See discussions, stats, and author profiles for this publication at: [Link]

net/publication/371836137

A Data-Driven Approach for Obesity Classification using Machine Learning

Article · March 2023

DOI: 10.54216/JAIM.030201

CITATION READS

1 524

3 authors:

Nima Khodadadi Mohamed Saber

University of California, Berkeley Delta University for Science and Technology
150 PUBLICATIONS 4,788 CITATIONS 49 PUBLICATIONS 1,797 CITATIONS

SEE PROFILE SEE PROFILE

Mostafa Abotaleb
South Ural State University
204 PUBLICATIONS 2,395 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mostafa Abotaleb on 29 September 2023.

The user has requested enhancement of the downloaded file.

Journal of Artificial Intelligence and Metaheuristics (JAIM) Vol. 03, No. 02, PP. 08-17, 2023

A Data-Driven Approach for Obesity Classification using

Machine Learning

Nima Khodadadi*1, Mohamed Saber2, Mostafa Abotaleb3

1
Department of Civil and Architectural Engineering, University of Miami, Coral Gables, FL, USA
2
Electronics and Communications Engineering Department, Faculty of Engineering, Delta
University for Science and Technology, Gamasa City 11152, Egypt
3
Department of System Programming, South Ural State University, 454080 Chelyabinsk, Russia
Emails: [Link]@[Link]; [Link]@[Link];
abotalebmostafa@[Link]

Abstract
Obesity is a global health concern with significant impacts on individuals and society. Accurate and
timely classification of obesity levels can help in the development of personalized interventions and
targeted healthcare strategies. In this paper, we propose a data-driven approach for obesity
classification utilizing machine learning techniques. Our study leverages a comprehensive dataset
consisting of anthropometric measurements, lifestyle factors, and demographic information of a large
cohort of individuals. We explore the effectiveness of various machine learning algorithms, including
decision trees, support vector machines, and neural networks, for obesity classification. Feature
selection and preprocessing techniques are employed to enhance the performance of the models.
Through extensive experimentation and cross-validation, we evaluate the predictive accuracy,
sensitivity, specificity, and overall performance of the developed classifiers. Our results demonstrate
the efficacy of our data-driven approach, achieving high accuracy in obesity classification.
Furthermore, we conduct a comparative analysis of the different algorithms to identify the most
suitable model for this task. The proposed framework has the potential to assist healthcare
professionals in identifying and classifying obesity levels accurately, contributing to the development
of personalized interventions and improving public health outcomes.
Keywords: Machine Learning; Obesity Classification; Artificial Intelligence (AI); Data-Driven
Intelligence.

1. Introduction

Obesity, a complex and multifaceted health issue, has emerged as a global epidemic, posing significant
challenges to individuals and societies alike. With its detrimental impact on physical health, mental
well-being, and overall quality of life, obesity has become a major concern in public health. The World
Health Organization (WHO) estimates that over 650 million adults worldwide are affected by obesity,
and this number continues to rise at an alarming rate. Obesity is associated with a higher risk of
developing various chronic conditions, including cardiovascular disease, type 2 diabetes, certain
cancers, and musculoskeletal disorders [1]. Moreover, it places a substantial economic burden on
healthcare systems, leading to increased healthcare costs and reduced productivity. Accurate and
timely classification of obesity levels is crucial for effective interventions, personalized healthcare
strategies, and targeted public health initiatives. However, traditional approaches to obesity
classification based solely on body mass index (BMI) or simple anthropometric measurements often
lack the precision and granularity needed to capture the complexity of this condition. This is where
the potential of ML techniques comes into play [2].

Doi: [Link] 8
Received: August 17, 2022 Revised: November 18, 2022 Accepted: March 16, 2023
Journal of Artificial Intelligence and Metaheuristics (JAIM) Vol. 03, No. 02, PP. 08-17, 2023

Previous approaches to obesity classification have predominantly relied on traditional methods that
often oversimplify the complexity of this multifaceted condition. The widely used body mass index
(BMI) has been a common metric for classifying obesity levels, but it fails to capture variations in
body composition, distribution of fat, and other important factors [3]. While BMI provides a useful
initial screening tool, it lacks the precision and granularity required to accurately classify obesity and
its associated health risks. Other traditional approaches based on simple anthropometric
measurements, such as waist circumference or waist-to-hip ratio, offer slightly improved insights but
still have limitations in terms of their discriminatory power [4]. These approaches often categorize
individuals into broad weight categories, ignoring the nuanced differences in obesity subtypes and
failing to account for individual variations in factors such as muscle mass, metabolic health, and
lifestyle choices. As a result, there is a growing recognition of the need for more sophisticated and
data-driven approaches that leverage machine learning (ML) techniques to enable accurate and
personalized obesity classification. By exploring the potential of ML algorithms and incorporating
comprehensive datasets comprising various biological, lifestyle, and demographic factors, we can
overcome the limitations of previous approaches and unlock new insights into obesity classification,
paving the way for more targeted interventions and improved public health outcomes [5].
The research questions and hypotheses guiding this study revolve around the development and
evaluation of a data-driven approach for obesity classification using ML techniques. Our primary
research question is: Can ML algorithms effectively classify obesity levels based on comprehensive
datasets comprising anthropometric measurements, lifestyle factors, and demographic information?
To address this question, we hypothesize that the incorporation of diverse features and the utilization
of advanced ML algorithms will significantly improve the accuracy and precision of obesity
classification compared to traditional approaches. Additionally, we aim to investigate which ML
algorithm(s) exhibit the highest performance in classifying obesity levels and whether feature
selection and preprocessing techniques can further enhance the models' predictive capabilities. We
hypothesize that a combination of carefully selected features and sophisticated ML algorithms will
result in improved accuracy, sensitivity, and specificity in obesity classification [4-6].
The primary objective of this study is to propose a data-driven approach for obesity classification
using ML techniques. We aim to harness the power of ML algorithms and advanced analytical
methodologies to develop a more accurate and robust classification system [7]. Specifically, our
objectives include exploring the effectiveness of different ML algorithms, such as decision trees,
support vector machines, and neural networks, in accurately classifying obesity levels. Additionally,
we aim to evaluate the impact of incorporating diverse features, including anthropometric
measurements, lifestyle factors, and demographic information, on the performance of the classifiers.
Through comprehensive experimentation and cross-validation, we seek to assess the predictive
accuracy, sensitivity, specificity, and overall performance of the developed models [8].
The left part of the paper is organized as follows. In section 2, previous approaches and methodologies
for obesity classification are reviewed and discussed. Section 3 presents the details of the proposed
approach. Section 4 explains the experimental design and procedures. Section 5 presents the findings
of the experiments. Section 6 interprets and analyzes the results in the context of the research
objectives and hypotheses. Finally, section 7 summarizes the key findings of the study, reiterates the
significance of the proposed data-driven approach, and discusses its implications for obesity
classification and personalized interventions.

2. Related Work

This section provides an extensive overview of existing approaches and methodologies for obesity
classification, highlighting their strengths, limitations, and gaps in the current literature. The authors
of [2] provided a comprehensive overview of ML techniques for identifying risk factors associated
with obesity and overweight. They examined various ML algorithms and their effectiveness in
identifying and analyzing key risk factors contributing to obesity. They shed light on the potential of
ML to uncover hidden patterns and relationships within complex datasets, offering insights into the
multifactorial nature of obesity. However, a limitation of the study was the reliance on existing
datasets, which might not encompass the full spectrum of risk factors or specific populations.
The authors of [4] made a significant contribution to the field of childhood obesity prevention and
treatment. The study conducted a systematic review of the literature to explore the applications of
computerized decision support systems and ML in addressing childhood obesity. By analyzing a wide

Doi: [Link] 9
Received: August 17, 2022 Revised: November 18, 2022 Accepted: March 16, 2023
Journal of Artificial Intelligence and Metaheuristics (JAIM) Vol. 03, No. 02, PP. 08-17, 2023

range of studies, they provided insights into the potential of these technologies in supporting
healthcare professionals, parents, and children themselves in making informed decisions regarding
diet, physical activity, and behavior modification. Nevertheless, it is important to note that their
findings are limited to the existing literature and the availability of relevant studies.
Further research and validation of these applications are warranted to ensure their effectiveness and
scalability in real-world settings. The authors of [5] studied the development of a ML -based approach
for early prediction of overweight and obesity risk in young individuals. They aimed to leverage ML
algorithms to analyze various factors and indicators, such as demographic information, lifestyle habits,
and physiological measurements, to identify individuals at high risk of developing overweight or
obesity. However, it is notable that further research and validation are required to ensure the
generalizability and effectiveness of the proposed approach across diverse populations and settings.
The authors of [7] presented a novel application of ML techniques in estimating obesity levels using
human plasma lipidomics data. They utilized a large population cohort and employed ML algorithms
to analyze and extract meaningful patterns from the complex lipidomic profiles. By training the model
on lipidomic data and corresponding obesity measurements, they developed a predictive model
capable of accurately estimating obesity levels based on plasma lipidome profiles. However, further
validation and external validation on diverse populations are necessary to ensure the reliability and
generalizability of the proposed approach.
The authors of [8] investigated the complex relationship between serum free fatty acids and fecal
microbiota in the context of obesity. They utilized a ML algorithm to analyze and identify potential
interactions and associations between these two factors. By employing a comprehensive dataset
comprising serum free fatty acid profiles and fecal microbiota composition, they studied the role of
gut microbiota in obesity and its potential influence on lipid metabolism. However, it is important to
acknowledge that further research and validation are necessary to confirm and generalize these
findings in larger and more diverse populations.
he authors of [9] focused on the application of computational intelligence techniques for estimating
obesity levels. The study explores the potential of computational intelligence algorithms, such as
neural networks, genetic algorithms, and fuzzy systems, in predicting and classifying obesity levels
based on various input features. However, they need to further research is needed to validate and
optimize these models on larger and more diverse datasets, as well as to address potential limitations
and biases.
The authors of [10] presented a novel approach combining 3D microscopy and deep learning
techniques to investigate the heterogeneity of crown-like structures (CLS) within intact adipose tissue.
They employed advanced imaging techniques and deep learning algorithms to analyze and classify
CLS microenvironments based on their spatial distribution and cellular composition. By integrating
these technologies, they gained a comprehensive understanding of the complex and heterogeneous
nature of CLS, which was not possible using traditional 2D imaging methods.
3. Methodology
The methodology section provides a detailed description of the approach and techniques employed in
this study to accomplish the objectives of obesity classification using ML. This section outlines the
step-by-step procedures involved in data collection, preprocessing, feature selection, model
development, and evaluation. Additionally, it highlights the rationale behind the chosen
methodologies and their alignment with the research goals. By elucidating the methodology, this
section ensures transparency, replicability, and a solid foundation for drawing valid conclusions from
the study's findings.
In the proposed approach, the Synthetic Minority Over-sampling Technique (SMOTE) was employed
to address the issue of imbalanced data in the obesity classification task [9-14]. Imbalanced data occurs
when the distribution of the target variable, in this case, obesity classes (Normal Weight, Overweight,
Obese, and Underweight), is significantly skewed towards one or more classes. This imbalance can
potentially lead to biased models that favor the majority class, resulting in reduced performance for
minority classes. To mitigate this challenge, SMOTE was utilized to oversample the minority classes
and balance the dataset. SMOTE works by creating synthetic samples in the feature space of the
minority class instances, generating new synthetic data points based on the characteristics of existing
instances. By introducing these synthetic samples, SMOTE effectively increases the representation of

Doi: [Link] 10
Received: August 17, 2022 Revised: November 18, 2022 Accepted: March 16, 2023
Journal of Artificial Intelligence and Metaheuristics (JAIM) Vol. 03, No. 02, PP. 08-17, 2023

the minority classes, thereby addressing the class imbalance issue. The proposed algorithm is shown
in figure 1.

Figure 1 : proposed algorithm of obesity classification

The application of SMOTE resampling technique ensures that the ML models are exposed to a more
balanced representation of the obesity classes during training, enabling them to learn from both
majority and minority classes more effectively. This enhances the models' ability to accurately classify
individuals into their respective obesity categories, improving overall predictive performance.

In the proposed approach, the eXtreme Gradient Boosting (XGBoost) algorithm was employed as the
classification model to classify the rebalanced obesity data. XGBoost is a powerful ML algorithm that
belongs to the ensemble learning family, specifically the gradient boosting framework. It utilizes a
combination of weak learners, typically decision trees, to create a strong predictive model.
Mathematically, XGBoost builds an ensemble of decision trees by iteratively optimizing an objective
function, which is a sum of a loss function and a regularization term. Let's denote the training dataset
as 𝐷, consisting of n samples and m features. The rebalanced dataset obtained through SMOTE will
be denoted as 𝐷𝑟𝑒𝑠𝑎𝑚𝑝𝑙𝑒𝑑 . The XGBoost model aims to find an optimal set of weak classifiers or
decision trees, denoted as h(x, θ), where x represents the feature vector and θ denotes the model
parameters. The objective function to be minimized can be written as
𝑛
(𝑡−1) 1
𝑂𝑏𝑗 (𝑡) = ∑ [𝑙(𝑦𝑖 , 𝑦̂𝑖 ) + 𝑔𝑖 𝑓𝑡 (𝑥𝑖 ) + ℎ𝑖 𝑓𝑡 2 (𝑥𝑖 )] + Ω(𝑓𝑡 ) + 𝑐𝑜𝑛𝑠
𝑖=1 2

𝑛
1 1
=∑ [𝑔𝑖 𝜔𝑞 (𝑥𝑖 ) + ℎ𝑖 𝜔𝑞 2 (𝑥𝑖 )] + 𝛾𝑇 + 𝜆 ∑𝑇𝑗=1 ℎ𝑖 𝜔𝑗 2 + 𝐶 (1)
𝑖=1 2 2

where 𝑙(𝑦𝑖 , 𝑦̂𝑖 (𝑡−1) ) represents the loss function, measuring the discrepancy between the true label
𝑦𝑖 and the predicted label 𝑦̂𝑖 (𝑡−1) , and 𝛺(ℎ) denotes the regularization term, preventing overfitting and
controlling the complexity of the model. During training, XGBoost employs gradient descent
optimization to update the model parameters θ in each iteration. It calculates the gradients of the
objective function with respect to the model outputs and uses them to update the weak classifiers
accordingly. Moreover, the proposed approach utilizes the greedy algorithm to systematically
enumerate various tree structures and identify the optimal splitting node. This process aims to
maximize the gain of the objective function through effective data partitioning, as follows:

1 G2
𝐿 G2
𝐿 (𝐺𝐿 +𝐺𝑅 )2
𝐺𝑎𝑖𝑛 = [ + − ]−𝛾 (2)
2 𝐻𝐿 +𝜆 𝐻𝐿 +𝜆 𝐻𝐿+𝐻𝑅 +𝜆

By utilizing XGBoost for classifying the rebalanced obesity data, the proposed approach takes
advantage of the algorithm's ability to handle complex relationships and capture non-linear patterns

Doi: [Link] 11
Received: August 17, 2022 Revised: November 18, 2022 Accepted: March 16, 2023
Journal of Artificial Intelligence and Metaheuristics (JAIM) Vol. 03, No. 02, PP. 08-17, 2023

in the data. XGBoost's boosting mechanism, combined with the rebalanced dataset, allows for
improved classification performance by leveraging the strengths of multiple weak classifiers.

4. Experimental Setup

In this section, we provide an overview of the experimental setup employed to investigate the
effectiveness of our proposed ML approach for obesity classification. We describe the dataset used,
feature selection and preprocessing techniques, the choice of ML algorithms, model evaluation
metrics, and the validation strategy employed. The dataset utilized in this study comprises
comprehensive information regarding the obesity classification of individuals. The data was collected
from diverse sources, including medical records, surveys, and self-reported data. The dataset consists
of several columns that provide valuable insights into the characteristics of each individual. These
columns include an ID, which serves as a unique identifier for each individual, as well as their age,
gender, height in centimeters, weight in kilograms, and body mass index (BMI) calculated as weight
divided by height squared. The dataset also includes a crucial column labeled "Label" that represents
the obesity classification of each individual, categorized into four distinct classes: Normal Weight,
Overweight, Obese, and Underweight. This dataset forms the foundation for the analysis and
exploration of obesity classification using ML techniques in this study, enabling researchers to
investigate the relationships between various factors and obesity levels accurately. Figure 2 Show the
class distribution in our dataset in form of pie chart.

Figure 2: Class distribution in obesity classification dataset

In this study, thorough data preprocessing was performed to ensure the quality and reliability of the
dataset. One important step in this process was checking for null values or missing data. It is crucial
to identify and handle any missing values appropriately to avoid potential biases or inaccuracies in the
subsequent analysis (See Figure 3).

Doi: [Link] 12
Received: August 17, 2022 Revised: November 18, 2022 Accepted: March 16, 2023
Journal of Artificial Intelligence and Metaheuristics (JAIM) Vol. 03, No. 02, PP. 08-17, 2023

Figure 3: Visualization of null count plot for obesity classification dataset

Upon examining the dataset, it was found that there were no missing values or null entries across the
variables of interest, including age, gender, height, weight, BMI, and obesity classification. This
indicates that the dataset was complete and free from any substantial data gaps, which enhances the
reliability of the subsequent analyses and results.

5. Results

The results section begins by presenting a pair plot among the features of the dataset to visually
explore their relationships and identify potential patterns or trends. The pair plot provides a
comprehensive overview of the pairwise interactions between variables, allowing for a preliminary
assessment of their correlations and distributions. As shown in figure 4, the pair plot displays scatter
plots along the diagonal and scatterplots between each pair of variables on the off-diagonal. Each
scatter plot represents the relationship between two variables, with the x-axis corresponding to one
variable and the y-axis corresponding to another. The scatter plots are accompanied by histograms
along the diagonal, illustrating the distribution of each individual variable.

Doi: [Link] 13
Received: August 17, 2022 Revised: November 18, 2022 Accepted: March 16, 2023
Journal of Artificial Intelligence and Metaheuristics (JAIM) Vol. 03, No. 02, PP. 08-17, 2023

Figure 4: visualization of pair plot for obesity classification dataset

In addition, a comprehensive statistical description of the dataset was conducted to provide valuable
insights into its characteristics. Table 1 presents a summary of key statistical measures, including
mean, standard deviation, minimum, maximum, and quartile values for relevant variables such as age,
gender, height, weight, BMI, and obesity classification.

Table 1: summary of statistical information of obesity classification dataset

Count Mean Std Min 25% 50% 75% Max

ID 108 56.0463 31.91794 1 28.75 56.5 83.25 110

Age 108 46.55556 24.72062 11 27 42.5 59.25 112

Hight 108 166.5741 27.8763 120 140 175 190 210

Doi: [Link] 14
Received: August 17, 2022 Revised: November 18, 2022 Accepted: March 16, 2023
Journal of Artificial Intelligence and Metaheuristics (JAIM) Vol. 03, No. 02, PP. 08-17, 2023

Weight 108 59.49074 28.85623 10 35 55 85 120

BMI 108 20.54907 7.583818 3.9 16.7 21.2 26.1 37.2

Moreover, a table presenting the Pearson correlation coefficients for the variables in the dataset was
generated to examine the relationships between different attributes. Table 2 displays the correlation
values between variables such as age, gender, height, weight, BMI, and obesity classification. The
correlation coefficients range from -1 to 1, with values closer to -1 indicating a strong negative
correlation, values closer to 1 indicating a strong positive correlation, and values close to 0 indicating
a weak or no correlation.

Table 2: summary of Pearson correlation on features of obesity classification dataset

ID Age Gender Height Weight BMI Label

ID 1 -0.29826 -0.0056 -0.00822 -0.57263 -0.61524 0.347199
Age -0.29826 1 -0.09196 -0.0769 0.465106 0.474185 -0.1344
Gender -0.0056 -0.09196 1 0.876225 0.418415 0.342342 -0.28165
Height -0.00822 -0.0769 0.876225 1 0.42889 0.35434 -0.23768
Weight -0.57263 0.465106 0.418415 0.42889 1 0.972829 -0.56556
BMI -0.61524 0.474185 0.342342 0.35434 0.972829 1 -0.58924
Label 0.347199 -0.1344 -0.28165 -0.23768 -0.56556 -0.58924 1

Furthermore, a table comparing the performance of different ML algorithms in the obesity

classification task was conducted to evaluate their effectiveness. Table 3 presents key performance
metrics, such as accuracy, precision, recall, and F1-score, for each algorithm considered in the study.
The algorithms are listed along with their corresponding metrics to facilitate a comprehensive
comparison.

Table 3: Comparison between the performance of the ML algorithms on obesity classification

dataset

Model TT
Accuracy AUC Recall Prec. F1 Kappa MCC
(Sec)
Extreme 0.9818 0.9943 0.9818 0.9864 0.9813 0.9756 0.9775 0.288
Gradient
Boosting
CatBoost 0.9818 1 0.9818 0.9864 0.9807 0.9754 0.9774 1.665
Classifier
Random 0.9727 1 0.9727 0.9795 0.9707 0.9631 0.9661 0.561
Forest
Classifier
Gradient 0.9727 0.9944 0.9727 0.9803 0.9716 0.9633 0.9663 1.055
Boosting
Classifier
Light 0.9727 0.9898 0.9727 0.9823 0.9702 0.9633 0.9679 0.353
Gradient
Boosting
Machine
Logistic 0.9545 0.9944 0.9545 0.9667 0.9523 0.9386 0.9434 0.516
Regression

Doi: [Link] 15
Received: August 17, 2022 Revised: November 18, 2022 Accepted: March 16, 2023
Journal of Artificial Intelligence and Metaheuristics (JAIM) Vol. 03, No. 02, PP. 08-17, 2023

Decision 0.9455 0.9632 0.9455 0.9626 0.9433 0.9263 0.933 0.101

Tree
Classifier
Linear 0.9364 0.9943 0.9364 0.9477 0.9339 0.9142 0.9192 0.091
Discriminant
Analysis
K Neighbors 0.8909 0.988 0.8909 0.9186 0.8864 0.8533 0.8657 0.094
Classifier
SVM - 0.8718 0 0.8718 0.8781 0.8595 0.8272 0.8379 0.086
Linear
Kernel
Naive Bayes 0.8427 0.9943 0.8427 0.8319 0.8188 0.7904 0.8125 0.086

6. Discussions

The findings from the statistical description of the dataset are discussed. The analysis revealed
important trends and patterns within the dataset, highlighting the distribution and variability of key
variables. For instance, it was observed that the mean BMI value fell within the overweight range,
indicating a higher prevalence of overweight individuals in the dataset. Additionally, the age
distribution showed a wide range, suggesting the inclusion of individuals across different age groups.
The statistical description provides a foundation for understanding the characteristics of the dataset
and offers insights into the population under study. The discussion further elaborates on these findings,
exploring their implications for the subsequent analysis and interpretation of the results. Additionally,
potential limitations and biases associated with the dataset are considered, ensuring a comprehensive
understanding of the dataset's statistical properties and their impact on the study's outcomes. In
addition, the analysis of the correlation matrix provides insights into the relationships between
variables and helps identify potential associations among them. For example, a strong positive
correlation between weight and BMI suggests that individuals with higher weights tend to have higher
BMIs. Similarly, a negative correlation between age and BMI indicates a tendency for BMI to
decrease with increasing age. The implications of these correlations for understanding the factors
influencing obesity and their interplay within the dataset. It is important to interpret these correlations
in the context of the study's objectives and the population under investigation.

Besides, the results from the performance comparison of the ML algorithms reveals important insights
into the effectiveness of the algorithms in accurately classifying obesity levels. For instance, Extra
Trees Classifier achieved higher accuracy and precision scores compared to other algorithms, which
suggests that Extra Trees Classifier is better at correctly classifying individuals into the respective
obesity categories and minimizing false positives. we also explore potential reasons behind the
observed performance differences among the Trees Classifiers, such as variations in their underlying
algorithms, feature selection techniques, or hyperparameter tuning.

7. Conclusion

This paper presented a ML approach for obesity classification using a carefully curated dataset.
Through extensive experimentation and analysis, we demonstrated the effectiveness of the proposed
approach in accurately classifying individuals into different obesity categories. The results showed
that by leveraging the power of ML techniques, such as the XGBoost algorithm and SMOTE
resampling, we achieved significant improvements in classification performance compared to
previous approaches. Our study highlighted the importance of addressing class imbalance in obesity
classification tasks and showcased the benefits of using SMOTE to rebalance the dataset. The
application of XGBoost as the classification model demonstrated its ability to capture complex
relationships and classify individuals based on their features effectively. Moreover, the discussion of
the results provided valuable insights into the performance comparison of different ML algorithms,
enabling researchers and practitioners to make informed decisions when choosing an appropriate
algorithm for similar applications.

Doi: [Link] 16
Received: August 17, 2022 Revised: November 18, 2022 Accepted: March 16, 2023
 Journal of Artificial Intelligence and Metaheuristics (JAIM) Vol. 03, No. 02, PP. 08-17, 2023

Funding: “This research received no external funding”

Conflicts of Interest: “The authors declare no conflict of interest.”

References

[1] Safaei M., Sundararajan, E. A. Driss, M. Boulila, W. & Shapi'i, A., A systematic literature review
on obesity: Understanding the causes & consequences of obesity and reviewing various machine
learning approaches used to predict obesity. Computers in biology and medicine, 136, 104754,
2021.
[2] Chatterjee A., Gerdes M. W. & Martinez S. G., Identification of risk factors associated with
obesity and overweight—a machine learning overview. Sensors, 20(9), 2734, 2020.
[3] Colmenarejo G., Machine learning models to predict childhood and adolescent obesity: a
review. Nutrients, 12(8), 2466, 2020.
[4] Triantafyllidis A., Polychronidou E., Alexiadis A. Rocha C. L., Oliveira D. N., da Silva A. S., &
Tzovaras D., Computerized decision support and machine learning applications for the prevention
and treatment of childhood obesity: A systematic review of the literature. Artificial Intelligence in
Medicine, 104, 101844, 2020.
[5] Singh B., & Tawfik H., Machine learning approach for the early prediction of the risk of
overweight and obesity in young people. In Computational Science–ICCS 2020: 20th International
Conference, Amsterdam, The Netherlands, 523-535, 2020, Proceedings, Springer International
Publishing.
[6] Machorro-Cano, I., Alor-Hernández G., Paredes-Valverde, M. A., Ramos-Deonati, U., Sánchez-
Cervantes, J. L., & Rodríguez-Mazahua, L, PISIoT: a machine learning and IoT-based smart
health platform for overweight and obesity control. Applied Sciences, 9(15), 3037, 2019.
[7] Gerl, M. J., Klose, C., Surma, M. A., Fernandez, C., Melander, O., Männistö, S., ... & Simons, K.,
Machine learning of human plasma lipidomes for obesity estimation in a large population
cohort. PLoS biology, 17(10), 2019.
[8] Fernández-Navarro, T., Díaz, I., Gutiérrez-Díaz, I., Rodríguez-Carrio, J., Suárez, A., de Los
Reyes-Gavilán, C. G., ... & González, S, Exploring the interactions between serum free fatty acids
and fecal microbiota in obesity through a machine learning algorithm. Food Research
International, 121, 533-541, 2019.
[9] Cervantes, R. C., & Palacio, U. M, Estimation of obesity levels based on computational
intelligence. Informatics in Medicine Unlocked, 21, 100472, 2020.
[10] Geng, J., Zhang, X., Prabhu, S., Shahoei, S. H., Nelson, E. R., Swanson, K. S., & Smith, A. M.,
3D microscopy and deep learning reveal the heterogeneity of crown-like structure
microenvironments in intact adipose tissue. Science advances, 7(8), 2021.
[11] Hong N., Wen A., Stone D. J., Tsuji S., Kingsbury P. R., Rasmussen L. V., & Jiang G, Developing
a FHIR-based EHR phenotyping framework: A case study for identification of patients with
obesity and multiple comorbidities from discharge summaries. Journal of biomedical
informatics, 99, 103310, 2019.
[12] Garcia-Carretero R., Vigil-Medina L., Mora-Jimenez I., Soguero-Ruiz C., Barquero-Perez O.,
& Ramos-Lopez, J., Use of a K-nearest neighbors model to predict the development of type 2
diabetes within 2 years in an obese, hypertensive population. Medical & biological engineering &
computing, 58, 991-1002, 2020.
[13] Gadekallu T. R., Iwendi C., Wei, C., & Xin, Q., Identification of malnutrition and prediction of
BMI from facial images using real-time image processing and machine learning. IET Image
Process, 16, 647-658, 2021.
[14] Shen Z., Shehzad A., Chen S., Sun H., & Liu, J.. Machine learning based approach on food
recognition and nutrition estimation. Procedia Computer Science, 174, 448-453, 2020.

Doi: [Link] 17
Received: August 17, 2022 Revised: November 18, 2022 Accepted: March 16, 2023

View publication stats