Artificial Intelligence-Machine Learning Techniques Promoting

SDG's: An Exploratory Approach

Pratyush Prasad

, Pranjal Prasad

and Asim Prasad

Manipal University Jaipur, India

Ajay Kumar Garg Engineering College, Ghaziabad, India

Amity University, Noida, India

Keywords: SDG, Artificial Intelligence, Machine Learning, Linear Models, Decision Trees, Random Forest, Naïve

Bayes.

Abstract: This article provides an overview of the Machine Learning (ML) techniques, models, and ideas utilized to

analyze datasets to achieve Sustainable Development Goals (SDG) targets. Furthermore, this study

investigates the use of Artificial Intelligence (AI) in facilitating the attainment of SDGs. An exploratory

approach and a concept-centric literature review were employed to address the study issues. The study reveals

the fundamental principles of Supervised, Unsupervised, and Reinforced ML methodologies, as well as

several algorithms like k-Nearest Neighbours, Linear Models, Naïve Bayes, Decision trees, Random forests,

gradient boosted decision trees, Support vector machines, and Neural networks. The study implications

pertain to the use of artificial intelligence and machine learning techniques to advance the aims of the SDGs

for the betterment of society, the economy, and the environment.

1 INTRODUCTION

The seventeen Sustainable Development Goals

(SDGs), agreed by the UN in 2015, call for

collaborative action to end poverty, protect the

environment, and promote peace and prosperity for

all by 2030. However, complex issues related to these

require the application of Artificial Intelligence(AI)

and Machine Learning(ML) algorithms, models to

support SDG objectives(International

Telecommunication Union, 2021). Machine and

system intelligence is called artificial intelligence

(AI). They can perform duties independently and

work with humans and nature due to their

intelligence. AI software can perceive, decide,

forecast, extract information, recognize patterns from

data, communicate, and think logically(Sætra, 2021).

ML as defined by Arthur Samuel is the “field of study

that allows computers to learn without being

explicitly programmed.” The growing field of data

science uses ML to extract knowledge from datasets.

It combines computer science, statistics, and

AI(Müller & Guido, 2016). ML uses computerized

approaches to solve problems using past data and

expertise without changing critical operations

(Sandhu, 2018). Unlike AI, ML involves discovering

patterns in datasets (data mining) and using them to

classify or predict occurrences related to a problem

(Alpaydın, 2004). ML enables intelligent machines to

maintain their talents. Statistical methods train

algorithms to classify or predict, providing data

mining insights. These insights guide application and

business decisions to improve growth indicators. ML

has several applications that relate to real-world

problems supporting automation (Khanum et al.,

2015) in fields related to bioinformatics(Tan &

Gilbert, 2003), Population Genetics (Schrider &

Kern, 2018), autonomous vehicle (AV), healthcare,

natural language processing (NLP), business

applications, intelligent robots, climate modeling,

gaming, voice processing, image processing(Rustam

et al., 2020), cancer detection(Prasad, 2023). ML

involves data storage, abstraction, generalization, and

evaluation(H, 2023). Data Storage stores and

retrieves large amounts of data, essential to ML.

Cognitive abstraction involves obtaining useful

information from a dataset. Developing broad

concepts that include all data is required. Abstraction

(knowledge generation) uses current and new models.

Training establishes model parameters from a dataset.

After training, the model abstracts the data to capture

its key points. Generalizing stored data knowledge

Prasad, P., Prasad, P. and Prasad, A.

Artiﬁcial Intelligence-Machine Learning Techniques Promoting SDG’s: An Exploratory Approach.

DOI: 10.5220/0013405700003882

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 2nd Pamir Transboundary Conference for Sustainable Societies (PAMIR-2 2023), pages 1717-1726

ISBN: 978-989-758-723-8

1717

allows it to be used for future decision-making or

action. These actions should be performed on similar

tasks but not identical to earlier ones. The goal of

generalization is to discover the data traits or qualities

that will be most relevant to future activities.

Evaluation involves systematic feedback to the user

to measure the effectiveness or utility of acquired

knowledge. Feedback is then used to improve

learning. ML system knowledge acquisition involves

decision process, error function, and model

optimization. In light of these deliberations,

considering the growing importance of ML, this

article, through an exploratory approach, aims to

provide answers to the following research questions:

RQ1: What ML techniques, models, and concepts

facilitate dataset analysis for SDG

targets?

RQ2: Does the application of AI promote the

achievement of SDGs?

2 LITERATURE REVIEW

The researcher conducted a critical analysis to present

the ML algorithm, models to understand what is

known about the study topic, the related concepts, and

the perspectives(Grant & Booth, 2009). The ML

algorithms (H, 2023; Müller & Guido, 2016) are of

three types: (1) Supervised ML, (2) Unsupervised

ML, and (3) Reinforced ML, as detailed in Figure 1.

Source: (Polzer, 2021).

Figure 1: Machine Learning Types.

2.1 Supervised ML

Supervised ML is where the user provides the

algorithm with a series of input-output pairs. The

input/output data pair teaches the ML algorithm. The

program finds input-based techniques for output

generation. Supervised ML train models to produce

the desired output using a training set. The training

dataset contains input data and target outputs to help

the model learn iteratively. The approach uses a loss

function to evaluate its performance and iteratively

adjust its settings to minimize error until it reaches a

desirable accuracy. Supervised ML techniques work

in two steps. Analytical tasks begin with training data

analysis. These algorithms then create dependent

functions to map new attribute instances. ML

approaches have found that a training subset of 66%

of the data can achieve the desired result while

minimizing processing needs. (Ng, 2005). The

process diagram of supervised ML is in Figure 2. The

Supervised ML problems are of two types:

Source:

https://ebrary.net/136995/computer_science/machine_learning

Figure 2: Process diagram for Supervised Learning.

(a) Classification: Classification refers to the

systematic procedure of utilizing a model to

make predictions about values that are not yet

known, specifically output variables, by

leveraging a set of known values, namely input

variables(Muhammad & Yan, 2015). It aims to

predict a class label (Figure 3).

Source: https://www.javatpoint.com/classification-

algorithm-in-machine-learning

Figure 3: Class Labelling.

It uses an algorithm to assign test data into specific

categories accurately. It recognizes specific entities

within the Dataset and attempts to draw some

conclusions on how those entities should be labeled

or defined. The classifier is the algorithm that is

utilized to classify a given dataset. Classification

problems are either binary classification or multiclass

classification. A binary classifier has two possible

PAMIR-2 2023 - The Second Pamir Transboundary Conference for Sustainable Societies- | PAMIR

1718

outcomes: Yes or No, Male or Female, etc. The

multiclass classifier has more than two outcomes, like

classifying fruits, vegetables etc. Figure 4 shows a

general classification architecture.

Figure 4: General Classification Architecture.

(b) Regression: This technique helps understand

dependent and independent variables. Predicting

a real or continuous number is the goal. Common

regression methods include linear, logistic, and

polynomial. If the outcome is continuous, it's

regression; otherwise, classification. Different

ML algorithms (Bhavsar & Ganatra, 2012) are k-

Nearest Neighbors, Linear Models, Naïve Bayes,

Decision trees, Random forests, Gradient

boosted decision trees, Support vector machines,

and Neural networks( Deep Learning). These are

eleborrated herein.

(i) k-Nearest Neighbors(KNN): K-nearest

neighbor (Lindholm et al., 2019) is an

instance-based supervised ML learning

method that does not rely on parameters and

is one of the simplest to understand using

small datasets. The idea is that "like"

samples tend to cluster together. K-nearest

neighbor classifiers are used to determine

the most common class label given an

unlabeled sample by searching the pattern

space for the k-objects that are closest to it

(Burges, 1998). If k=1, then the unknown

sample is placed in the training sample class

that most closely matches it in the pattern

space. Figure 5 shows the KNN.

Source: https://www.geeksforgeeks.org/k-nearest-neighbours/

Figure 5: KNN Visualization.

(ii) Linear models:

Linear models (Matloff,

2017) are another machine-learning

technique class that explicitly learns from

labeled datasets and maps the data points to

the best-performing linear functions. This

can be applied for prediction purposes on

large multidimensional datasets.

(iii) Naïve Bayes: A category of supervised

learning algorithms known as naive Bayes

methods utilize Bayes' theorem with the

"naive" assumption that each pair of features

is conditionally independent given the value

of the class variable(Maertens et al., 2017).

It is used for Sentiment Analysis, Text

Classification, Credit Scoring, Medical Data

Classification, and Text Filtering for Spam.

The naive Bayesian classifiers presume that,

given the class variables, the value of one

characteristic is independent of the value of

any other characteristic. Accordingly, the

following equation gives the posterior

probability (Gianey & Choudhary, 2018).

(1)

Where,

(iv) Decision trees: A decision tree is a graphical

representation used to facilitate decision-

making or produce numerical forecasts by

utilizing the information contained within a

given dataset. It is a form of supervised

learning methodology that is utilized to

make predictions about response values,

achieved by acquiring decision rules formed

from the Dataset's features. The decision tree

comprises three essential components: a root

node, leaf nodes, and branches. Regardless

of the precise form of the decision tree

employed, the process invariably

commences with a distinct decision. The

choice is visually represented by a box

serving as the root node. A tree structure's

root and leaf nodes include inquiries or

criteria that necessitate a response. Nodes

are typically observed in the form of squares

or circles. In this context, squares are

utilized to symbolize decisions, whilst

circles are employed to indicate ambiguous

outcomes(Müller & Guido, 2016) . Decision

Tree models can potentially be employed in

both regression and classification scenarios.

Artiﬁcial Intelligence-Machine Learning Techniques Promoting SDG’s: An Exploratory Approach

1719

Hence, they are frequently described as

Classification And Regression Trees

(CART). Decision trees are utilized in the

medical domain for diagnostic purposes and

in risk management, personal management,

corporate strategy, financial management,

and project management requirements.

Figure 6 shows a Decision Tree structure.

Source:(Uddin et al., 2019).

Figure 6: Decision Tree.

(v) Random forests: The Random Forest

algorithm (Schrider & Kern, 2018)

integrates the predictions of numerous

decision trees in order to arrive at a

consolidated outcome, as shown in Figure 7.

Source:(David, 2020).

Figure 7: Random Forest.

The high level of user-friendliness and

adaptability of this tool has significantly

contributed to its widespread adoption since

it effectively addresses both classification

and regression tasks. The Random Forest

algorithm is extensively employed in

domains such as E-commerce, banking,

medicine, and the stock market. In the

context of the Banking business, this

technique can be employed to identify

customers who are likely to fail on a loan. It

is employed to forecast the factors that

contribute to optimal functioning. It helps

predict customer behavior and evaluate

medical records.

(vi) Gradient-boosted decision trees: Gradient-

boosted decision trees (Natekin & Knoll,

2013), also referred to as Gradient boosting

machine (GBM) or Gradient Boosted

Regression Tree (GBRT) represent a

machine learning methodology aimed at

enhancing the prediction efficacy of a model

by iteratively refining the learning

process(Chen & Guestrin, 2016; Z. Zhang &

Jung, 2021). The concept of Gradient refers

to the rate of change of a function with

respect to its independent Boosting is

particularly advantageous in scenarios when

the data has a lower number of dimensions,

where a basic linear model exhibits poor

performance, interpretability is of lesser

importance, and there are no stringent

constraints on latency. Boosting algorithms

have demonstrated their suitability for

artificial intelligence projects in several

industries, including a wide spectrum of

sectors. These algorithms have proven to be

effective and efficient in enhancing the

performance of AI systems. In healthcare,

boosting techniques are employed to

mitigate errors in the prediction of medical

data, namely in areas like the estimation of

cardiovascular risk factors and the prognosis

of cancer patient survival rates. Gradient

boosting is a commonly employed technique

in the field of marketing to optimize the

allocation of budget across various channels,

with the ultimate goal of maximizing the

return on investments. Figure 8 shows a

GBM learning model.

Source:

(T. Zhang et al., 2021).

Figure 8: GBM learning model.

(vii) Support vector machine(SVM): The SVM

(Vanneschi & Silva, 2023) is a highly

effective ML technique that has

demonstrated its versatility in numerous

applications, encompassing text

classification, picture classification, spam

PAMIR-2 2023 - The Second Pamir Transboundary Conference for Sustainable Societies- | PAMIR

1720

detection, handwriting identification, gene

expression analysis, face detection, and

anomaly detection. SVMs exhibit

adaptability and efficiency across various

applications because they can effectively

handle high-dimensional data and capture

nonlinear correlations. Its algorithms have

demonstrated high efficacy in identifying

the largest separation hyperplane across

several classes within the target

feature(Awad & Khanna, 2015; Jakkula,

2011). Figure 9 displays SVM for binary

classification of data.

Source: Source:(Fan et al., 2021).

Figure 9: SVM for Binary Classification.

(viii) Neural networks: A neural network

(Gurney, 1997) is a type of ML model that is

specifically designed to replicate the

functional and structural characteristics of

the human brain. Neural networks, also

known as artificial neural networks (ANNs)

or deep neural networks have gained

significant popularity as machine learning

techniques that aim to replicate the learning

mechanisms observed in biological

organisms(Charu C. Aggarwal, 2018). This

deep learning technology falls within the

broader domain of AI.. Many people use

ANN to learn computers in a way that is

similar to how living things learn. The cells

that make up the nervous system are called

neurons. Axons and dendrites link the

neurons to each other. The areas where

axons and dendrites meet are called

synapses. Synaptic connections often

change how strong they are in reaction to

things outside of the brain. Figure 10

displays the biological neural network

(BNN) connections. The way living things

learn changed because of this. Neural

networks can build complex models for

large datasets composed of interconnected

neurons, which work together to address

complex challenges. Neural networks are

extensively employed throughout many

domains, encompassing image

identification, predictive modeling, and

natural language processing (NLP). Since

2000, notable instances of commercial

applications have demonstrated

considerable significance(Mijwel et al.,

2019). These include using handwriting

recognition for cheque processing,

transcribing speech into text, analyzing data

for oil and gas exploration, predicting

weather patterns, and implementing facial

recognition technology. Neural networks are

sensitive to parameter choice and data

scaling. Figure 11 shows an ANN general

structure.

Source: (Westbrook et al., 2010).

Figure 10: BNN.

Source: (Sairamya et al., 2019).

Figure 11: ANN General Structure.

2.2 Unsupervised ML

An unsupervised ML algorithm has no known output

and teacher[1]. Unsupervised ML approaches can

find patterns and relationships in a dataset without a

preset output variable, making them useful for

description jobs. This type of ML classification is

called unsupervised ML because there is no response

variable to guide the study. Unsupervised ML seeks

latent dimensions, components, clusters, and

trajectories in data. Principle components analysis,

Artiﬁcial Intelligence-Machine Learning Techniques Promoting SDG’s: An Exploratory Approach

1721

factor analysis, and mixture modeling are

unsupervised learning methods(Hastie et al., 2006).

Source: (Morimoto & Ponton, 2021).

Figure 12: Supervised and Unsupervised Learning Model.

Figure 12 depicts a typical supervised and

unsupervised learning model. The types of

unsupervised ML algorithms are discussed

herein.

(a) Dataset Transformation: Dataset

Transformation is the process of

transforming a dataset for ease of

understanding. This involves reducing the

dimension of the data set with high

dimensions and many features with fewer

features without compromising its essential

characteristics.

(b) Clustering: Clustering is a data mining

technique that facilitates the organization of

unlabelled data by grouping them based on

their commonalities or

differences(Alloghani et al., 2020).

Clustering refers to the procedure of

grouping a Dataset based on the similarity of

its components. Clustering algorithms

(Khanum et al., 2015) generate distinct

groups characterized by similar items within

the various categories. Clustering techniques

are utilized for the analysis of raw and

unclassified data elements, aiming to

arrange them into separate groups that

exhibit inherent structures or patterns within

the data. Figure 13 shows a clustering

model. Clustering methods can be

categorized into various classes, including

exclusive, overlapping, hierarchical, and

probabilistic.

Source: (Dutt et al., 2019).

Figure 13: Clustering Model.

(i) Exclusive clustering: Exclusive clustering is

a type of clustering that enforces the

constraint that each data point can belong to

only one cluster. The K-means clustering

algorithm exemplifies the concept of

exclusive clustering.

Pattern identification,

image analysis, consumer analytics, market

segmentation, social network analysis, and

many more domains apply the clustering

technique.

(ii) Overlapping: Overlapping clustering

techniques permit data points to be attached

to multiple clusters. Partitioning methods

are more prevalent than overlapping

clustering algorithms due to their simplicity

and effectiveness on large datasets.

(iii) Hierarchical clustering: Hierarchical

Clustering aims to identify inherent clusters

by considering the Dataset's attributes. The

primary objective of the hierarchical

clustering algorithm is to identify and

construct a hierarchical structure that reveals

the presence of nested groups within the

Dataset. The concept resembles the

biological taxonomy in classifying

organisms within the plant or animal

kingdom. Hierarchical clusters are typically

depicted through a hierarchical tree structure

referred to as a dendrogram. Hierarchical

clustering can be classified into two distinct

approaches: agglomerative, also known as

the bottom-up technique, and divisive, also

referred to as the top-down approach.

(iv) Probabilistic clustering: Probabilistic

clustering is a technique that addresses

density estimation or "soft" clustering

challenges. It is utilized for the purpose of

addressing density estimation or "soft"

clustering difficulties. Probabilistic

clustering is a technique wherein data points

PAMIR-2 2023 - The Second Pamir Transboundary Conference for Sustainable Societies- | PAMIR

1722

are grouped based on their probability of

belonging to a specific distribution.

unsupervised learning methodology that

examines the interdependence between

different data elements and is designed to

enhance cost-effectiveness. Market basket

analysis is a commonly employed approach

for examining the association between

various items, enabling firms to enhance

their comprehension of these relationships.

(d) Apriori algorithms: Apriori is an

unsupervised learning methodology due to

its frequent application in identifying and

extracting remarkable patterns and

associations. The Apriori algorithm can be

adapted to perform classification tasks using

labeled data. The popularity of Apriori

algorithms has been primarily driven by

their application in market basket research,

resulting in the development of diverse

recommendation engines for music

platforms and online stores.

2.3 Reinforced ML

ML is a scientific study of decision-making, and

reinforcement ML (Naeem et al., 2020; Sutton &

Barto, 2018) is a subfield. A software agent interacts

with an unknown environment. The agent's behaviors

reveal the environment's dynamics. To maximize

reward, obtain the best actions in a specific context.

It involves using appropriate measures to maximize

advantages in a given setting. This self-teaching

system learns by trial and error. The agent optimizes

rewards through trial-and-error behavior. As the

reinforcement agent chooses how to complete the job,

reinforcement learning has no predetermined answer.

The machine may learn from its own experiences if

there is no training dataset. This data is collected

using trial-and-error ML techniques. Reinforcement

learning algorithms learn from observed outcomes

and select the best action. After each step, the

algorithm receives feedback to determine whether its

choice was correct, neutral, or erroneous. This

technique is useful in automated systems that must

make many incremental judgments without human

interaction. Reinforced Learning has an agent,

environment, policy, reward signal, and value

function. Environment models may also be present. A

common reinforcement learning model is in Figure

14. Reinforcement learning is used in robotics,

autonomous control, healthcare, communication and

networking, gaming, natural language processing,

scheduling management, and self-organized

systems(Naeem et al., 2020).

Source: (Mahesh Batta, 2020).

Figure 14: Reinforced Learning.

3 AI-ML AND SDGs

(Vinuesa et al., 2020) grouped SDGs into societal,

economic, and environmental categories, arguing that

the increasing influence of AI through technological

improvement will likely enable 134 (79 %) targets

across all SDGs. Sixty-seven societal targets (82%)

benefit from AI that positively impact the provision

of energy services, water, food, good health, and low

carbon systems, promoting a circular economy.

Forty-two economic group targets (70%) will benefit

from AI-enabled technologies related to decent work,

industry innovation, infrastructure, and inequality.

Further, twenty five targets (93%) from the

environmental group will be positively impacted.

These relate to life below water, life on land, and

climate action. As AI may inhibit some targets, it is

necessary to overcome safety, transparency, and

ethical standards and put regulatory standards in place

to plug perceived gaps.(Sætra, 2021) supports these

arguments, viewing AI as part of a sociotechnical

system that includes bigger structures economic and

political processes, not as a standalone tool. (Nasir et

al., 2023) imply that while the advancement of AI

technology is concentrated on enhancing present

economic growth, significant societal and

environmental challenges may be overlooked. Thus

AI is poised to assume a more prominent position in

the field of achieving SDGs. Its potential to provide

support and facilitate coordination in this domain is

expected to grow significantly in the future(Leal

Filho et al., 2023).

Artiﬁcial Intelligence-Machine Learning Techniques Promoting SDG’s: An Exploratory Approach

1723

4 FINDINGS AND CONCLUSION

The findings briefly present the answers to the

research questions.

RQ1: What ML techniques, models, and concepts

facilitate dataset analysis for SDG targets?

In a general sense, machine learning encompasses

the use of diverse models to discern patterns inside

data and subsequently generate precise predictions by

leveraging the observed patterns. These subjects are

interconnected with supervised learning, a technique

that uses training data to teach the model.

Generalization refers to the ability of a machine

learning model to make accurate predictions about

unseen data based on its training data. Overfitting

occurs when a model exhibits excessive fidelity to the

training data, diminishing the ability to generalize to

unseen data. Underfitting occurs when the model fails

to predict outcomes accurately using training and

fresh, unseen data. Supervised learning involves

partitioning data into three distinct categories:

training, development, and testing datasets. The test

dataset is employed post-model development to

evaluate the model's performance on previously

unseen data.

Additionally, the selection of pertinent fields

within a dataset was deliberated upon. Subsequently,

an analysis was conducted on ANN, which serve as

the first model inside this sequence of blog entries.

Neural networks typically consist of three layers: an

input layer, a hidden layer, and an output layer.

Neural networks emerged as pioneering machine

learning models, with subsequent investigations

delving into numerous versions of this paradigm. The

utilization of several hidden layers in deep neural

networks enhances their performance in certain tasks

compared to simple neural networks, owing to their

inherent complexity.

The research presented Supervised,

Unsupervised, and Reinforced ML algorithms.

Under Supervised ML, the classification and

regression models were deliberated. The different ML

algorithms presented are k-Nearest Neighbors, Linear

Models, Naïve Bayes, Decision trees, Random

forests, Gradient boosted decision trees, Support

vector machines, and Neural networks. The types of

unsupervised ML and reinforced ML with concepts

were discussed.

RQ2: Does the application of AI promote the

achievement of SDGs?

The findings reveal that AI-ML research

supports and promotes the achievement of SDG

targets. AI-ML helps to model worldwide complex

challenges related to overcoming poverty, inequality,

climate change, environmental degradation, peace,

and justice(Leal Filho et al., 2023). AI can help solve

humanity's biggest problems in almost all fields, like

human health. agriculture and forest ecosystems that

affect our planet, but large-scale AI adoption also

poses unanticipated hazards(Holzinger et al., 2021).

Thus, stakeholders, governments, policymakers,

industry, and academia must ensure that AI is

developed with these potential threats in mind. Also,

it needs to be ensured that AI applications are safe,

traceable, transparent, explainable, valid, and

verifiable. This will be possible if stakeholders

employ trustworthy and ethical AI and avoid

misusing AI technologies. AI provides many

opportunities to solve complex problems, manage

climate change, and help nations achieve their stated

targets for net zero carbon emissions in the long run.

This response aims to outline the essential principles

and insights that must be embraced to ensure a

constructive transformation of AI advancements and

implementations, ultimately facilitating the

achievement of the SDGs by the year 2030.

REFERENCES

Alloghani, M., Al-Jumeily, D., Mustafina, J., Hussain, A.,

& Aljaaf, A. J. (2020). A Systematic Review on

Supervised and Unsupervised Machine Learning

Algorithms for Data Science.

https://doi.org/10.1007/978-3-030-22475-2_1

Alpaydın, E. (2004). Introduction to Machine Learning

(2nd ed.). The MIT Press.

Awad, M., & Khanna, R. (2015). Efficient learning

machines: Theories, concepts, and applications for

engineers and system designers. In Efficient Learning

Machines: Theories, Concepts, and Applications for

Engineers and System Designers (Issue April 2015, pp.

1–248). https://doi.org/10.1007/978-1-4302-5990-9

Bhavsar, H., & Ganatra, A. (2012). A Comparative Study

of Training Algorithms for Supervised Machine

Learning. International Journal of Soft Computing and

Engineering, 2(4), 74–81.

Burges, C. J. C. (1998). A Tutorial on Support Vector

Machines for Pattern Recognition. Data Mining and

Knowledge Discovery, 2(2), 121–167.

https://doi.org/10.1023/A:1009715923555

Charu C. Aggarwal. (2018). Neural Networks and Deep

Learning. In Neural Networks and Deep Learning.

Springer. https://doi.org/110.1007/978-3-319-94463-0

Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree

Boosting System. KDD ’16: Proceedings of the 22nd

PAMIR-2 2023 - The Second Pamir Transboundary Conference for Sustainable Societies- | PAMIR

1724

ACM SIGKDD International Conference on

Knowledge Discovery and Data Mining.

David, D. (2020). Random Forest Classifier Tutorial: How

to Use Tree-Based Algorithms for Machine Learning.

FreeCodeCamp.

https://www.freecodecamp.org/news/how-to-use-the-

tree-based-algorithm-for-machine-learning/

Dutt, S., Chandramouli, S., & Das, A. K. (2019). Machine

Learning. In Pearson. Pearson India Educa?ion

Services Pvt. Ltd.

Fan, J., Lee, J., & Lee, Y. (2021). A transfer learning

architecture based on a support vector machine for

histopathology image classification. Applied Sciences

(Switzerland), 11(14).

https://doi.org/10.3390/app11146380

Gianey, H. K., & Choudhary, R. (2018). Comprehensive

Review On Supervised Machine Learning Algorithms.

Proceedings - 2017 International Conference on

Machine Learning and Data Science, MLDS 2017,

2018-Janua(December 2017), 38–43.

https://doi.org/10.1109/MLDS.2017.11

Grant, M. J., & Booth, A. (2009). A typology of reviews:

An analysis of 14 review types and associated

methodologies. Health Information and Libraries

Journal, 26(2), 91–108. https://doi.org/10.1111/j.1471-

1842.2009.00848.x

Gurney, K. (1997). Introduction to neural networks. In

Routledge (Vol. 317). UCL Press Limited.

https://doi.org/10.1016/S0140-6736(95)91746-2

H, V. M. (2023). What Is Machine Learning? Components

And Applications. VTUPulse.

https://www.vtupulse.com/machine-learning/what-is-

machine-learning-components-and-applications/

Hastie, T., Tibshirani, R., James, G., & Witten, D. (2006).

An Introduction to Statistical Learning Second Edition.

Springer Texts, 102, 618.

Holzinger, A., Weippl, E., Tjoa, A. M., & Kieseberg, P.

(2021). Digital Transformation for Sustainable

Development Goals (SDGs) - A Security, Safety and

Privacy Perspective on AI. International Cross-

Domain Conference for Machine Learning and

Knowledge Extraction, 1–20.

https://doi.org/10.1007/978-3-030-84060-0

International Telecommunication Union. (2021). United

Nations Activities on Artificial Intelligence.

Jakkula, V. (2011). Tutorial on Support Vector Machine

(SVM). School of EECS, Washington State University,

1–13.

http://www.ccs.neu.edu/course/cs5100f11/resources/ja

kkula.pdf

Khanum, M., Mahboob, T., Imtiaz, W., Abdul Ghafoor, H.,

& Sehar, R. (2015). A Survey on Unsupervised

Machine Learning Algorithms for Automation,

Classification and Maintenance. International Journal

of Computer Applications, 119(13), 34–39.

https://doi.org/10.5120/21131-4058

Leal Filho, W., Yang, P., Eustachio, J. H. P. P., Azul, A.

M., Gellers, J. C., Gielczyk, A., Dinis, M. A. P., &

Kozlova, V. (2023). Deploying digitalisation and

artificial intelligence in sustainable development

research. In Environment, Development and

Sustainability (Vol. 25, Issue 6). Springer Netherlands.

https://doi.org/10.1007/s10668-022-02252-3

Lindholm, A., Wahlström, N., Lindsten, F., & Schön, T. B.

(2019). Supervised Machine Learning: Statistical

Machine Learning course (p. 112).

http://www.it.uu.se/edu/course/homepage/sml/literatur

e/lecture_notes.pdf

Maertens, R. M., Long, A. S., & White, P. A. (2017).

Performance of the in vitro transgene mutation assay in

MutaMouse FE1 cells: Evaluation of nine misleading

(“False”) positive chemicals. Environmental and

Molecular Mutagenesis, 58(8), 582–591.

https://doi.org/10.1002/em.22125

Mahesh Batta. (2020). Machine Learning Algorithms - A

Review. International Journal of Science and Research

(IJSR), January 2019.

https://doi.org/10.21275/ART20203995

Matloff, N. (2017). Statistical Regression and

Classification. Statistical Regression and

Classification. https://doi.org/10.1201/9781315119588

Mijwel, M. M., Esen, A., & Shamil, A. (2019). Overview

of Neural Networks. In Neural Networks and Fuzzy

Systems (Issue April). https://doi.org/10.1007/978-1-

4615-6253-5_1

Morimoto, J., & Ponton, F. (2021). Virtual reality in

biology: could we become virtual naturalists?

Evolution: Education and Outreach, 14(1), 1–14.

https://doi.org/10.1186/s12052-021-00147-x

Muhammad, I., & Yan, Z. (2015). Supervised Machine

Learning Approaches: a Survey. ICTACT Journal on

Soft Computing, 05(03), 946–952.

https://doi.org/10.21917/ijsc.2015.0133

Müller, A. C., & Guido, S. (2016). Introduction to Machine

Learning with Python (D. Schanafelt (ed.); 3rd ed.).

O’Reilly Media, Inc.

https://www.oreilly.com/library/view/introduction-to-

machine/9781449369880/

Naeem, M., Rizvi, S. T. H., & Coronato, A. (2020). A

Gentle Introduction to Reinforcement Learning and its

Application in Different Fields. IEEE Access, 8,

209320–209344.

https://doi.org/10.1109/ACCESS.2020.3038605

Nasir, O., Javed, R. T., Gupta, S., Vinuesa, R., & Qadir, J.

(2023). Artificial intelligence and sustainable

development goals nexus via four vantage points.

Technology in Society, 72(November 2022), 102171.

https://doi.org/10.1016/j.techsoc.2022.102171

Natekin, A., & Knoll, A. (2013). Gradient boosting

machines, a tutorial. Frontiers in Neurorobotics,

7(DEC). https://doi.org/10.3389/fnbot.2013.00021

Ng, A. (2005). Lecture notes Supervised Learning (Issue 0).

Polzer, D. (2021).

7 of the Most Used Regression

Algorithms and How to Choose the Right One. Towards

Data Science. https://towardsdatascience.com/7-of-the-

most-commonly-used-regression-algorithms-and-how-

to-choose-the-right-one-fc3c8890f9e3

Prasad, P. (2023). The Application of Machine Learning

Techniques to the Diagnosis of Breast Cancer. 2023 3rd

International Conference on Artificial Intelligence and

Artiﬁcial Intelligence-Machine Learning Techniques Promoting SDG’s: An Exploratory Approach

1725

Signal Processing (AISP).

https://doi.org/10.1109/AISP57993.2023.10134824

Rustam, F., Reshi, A. A., Mehmood, A., Ullah, S., On, B.

W., Aslam, W., & Choi, G. S. (2020). COVID-19

Future Forecasting Using Supervised Machine

Learning Models. IEEE Access, 8, 101489–101499.

https://doi.org/10.1109/ACCESS.2020.2997311

Sætra, H. S. (2021). Ai in context and the sustainable

development goals: Factoring in the unsustainability of

the sociotechnical system. Sustainability (Switzerland),

13(4), 1–19. https://doi.org/10.3390/su13041738

Sairamya, N. J., Susmitha, L., George, S. T., & Subathra,

M. S. P. (2019). Hybrid Approach for Classification of

Electroencephalographic Signals Using Time–

Frequency Images With Wavelets and Texture Features

- ScienceDirect. In Intelligent Data Analysis for

Biomedical Applications Challenges and Solutions (pp.

253–273). Academic Press.

https://doi.org/10.1016/B978-0-12-815553-0.00013-6

Sandhu, T. H. (2018). Machine learning and natural

language processing - a review. International Journal

of Advanced Research in Computer Science, 9(2), 582–

584.

Schrider, D. R., & Kern, A. D. (2018). Supervised Machine

Learning for Population Genetics: A New Paradigm.

Trends in Genetics, 34(4), 301–312.

https://doi.org/10.1016/j.tig.2017.12.005

Sutton, R. S., & Barto, A. G. (2018). Reinforcement

Learning- An Introduction (2nd ed.). The MIT Press.

Tan, A. C., & Gilbert, D. (2003). An Empirical Comparison

of Supervised Machine Learning Techniques in

Bioinformatics BT - First Asia-Pacific Bioinformatics

Conference (APBC2003). 19(Apbc), 219–222.

http://www.cs.waikato.ac.nz/~ml/weka/%0Ahttp://crpi

t.com/confpapers/CRPITV19Tan.pdf

Uddin, S., Khan, A., Hossain, M. E., & Moni, M. A. (2019).

Comparing different supervised machine learning

algorithms for disease prediction. BMC Medical

Informatics and Decision Making, 19(1), 1–16.

https://doi.org/10.1186/s12911-019-1004-8

Vanneschi, L., & Silva, S. (2023). Support Vector

Machines. Natural Computing Series, 271–281.

https://doi.org/10.1007/978-3-031-17922-8_10

Vinuesa, R., Azizpour, H., Leite, I., Balaam, M., Dignum,

V., Domisch, S., Felländer, A., Langhans, S. D.,

Tegmark, M., & Fuso Nerini, F. (2020). The role of

artificial intelligence in achieving the Sustainable

Development Goals. Nature Communications, 11(1),

1–10. https://doi.org/10.1038/s41467-019-14108-y

Westbrook, A. L., Hannigan, D. A., Developer, C.,

Dougherty, M., College, H., Friedman, D., & Carolina,

N. (2010).

The Brain : Understanding Neurobiology

Through the Study of Addiction.

Zhang, T., Lin, W., Vogelmann, A. M., Zhang, M., Xie, S.,

Qin, Y., & Golaz, J. C. (2021). Improving Convection

Trigger Functions in Deep Convective

Parameterization Schemes Using Machine Learning.

Journal of Advances in Modeling Earth Systems, 13(5),

1–19. https://doi.org/10.1029/2020MS002365

Zhang, Z., & Jung, C. (2021). GBDT-MO: Gradient-

Boosted Decision Trees for Multiple Outputs. IEEE

Transactions on Neural Networks and Learning

Systems, 32(7), 3156–3167.

https://doi.org/10.1109/TNNLS.2020.3009776

PAMIR-2 2023 - The Second Pamir Transboundary Conference for Sustainable Societies- | PAMIR

1726