center with
minimized energy consumption through effective
distribution of workload and dynamic adjustment
of cooling. For example, predictive models
estimating a peak period and pre-dedicating
resources avoid overprovisioning, hence a
reduction in utility bills. In return, these savings
might be reinvested to facilitate an increase in
infrastructure by applying more advanced
technologies, again encouraging a cycle of growth
and innovation.

Improved Efficiency and scalability. Optimized
resource use ensures data centers operate at peak
in handling increased workloads without
increased proportional consumption of resources.
This aspect of scalability is important in a business
perspective where demands for digital services are
growing each day. Efficient operations minimize
downtime and improve the reliability of services,
thus improving customer satisfaction and
retention (Islam et al., 2024).

Data-Driven Decision-Making. Adopting ML-driven

resource optimization encourages a fact-based
decision-making culture. With predictive analytics
and real-time monitoring, operators gain profound
insight into the performance metrics they need to
make informed strategic planning. This places the
decision-makers in a good position to develop an
investment plan in energy-efficient hardware or
renewable energy sources, based on forecasts from
data-driven projections about cost savings in the
long term (Hasanuzzaman et al., 2023).

Data Collection and Preprocessing

Data Sources

The dataset was retrieved from the GitHub
repository, which provided a rich dataset of
resource usage metrics and operational data from
U.S. data centers. It contained complex and fine-
grained information that was necessary to
optimize data center performance and deal with
incomplete data challenges. A detailed description
of the dataset and its key attributes was provided.
It was designed for analyzing resource usage
patterns in data centers, putting much emphasis on
energy efficiency, workload distribution, and
operational reliability (Pro-AI-Robikul, 2023). This
integration of time-series data with sensor
readings and performance logs provided a
comprehensive overview of resource consumption
and environmental conditions in data center
operations. This dataset was curated for the
engagement of machine learning models in the
study and optimization of resource consumption
along with the challenges of missing data. The
analyst will have more actionable insights into
data center operation for efficient and sustainable
performance in the digital economy by making this
dataset accessible.

Data Pre-processing

A code snippet in Python was incorporated into the
preprocessing pipeline of machine learning:
Firstly, the code started by standardizing column

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

125

https://www.theamericanjournals.com/index.php/tajet

names to ensure consistency across multiple
datasets. Secondly, it concatenated these datasets
vertically, eliminating any duplicate rows. To
handle missing values, the code filled in missing
entries with the median value of the respective
column. Thirdly, the dataset was reordered based
on timestamps for chronological analysis. Finally,
the resulting merged dataset was saved as a CSV
file for any further use. The cleaning process
ensured that all data quality and consistency have
been pre-processed and increased, in terms of
completeness for it to be suitable for model
training and evaluation. Handling missing values is

an important data pre-processing step. Imputation
techniques included the replacement of missing
values with estimated ones. Mean, median, or
mode were used to replace the values in simple
imputation methods. Other versions, such as more
sophisticated ones like K-Nearest Neighbors,
impute based on the values of similar data points.
Another interpolation technique was deployed
involving the estimation of missing values from the
trend of the surrounding data points. Depending
on the complexity of the trend, either simple or
spline interpolation is applied.

Exploratory Data Analysis (EDA)

Figure 1: Displays the Distribution of CPU Usage

This histogram displays the distribution of CPU
usage with a superimposed density curve. There is
a highly right-skewed, or positively skewed,
distribution with several distinct peaks. The most
prominent peak near the 0 mark reflects very high
frequency, about 95,000, of instances with low CPU
usage. Several other, much smaller peaks exist at
intervals, notably around the values for CPU usage

of 1, 2, and 3, but their frequencies are much lower,
in the range of 5,000 to 10,000. This pattern
suggests that the system is normally running at
low CPU utilization, with periods of moderate to
high usage creating secondary peaks. The long
right tail extending beyond CPU usage of 5
indicates very high CPU utilization, though these
occur with very low frequency.

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

126

https://www.theamericanjournals.com/index.php/tajet

Figure 2: Portrays GPU Usage Over Time

This time series chart above shows a pattern of
GPU usage during almost a year, from mid-2022 to
mid-2023, in a highly turbulent manner. The
values range from -2.5 to 1.0 with very sudden
fluctuation between these two bounds. Notice the
high activities in September and November of
2022, that it is denser and has fluctuations at a very
fast pace. Starting in January 2023, this pattern
becomes much more random and less dense with

clearly separated spikes. Such a negative value for
this metric is somewhat unusual for the case of
GPU usage; they might reflect either a
measurement scale that is normalized, including
possible anomalies in monitoring. Thus, these
would reflect batch job-type processing workloads
in nature and probably mean that periodic and not
quite steady, strong computations should run on
the GPUs but also could run in discontinuity.

Figure 3: Exhibits Memory Usage Minus Disk Cache Over Time

The above time series graph shows memory usage,
minus disk cache, over about a year from mid-2022
to mid-2023; it depicts raw measurements in gray
and a rolling average in green. Overall, memory

usage is around 0 GB, with a baseline between -1
GB and 0 GB, but there are several large spikes
throughout the period. The most prominent spikes
are in November 2022, reaching about 5 GB, and in

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

127

https://www.theamericanjournals.com/index.php/tajet

May of 2023, whereby the usage spikes over 6 GB.
The moving average is represented by the green
line, which smooths

the high-frequency

fluctuations and reveals the underlying patterns in
the memory usage spikes, usually short-lived
before returning to baseline. The negative values of

the baseline indicate that essentially, the disk
cache exceeds that amount of real memory usage
in normal operation, which in general is a
characteristic of well-optimized systems where
memory not used is used for caching.

Figure 4: Visualizes Hexbin Plot: CPU vs. GPU Usage

This hexbin plot visualizes the relationship
between CPU and GPU usage despite color
intensity showing the density. We notice several
interesting patterns in that plot: there is for
example a strong concentration at CPU usage of
around 0 and GPU usage within 0.2, shown by the
quite dark blue hexagons falling in this region. The
vertical distribution of GPU usage spans from -2.5
to 1.0, while CPU usage extends from -1 to about 5.
In particular, there are distinct horizontal bands of
activity at GPU usage levels of approximately -2.5,
0.2, and 1.0, indicating that the GPU tends to
operate at these specific levels regardless of CPU
utilization. The marginal histograms on the edges
give some indication that the CPU usage has a
right-skewed distribution with many peaks, but
the usage of the GPU seems more discrete, even
multi-modal, with quite clear states of preference.
This is a pattern indicative of CPU and GPU usage
being relatively independent, with significant GPU

operations happening in relatively distinct states
rather than continuously over time.

METHODOLOGY

Feature Engineering and Selection

Feature engineering in this regard should focus on
the derivation of meaningful temporal and
behavioral patterns from the data on system
resource usage. Key engineered features could
include rolling statistics, standard deviation, min,
and max-over various time windows, such as 1-
minute, 5-minute, and 1-hour for CPU, GPU, and
memory usage. Other features could be created by
calculating the rate of change, peak-to-average
ratios, and utilization patterns during specific
periods of peak hours versus off-hours. Most
valuable in this respect can be the cross-metric
feature-engineered variables, such as the ratio
between CPU and GPU utilization, memory usage
normalized by the CPU activity, or composite

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

128

https://www.theamericanjournals.com/index.php/tajet

metrics that express the overall system load. Time-
based features should encode time components,
including an hour of the day, day of the week, and
month, by using respective sine and cosine
transformations to capture the periodic patterns of
resource utilization.

Feature selection was driven by both statistical
significance and domain relevance. The most
important feature selection criteria involved the
importance score of features from tree-based
models such as Random Forests and XG-Boost,
correlation analysis to identify and remove
redundant features, and stability selection by
iterating multiple times through model training.
Features that show high VIFs were kept with great
care so that multicollinearity cannot take place.
Feature selection gave priority to features that
demonstrated consistency with their predictive
power across time and system states. Besides, the
features should be checked for their computational
efficiency and real-time availability since this is
seemingly a monitoring system to which rapid
calculations would be advantageous. Such a
combination of filter methods-statistical tests,
wrapper methods-recursive feature elimination,
and embedded methods-LASSO regularization
provided a robust approach toward identifying the
most relevant features while preserving model
interpretability.

Model Selection and Justification

Analysis of resource utilization in US data centers
was accomplished using the application of various
models for machine learning, most notably,
Logistic Regression, Random Forest, and Support
Vector Machines; each best fits different
characteristics of either or both of the aspects
relating to the data under discussion or even for
the objective of performing prediction: **Logistic
Regression** represents the statistical model used
while attempting to achieve binary classifications
such as overloading versus not overloading a

certain server within predetermined conditions. It
is simple and interpretable; hence, it forms a very
good starting point for selecting the main
predictors of operational anomalies. **Random
Forest** is an ensemble learning method based on
decision trees, especially efficient while dealing
with complex nonlinear relationships between
data. By aggregating multiple trees' predictions,
robustness and accuracy are gained for tasks such
as the prediction of energy consumption patterns
or the classification of server workloads. They are
particularly effective in classification and
regression problems in high-dimensional spaces.
**Support Vector Machines **classify data by
drawing a hyperplane that separates data points,
making them useful in anomaly detection in
resource usage or classification of the efficiency of
cooling systems under changing conditions. Each
of these models brings distinct strengths to the
table, enabling a comprehensive approach to data
center optimization. A choice of these models is
justified according to the characteristics of the
dataset and possible goals of the analysis.

Logistic regression fits binary outcomes and,
therefore, stands as a baseline model in many
respects, such as interpretability and low
computational needs. Random Forest handles
missing data in these data well, and generally
provides superior performances on noisy and
complicated data; hence, it also offers excellent
representations concerning incomplete and
multifaceted datasets. This will also shed light on
the most influential variables causing resource
utilization based on the feature importance of this
metric. Also, since events like failure are so rare
and critical in a data center, SVMs can handle high-
dimensional and imbalanced datasets. All these
factors together bring about interpretability,
robustness, and precision, hence providing great
help in predicting and reaching effective decisions
over the optimal utilization of resources in a data
center.

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

129

https://www.theamericanjournals.com/index.php/tajet

Train and Test Framework

A rigorous training and testing framework is
pivotal to guarantee that the machine learning
algorithms generalize well to unseen data and
perform effectively in real-world scenarios. The
commonly used splitting of data into the training-
testing set is in an 80-20% fashion, where model
training is done on 80% of the data, and the rest of
it, 20% is used for testing purposes. This division
ensures that the model is evaluated on a subset of
data it has not seen during training, providing an
unbiased assessment of its predictive accuracy.
Care is taken to ensure the data split maintains the
distribution of key variables, particularly when
dealing with imbalanced classes, such as anomalies
in server workloads or rare energy efficiency
issues. Stratified sampling is used in classification
to maintain class proportions in both training and
testing sets.

To enhance further the reliability of model
evaluation, k-fold cross-validation is applied. In
this technique, a training set is divided into k
subsets or "folds." Training on k-1 folds and
validation on the remaining one is performed by
the model while rotating through all folds until
each subset has served as the validation set.
Typical choices for k are 5 or 10; the choice
balances computational efficiency with the
thoroughness of evaluation. This helps to avoid
overfitting on a particular training subset and
provides a better, more realistic estimate of the
performance of the model. If there is a time series
or a sequence dependency, methods such as time-
based cross-validation or walk-forward validation
should be applied to preserve order when splitting
the data.

Hyperparameter Tuning

Model hyperparameter optimization is a crucial
process to avoid underfitting or overfitting to get
the peak performance. Grid search is an exhaustive
approach where a model needs to consider every

different combination with pre-specified values of
hyperparameters and then select the best setting.
For instance, a Random Forest model would make
use of grid search to tune the number of trees via
n_estimators, maximum tree depth via max_depth,
and minimum number of samples required to split
a

node

via

min_samples_split.

While

computationally expensive, the warranty with grid
search is that the best combination of parameters
within the predefined space will be found.

As a more computationally efficient alternative,
random search randomly samples a subset of the
hyperparameter combinations and evaluates
them. Random search often outperforms grid
search in finding near-optimal solutions at
considerably lower computational costs. Random
search was especially useful in models with lots of
hyperparameters, such as Support Vector
Machines or neural networks. Besides, more
complex approaches like Bayesian optimization an
approach that iteratively builds a probabilistic
model of the objective function and selects
hyperparameters to evaluate, based on expected
improvement also possible. Hyperparameter
tuning integrates with cross-validation to ensure
the performance of selected parameters is not
specific to one dataset split. Implemented together,
these approaches provide that machine learning
models are optimized for precision and reliability,
therefore increasing their usefulness in resource
optimization for data centers.

Performance Evaluation Metrics

Evaluation of machine learning models requires a
set of complete metrics that describe various
aspects of model effectiveness, in particular for
tasks like classification and regression. For
classification, accuracy is a basic metric expressing
the number of properly predicted instances
against the total within a dataset. However, it is not
enough in certain cases, especially when one is
dealing with datasets that have imbalances in class

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

130

https://www.theamericanjournals.com/index.php/tajet

distribution. It means that precision, recall, and the
harmonic mean between the two-sensitivity F1 are
more understandable for this sort of analysis:
Precision describes what fraction of true positive
predictions comprise all positive predictions. In
cases where there is a high cost related to the so-
called "false alarm", the Precision would prove
itself rather valuable in such scenario analysis. On
the contrary, Recall/Sensitivity is computed in the
identification rate concerning positive predictions
classified as such; this remains fundamental when
poor detection entails huge damage to important
aspects with sensitive or fatal consequences that
have to do with fault anomaly detection. Another
important metric in binary classification is the
AUC-ROC curve, which stands for the Area Under
the Receiver Operating Characteristic Curve. It
gives a model the ability to trace the generally
positive cases against the generally negative cases
at different threshold levels.

RESULTS

Model Performance

a) Logistic Regression

An appropriate Python code snippet was deployed
during the implementation of logistic regression
for classification. It instantiated a Logistic-
Regression object, trained the instance on the
training data, X_train, and y_train, and utilized the
trained model to make predictions on the test data,
X_test to produce the predicted labels, y_pred_lr.
The classification report summarized the
precision, recall, F1-score, and support for each
class in detail. Additionally, the accuracy score is
also calculated to check the overall correctness of
the prediction. This report helped the analyst to
understand the strengths and weaknesses of the
model, informing further improvements or
decisions from its predictions.

Output:

Table 1: Exhibits the Logistic Regression Classification Report Results

Logistic Regression Classification Report Results
precision recall f1-score support

0 0.98 0.97 0.98 26622
1 0.96 0.97 0.97 20499

accuracy

0.97 47121

macro avg

0.97 0.97 0.97 47121

weighted avg

0.97 0.97 0.97 47121

Accuracy: 0.97

The classification report above essentially gives
the details of the performance of the Logistic
Regression model. This model has given a very
good performance for class 0 with a high precision
of 0.98, recall of 0.97, and F1-score of 0.98,
reflecting well on its accuracy in identifying true
positive instances. For class 1, the model has had a
good performance at precision at 0.96, recall at

0.97, and an F1-score of 0.97; hence, it classified
most of the true positive instances correctly for
this class. The overall accuracy is 0.97, which
means the model can predict class labels correctly
for about 97% of the test instances. That means the
logistic regression model does quite a good job of
classifying data into two classes with high accuracy
while the performance is balanced in both classes.

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

131

https://www.theamericanjournals.com/index.php/tajet

b) Random Forest

Python code snippet implemented a random forest
classifier. It started by instantiating a Random-
Forest-Classifier object with a specified number of
trees, n_estimators=100, and a random state for
reproducibility. The model trained on the train
data X_train, y_train. After training, the model was
used on the test set X_test to make predictions,

y_pred_rf. The classification report presented the
model's performance in detail with precision,
recall, F1-score, and support for each class. Other
than that, the accuracy score is also calculated as a
measure of the overall correctness of the
predictions. This report assisted in understanding
the strengths and weaknesses of the model and
thus further helped in making improvements and
decisions based on its predictions.

Output:

Table 2: Depicts the Random Forest Classification Report Results

Random Forest Classification Report Results

precision recall f1-score support

0 1.00 1.00 1.00 26622
1 1.00 1.00 1.00 20499

accuracy

1.00 47121

macro avg

1.00 1.00 1.00 47121

weighted avg

1.00 1.00 1.00 47121

Accuracy: 1.00

The classification report of the Random Forest
model gives an outstanding performance. For class
0 and class 1, the model has a perfect precision,
recall, and F1-score of 1.00. That means the model
correctly identified all true positive instances and
did not have any false positives or false negatives.
The overall accuracy of the model is also 1.00,
which further confirms the ability of the model to
classify the data with 100% accuracy. Results
obtained here show that Random Forest is
outstanding in classifying the data into two classes
of remarkably good accuracy and reliability
predictions.

c) Support Vector Machines

A suitable Python code fragment facilitated the
implementation of a Support Vector Machine
(SVM) for classification tasks. The code begins by
generating an SVC object with an 'rbf' kernel and a
specified random state for reproducibility. The
model was then trained on the training data
(X_train, y_train). After training, the model was
used to make predictions on the test data (X_test),
generating the predicted labels (y_pred_svc). The
classification report had the performance of the
model in detail: precision, recall, F1-score, and
support for each class. Besides, the accuracy score
was computed as a measure that deals with the
overall correctness of the predictions.

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

132

https://www.theamericanjournals.com/index.php/tajet

Output:

Table 3: Showcases the Support Vector Machine Classification Report

Support Vector Machine Classification Report Results
precision recall f1-score support

0 0.99 1.00 0.99 26622
1 1.00 0.99 0.99 20499

accuracy

0.99 47121

macro avg

0.99 0.99 0.99 47121

weighted avg

0.99 0.99 0.99 47121

Accuracy: 0.99

The overall classification report of the SVM model
is very impressive. In class 0, the model has
performed well with high precision (0.99), recall
(1.00), and F1-score (0.99); hence, it is very well at
selecting true positive instances of this class.
Similarly, for class 1, the precision (1.00), recall
(0.99), and F1-score (0.99) are good, meaning true
positive instances of this class can be correctly
classified by the model. The overall accuracy of the
model is 0.99, with the model correctly predicting
the class labels for about 99% of the test instances.
These results thus indicate that the SVM model will
be effective in classifying the data into the two
classes with high accuracy and balanced
performances across both classes.

ROC Curve Model Comparison

The code snippet in Python compared three
classification models: Logistic Regression, Random
Forest, and SVM. For each model, it calculated the
ROC curve and AUC. For those whose default is to
predict with predict_proba such as Logistic
Regression and Random Forest, it took the
probabilities to plot the ROC curve. In the case of
the SVM, which does not have predict_proba, the
decision function was used instead. All the ROC
curves were plotted on the same graph for visual
comparison. The AUC scores for each model will
also be printed as a numerical measure for
comparison.

This

analysis

facilitated

an

understanding of the relative strengths and
weaknesses of the different models, aiding in
model selection and decision-making.

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

133

https://www.theamericanjournals.com/index.php/tajet

Output:

Figure 5: Visualizes ROC Curve Comparison of Models

This ROC curve comparison presents three
classification models: Logistic Regression, Random
Forest, and SVM. It plots the TPR against FPR at
various settings of threshold values. The closer it
goes to the upper-left corner, the better the
performance. Performance has been quantified by
AUC. In this case, the Random Forest and the SVM
models show an AUC score of 1.00, indicating
outstanding discrimination between the positive
and negative classes. The Logistic Regression
model also yields quite a good AUC of 0.99. From a
visual point of view, this would indicate that the
Random Forest and the SVM are the most capable
classifiers for this dataset, while the Logistic
Regression provides results that are slightly less
discriminative.

Comparison of Models Accuracy

The executed code snippet in Python compared the
performance of a few classification models by
calculating their training and testing accuracy.
Then, it created a histogram chart comparing the
training and testing accuracies. Bars were colored
differently to differentiate between the training
and testing accuracies. Bars were annotated with
their accuracy values. It finally analyzed the gap
between the training and testing accuracies for
probable overfitting. Models with big gaps
between training and testing accuracies were
flagged for probable overfitting, while models with
smaller gaps were able to generalize well. This
analysis helped in choosing the best model and
also finding possible issues like overfitting.

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

134

https://www.theamericanjournals.com/index.php/tajet

Figure 6: Depicts Training vs. Testing Accuracy of Models

The above bar chart shows the training and testing
accuracies of three classification models: Logistic
Regression, Random Forest, and Support Vector
Machine. The blue bars represent the training
accuracy, while the red bars represent the test
accuracy. Both Random Forest and SVM have very
similar training and test accuracies, indicating that
they generalize well to unseen data and are not
prone to overfitting. Logistic Regression presents
slightly larger differences between the accuracy
obtained for training and testing, which can be
considered overfitting. Retrospectively, the
Random Forest and SVM models seem to be robust
and reliable, placing probably the Random Forest
slightly above, given their performance is nearly
perfect for training and testing.

Feature Importance and Analysis

Understanding the feature importance of a
machine learning model sheds light on the data
itself and can be used for better interpretability.
With algorithms like Random Forest and Gradient
Boosting, the feature importance score is handy in
understanding the value of each feature in the
prediction made by the model. By evaluating
feature importance scores, the analyst can
pinpoint the key features that substantially
influence resource optimization. These features

might include:

Historical Resource Utilization: Past trends in
resource consumption can be used to predict
future trends and optimize resource utilization.

Workload Patterns: This is similarly critical as it
defines the workload nature, for instance, the peak
times of use and which applications use what
amount of resources.

External Factors: Weather conditions, holidays, or
economic indicators are some of the factors that
may influence the demand for resources and hence
must

be

considered

while

performing

optimization.

Hardware/Software

Configuration:

Specific

hardware and software configurations at the Data
Center level influence resource utilization and
should be factored into the optimization models at
work.

Handling Incomplete Data

Incomplete data is a prevalent challenge in real-
world datasets, and it can substantially influence
the performance of machine learning models. This
can be addressed in several ways:

Imputation: This is the process of filling up empty
values with estimated values. The most common of

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

135

https://www.theamericanjournals.com/index.php/tajet

such imputation techniques include mean
imputation, median imputation, and mode
imputation, and some of their advanced versions
include KNN imputation and multiple imputation.

Deletion: A straightforward approach is to remove
instances with missing values. However, this can
lead to significant data loss, especially if the
missing values are not missing at random.

Model-based imputation: This would involve the
training of a predictive model for observing the
missing values given the rest of the features.

Advanced Techniques: More advanced techniques,
such as multiple imputation and probabilistic
imputation, account for uncertainty in the
imputation process itself.

Implications for Data Center Management

The application of machine learning techniques
holds huge potential for the reformation of
resource management in US data centers. These
models analyze a pattern in historical data to
predict future resource demands, thus allowing
optimized resource allotment and minimizing
operational costs. It would also be possible to use
machine learning algorithms in forecasting the
periods of peak use so that the data center
operators can pre-scale their resources and avoid
service disruption. Besides that, predictive
maintenance models show the possibility of
hardware failures before they occur, which
reduces both downtime and maintenance costs.
Some recommendations for effectively integrating
machine learning into data center operations could
be:

Data Quality and Preparation: Basically, the quality
of data and how well it is prepared is considered
vital in training good machine learning models. It
is very important to clean, pre-process, and
feature-engineer with the intent of sustaining data
integrity and relevance.

Model Selection and Training: The usage of a
machine learning algorithm should be relevant to
a certain use case and the nature of the data. Model
training requires careful considerations together
with hyperparameter tuning.

Continuous Monitoring and Evaluation: A machine
learning model should be continuously monitored
and evaluated regarding performance and other
potential issues that can be resolved. Regular
retraining may be required given changing data
patterns and operational needs.

Collaboration and Skills: The use of machine
learning within data centers requires collaboration
among data scientists, IT professionals, and
domain experts. A multidisciplinary approach will
be able to ensure models are aligned with business
objectives and technical constraints.

Ethics: The design and deployment of machine
learning have to be considered in ethical
dimensions, such as data privacy and bias.
Operations of AI must be fair and accountable,
embracing transparent and responsible AI
practices.

Business Development Impact in the USA

Efficient resource usage management is critical for
data centers, as it directly impacts operational
costs, performance, and sustainability. This project
provides significant business value in the following
ways:

1.

Operational Efficiency:

o

Accurate classification of resource usage

enables proactive management of workloads,
preventing resource bottlenecks and over-
utilization.

2.

Cost Optimization:

o

Identifying underutilized or overutilized

resources helps in reallocation and scaling
decisions, reducing unnecessary operational
expenses.

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

136

https://www.theamericanjournals.com/index.php/tajet

3.

Improved Decision-Making:

o

Data-driven insights into resource patterns

support better scheduling and allocation
strategies, improving the reliability of data center
services.

4.

Sustainability:

o

Efficient resource utilization reduces energy

consumption and carbon footprint, aligning
operations with environmental goals.

5.

Fault Prevention:

o

Predicting abnormal usage patterns allows

for early detection of potential issues, minimizing
downtime,

and

maintaining

service-level

agreements (SLAs).

Challenges and Limitations

The deployment of machine learning to data center
resource optimization is not without its limitations
and challenges. Quite a few large ethical worries
remain concerning the use of operational data for
analytics; there is an amount of concerned
sentiment as to the privacy and security of this
data. Also, quite an amount of informed decisions
should be taken so that this does not end up with
biases within that data, leading towards an
outcome that can be unjust and biased. Other
important reasons that may hamper the
performance of the models of machine learning
relate to data quality poor quality means
inaccurate, incomplete, or noisy data. The
prediction resulting from such data would thus be
biased and unreliable. In machine learning,
interpretability for complicated models is hard
since one does not understand why particular
decisions were made. Finally, the models trained
on particular data centers will most likely be
narrow in generalization when applied to other
data centers that are characterized by different
configurations and operational features.

Future Research Directions

To address the limitations, and challenges and
further advance the domain of data center
resource optimization, several future research
directions can be explored. For instance, the use of
even larger and more diverse data sets may be one
venue to improve the models by providing higher
accuracy and generalizability. This could be made
possible through data center collaborations where
anonymized and aggregated data can be shared,
creating a larger dataset that can be
representative. Another exciting research area is
the

development

of

real-time

resource

optimization

and

predictive

maintenance

techniques. Using advanced machine learning
algorithms coupled with real-time streams of data,
timely decisions can be made that will go a long
way in improving resource utilization and system
reliability. That implies developing sophisticated
anomaly detection systems that can identify issues
well in advance before they strike or predictive
maintenance models that can forecast failures
quite accurately.

CONCLUSION

This study aimed to explore how different ML
techniques can be used for optimizing resource
utilization by data centers in the US, focusing on
strategies to handle incomplete data and
implications for business development. The
dataset was retrieved from the GitHub repository,
which provided a rich dataset of resource usage
metrics and operational data from U.S. data
centers. It contained complex and fine-grained
information that was necessary to optimize data
center performance and deal with incomplete data
challenges. A detailed description of the dataset
and its key attributes was provided. It was
designed for analyzing resource usage patterns in
data centers, putting much emphasis on energy
efficiency, workload distribution, and operational
reliability. This integration of time-series data with
sensor readings and performance logs provided a

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

137

https://www.theamericanjournals.com/index.php/tajet

comprehensive overview of resource consumption
and environmental conditions in data center
operations. This dataset was curated for the
engagement of machine learning models in the
study and optimization of resource consumption
along with the challenges of missing data. Analysis
of resource utilization in US data centers was
accomplished using the application of various
models for machine learning, most notably,
Logistic Regression, Random Forest, and Support
Vector Machines; Retrospectively, the Random
Forest and SVM models seem to be robust and
reliable, placing the Random Forest slightly above,
given their performance is nearly perfect for
training and testing. The application of machine
learning techniques holds huge potential for the
reformation of resource management in US data
centers. These models analyze a pattern in
historical data to predict future resource demands,
thus allowing optimized resource allotment and
minimizing operational costs.

REFERENCES

1.

Abdulazeez, D. H., & Askar, S. K. (2024). A Novel
Offloading Mechanism Leveraging Fuzzy Logic
and Deep Reinforcement Learning to Improve
IoT Application Performance in a Three-Layer
Architecture

Within

the

Fog-Cloud

Environment. IEEE Access.

2.

Alam, M., Islam, M. R., & Shil, S. K. (2023). AI-
Based Predictive Maintenance for US
Manufacturing: Reducing Downtime and
Increasing Productivity. International Journal
of Advanced Engineering Technologies and
Innovations, 1(01), 541-567.

3.

Al Mukaddim, A., Mohaimin, M. R., Hider, M. A.,
Karmakar, M., Nasiruddin, M., Alam, S., &
Anonna, F. R. (2024). Improving Rainfall
Prediction Accuracy in the USA Using
Advanced Machine Learning Techniques.
Journal of Environmental and Agricultural
Studies, 5(3), 23-34.

4.

Al Mukaddim, A., Nasiruddin, M., & Hider, M. A.
(2023). Blockchain Technology for Secure and
Transparent Supply Chain Management: A
Pathway to Enhanced Trust and Efficiency.
International Journal of Advanced Engineering
Technologies and Innovations, 1(01), 419-446.

5.

Buiya, M. R., Laskar, A. N., Islam, M. R.,
Sawalmeh, S. K. S., Roy, M. S. R. C., Roy, R. E. R.
S., & Sumsuzoha, M. (2024). Detecting IoT
Cyberattacks: Advanced Machine Learning
Models for Enhanced Security in Network
Traffic. Journal of Computer Science and
Technology Studies, 6(4), 142-152.

6.

Buiya, M. R., Alam, M., & Islam, M. R. (2023).
Leveraging Big Data Analytics for Advanced
Cybersecurity: Proactive Strategies and
Solutions. International Journal of Machine
Learning Research in Cybersecurity and
Artificial Intelligence, 14(1), 882-916.

7.

Debnath, P., Karmakar, M., Khan, M. T., Khan, M.
A., Al Sayeed, A., Rahman, A., & Sumon, M. F. I.
(2024). Seismic Activity Analysis in California:
Patterns, Trends, and Predictive Modeling.
Journal of Computer Science and Technology
Studies, 6(5), 50-60.

8.

Ezeigweneme, C. A., Umoh, A. A., Ilojianya, V. I.,
& Adegbite, A. O. (2024). Telecommunications
energy

efficiency:

optimizing

network

infrastructure for sustainability. Computer
Science & IT Research Journal, 5(1), 26-40.

9.

Feng, Y., Qi, Y., Li, H., Wang, X., & Tian, J. (2024,
July). Leveraging federated learning and edge
computing for recommendation systems
within cloud computing networks. In Third
International Symposium on Computer
Applications and Information Systems (ISCAIS
2024) (Vol. 13210, pp. 279-287). SPIE.

10.

Gazi, M. S., Nasiruddin, M., Dutta, S., Sikder, R.,
Huda, C. B., & Islam, M. Z. (2024). Employee
Attrition Prediction in the USA: A Machine

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

138

https://www.theamericanjournals.com/index.php/tajet

Learning Approach for HR Analytics and Talent
Retention Strategies. Journal of Business and
Management Studies, 6(3), 47-59.

11.

Gebreyesus, Yibrah, Damian Dalton, Sebastian
Nixon, Davide De Chiara, and Marta Chinnici.
"Machine

learning

for

data

center

optimizations: feature selection using Shapley
additive exPlanation (SHAP)." Future Internet
15, no. 3 (2023): 88.

12.

Gong, Y., Huang, J., Liu, B., Xu, J., Wu, B., & Zhang,
Y. (2024). Dynamic resource allocation for
virtual machine migration optimization using
machine

learning.

arXiv

preprint

arXiv:2403.13619.

13.

Hasanuzzaman, M., Hossain, S., & Shil, S. K.
(2023). Enhancing Disaster Management
through AI-Driven Predictive Analytics:
Improving Preparedness and Response.
International Journal of Advanced Engineering
Technologies and Innovations, 1(01), 533-562.

14.

He, N., Yang, S., Li, F., Trajanovski, S., Zhu, L.,
Wang, Y., & Fu, X. (2023). Leveraging deep
reinforcement

learning

with

attention

mechanism for virtual network function
placement and routing. IEEE Transactions on
Parallel and Distributed Systems, 34(4), 1186-
1201.

15.

Islam, M. R., Shawon, R. E. R., & Sumsuzoha, M.
(2023). Personalized Marketing Strategies in
the US Retail Industry: Leveraging Machine
Learning for Better Customer Engagement.
International Journal of Machine Learning
Research in Cybersecurity and Artificial
Intelligence, 14(1), 750-774.

16.

Islam, M. R., Nasiruddin, M., Karmakar, M.,
Akter, R., Khan, M. T., Sayeed, A. A., & Amin, A.
(2024). Leveraging Advanced Machine
Learning Algorithms for Enhanced Cyberattack
Detection on US Business Networks. Journal of
Business and Management Studies, 6(5), 213-

224.

17.

Islam, M. Z., Islam, M. S., Al Montaser, M. A.,
Rasel, M. A. B., Bhowmik, P. K., & Dalim, H. M.
(2024). EVALUATING THE EFFECTIVENESS OF
MACHINE LEARNING ALGORITHMS IN
PREDICTING CRYPTOCURRENCY PRICES
UNDER MARKET VOLATILITY: A STUDY
BASED ON THE USA FINANCIAL MARKET. The
American Journal of Management and
Economics Innovations, 6(12), 15-38.

18.

Karmakar, M., Debnath, P., & Khan, M. A.
(2024). AI-Powered Solutions for Traffic
Management in US Cities: Reducing Congestion
and Emissions. International Journal of
Advanced Engineering Technologies and
Innovations, 2(1), 194-222.

19.

Khan, M. T., Akter, R., Dalim, H. M., Sayeed, A. A.,
Anonna, F. R., Mohaimin, M. R., & Karmakar, M.
(2024). Predictive Modeling of US Stock
Market and Commodities: Impact of Economic
Indicators and Geopolitical Events Using
Machine. Journal of Economics, Finance and
Accounting Studies, 6(6), 17-33.

20.

Krothapalli, B., Shanmugam, L., & Mohammed,
S. B. (2023). Machine Learning Algorithms for
Efficient Storage Management in Resource-
Limited Systems: Techniques and Applications.
Journal of Artificial Intelligence Research and
Applications, 3(1), 406-442.

21.

Kumari, S. (2024). Cloud Transformation for
Mobile Products: Leveraging AI to Automate
Infrastructure Management, Scalability, and
Cost Efficiency. Journal of Computational
Intelligence and Robotics, 4(1), 130-151.

22.

Li, H., Wang, X., Feng, Y., Qi, Y., & Tian, J. (2024).
Integration Methods and Advantages of
Machine

Learning

with

Cloud

Data

Warehouses.

International

Journal

of

Computer

Science

and

Information

Technology, 2(1), 348-358.

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

139

https://www.theamericanjournals.com/index.php/tajet

23.

Ma, Y., Shen, Z., & Shen, J. (2024). Cloud
Computing and Hyperscale Data Centers: A
Comparative Study of Usage Patterns. Journal
of Theory and Practice of Engineering Science,
4(06), 11-19.

24.

Nasiruddin, M., Al Mukaddim, A., & Hider, M. A.
(2023). Optimizing Renewable Energy Systems
Using Artificial Intelligence: Enhancing
Efficiency and Sustainability. International
Journal of Machine Learning Research in
Cybersecurity and Artificial Intelligence, 14(1),
846-881.

25.

Okusi, O. (2024). Leveraging AI and machine
learning for the protection of critical national
infrastructure. Asian Journal of Research in
Computer Science, 17(10), 1-11.

26.

Pro-AI-Rokibul.

(2024).

Enhancing-Data-

Center-Operations-with-Machine-Learning-
Addressing-Missing-Resource-Usage-
Data/README.md

at

main

·

proAIrokibul/Enhancing-Data-Center-
Operations-with-Machine-Learning-
Addressing-Missing-Resource-Usage-Data.
GitHub.
https://github.com/proAIrokibul/Enhancing-
Data-Center-Operations-with-Machine-
Learning-Addressing-Missing-Resource-
Usage-Data/blob/main/README.md

27.

Rahman, A., Debnath, P., Ahmed, A., Dalim, H.
M., Karmakar, M., Sumon, M. F. I., & Khan, M. A.
(2024). Machine learning and network
analysis for financial crime detection: Mapping
and identifying illicit transaction patterns in
global black money transactions. Gulf Journal
of Advance Business Research, 2(6), 250-272.

28.

Rahman, M. K., Dalim, H. M., & Hossain, M. S.
(2023). AI-Powered Solutions for Enhancing
National Cybersecurity: Predictive Analytics
and Threat Mitigation. International Journal of
Machine Learning Research in Cybersecurity

and Artificial Intelligence, 14(1), 1036-1069.

29.

Shawon, R. E. R., Dalim, H. M., Shil, S. K., Gurung,
N., Hasanuzzaman, M., Hossain, S., & Rahman,
T. (2024). Assessing Geopolitical Risks and
Their Economic Impact on the USA Using Data
Analytics. Journal of Economics, Finance and
Accounting Studies, 6(6), 05-16.

30.

Shawon, R. E. R., Miah, M. N. I., & Islam, M. Z.
(2023). Enhancing US Education Systems with
AI: Personalized Learning and Academic
Performance Prediction. International Journal
of Advanced Engineering Technologies and
Innovations, 1(01), 518-540.

31.

Shawon, R. E. R., Rahman, A., Islam, M. R.,
Debnath, P., Sumon, M. F. I., Khan, M. A., & Miah,
M. N. I. (2024). AI-Driven Predictive Modeling
of US Economic Trends: Insights and
Innovations. Journal of Humanities and Social
Sciences Studies, 6(10), 01-15.

32.

Shawon, R. E. R., Chowdhury, M. S. R., &
Rahman, T. (2023). Transforming Urban Living
in the USA: The Role of IoT in Developing Smart
Cities. International Journal of Machine
Learning Research in Cybersecurity and
Artificial Intelligence, 14(1), 917-953.

33.

Sumon, M. F. I., Osiujjaman, M., Khan, M. A.,
Rahman, A., Uddin, M. K., Pant, L., & Debnath, P.
(2024). Environmental and Socio-Economic
Impact Assessment of Renewable Energy
Using Machine Learning Models. Journal of
Economics, Finance and Accounting Studies,
6(5), 112-122.

34.

Shil, S. K., Chowdhury, M. S. R., Tannier, N. R.,
Tarafder, M. T. R., Akter, R., Gurung, N., & Sizan,
M. M. H. (2024). Forecasting Electric Vehicle
Adoption in the USA Using Machine Learning
Models. Journal of Computer Science and
Technology Studies, 6(5), 61-74.

35.

Wang, Y., Bao, Q., Wang, J., Su, G., & Xu, X.

THE USA JOURNALS

THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN

–

2689-0984)

VOLUME 06 ISSUE12

140

https://www.theamericanjournals.com/index.php/tajet

(2024). Cloud Computing for Large-Scale
Resource Computation and Storage in Machine
Learning. Journal of Theory and Practice of
Engineering Science, 4(03), 163-171.

36.

Zhang, Y., Liu, B., Gong, Y., Huang, J., Xu, J., &
Wan, W. (2024, April). Application of machine
learning optimization in cloud computing
resource scheduling and management. In
Proceedings of the 5th International
Conference on Computer Information and Big
Data Applications (pp. 171-175).