PERSONALIZED ADVISOR OF CLIMATE

CHANGE

DISSERTATION

SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE AWARD OF THE DEGREE OF
MASTER OF TECHNOLOGY
IN
DATA SCIENCE

Submitted by

RAVIRAJ (24/DSC/28)

Under the supervision of

Dr. ABHILASHA SHARMA (ASSOCIATE Prof. SE DTU)

DEPARTMENT OF DATA SCIENCE AND ENGINEERING

DELHI TECHNOLOGICAL UNIVERSITY
(Formerly Delhi College of Engineering)
Bawana Road, Delhi 110042

DECEMBER, 2024
 DEPARTMENT OF SOFTWARE ENGINEERING
DELHI TECHNOLOGICAL UNIVERSITY
(Formerly Delhi College of Engineering)
Bawana Road, Delhi-110042

CANDIDATE’S DECLARATION

I, RAVIRAJ, Roll No – 24/DSC/28 students of M.Tech (DEPARTMENT OF SOFT-

WARE ENGINEERING ),hereby declare that the project Dissertation titled “PERSON-
ALIZED ADVISOR OF CLIMATE CHANGE ” which is submitted by us to the redDe-
partment OF SOFTWARE ENGINEERING , Delhi Technological University, Delhi in
partial fulfilment of the requirement for the award of degree of Master of Technology, is
original and not copied from any source without proper citation. This work has not pre-
viously formed the basis for the award of any Degree, Diploma Associateship, Fellowship
or other similar title or recognition.

Place: Delhi RAVIRAJ

Date:

i
 DEPARTMENT OF SOFTWARE ENGINEERING
DELHI TECHNOLOGICAL UNIVERSITY
(Formerly Delhi College of Engineering)
Bawana Road, Delhi-110042

CERTIFICATE

I hereby certify that the Project Dissertation titled “PERSONALIZED ADVISOR

OF CLIMATE CHANGE ” which is submitted by RAVIRAJ, Roll No – 24/DSC/28, DE-
PARTMENT OF SOFTWARE ENGINEERING ,Delhi Technological University, Delhi
in partial fulfilment of the requirement for the award of the degree of MASTER of Tech-
nology, is a record of the project work carried out by the students under my supervision.
To the best of my knowledge this work has not been submitted in part or full for any
Degree or Diploma to this University or elsewhere.

Place: Delhi Dr. ABHILASHA SHARMA

(ASSOCIATE Prof. SE DTU)

Date: SUPERVISOR

ii
 DEPARTMENT OF SOFTWARE ENGINEERING
DELHI TECHNOLOGICAL UNIVERSITY
(Formerly Delhi College of Engineering)
Bawana Road, Delhi-110042

ACKNOWLEDGEMENT

We wish to express our sincerest gratitude to Dr. ABHILASHA SHARMA for her
continuous guidance and mentorship that she provided us during the project. She showed
us the path to achieve our targets by explaining all the tasks to be done and explained
to us the importance of this project as well as its industrial relevance. She was always
ready to help us and clear our doubts regarding any hurdles in this project. Without her
constant support and motivation, this project would not have been successful.

Place: Delhi RAVIRAJ

Date:

iii
 ABSTRACT

Abstracts contain most of the following kinds of information in brief form. The body
of your paper will, of course, develop and explain these ideas much more fully. As you
will see in the samples below, the proportion of your abstract that you devote to each
kind of information—and the sequence of that information—will vary, depending on the
nature and genre of the paper that you are summarizing in your abstract. And in some
cases, some of this information is implied, rather than stated explicitly. The Publication
Manual of the American Psychological Association, which is widely used in the social
sciences, gives specific guidelines for what to include in the abstract for different kinds
of papers—for empirical studies, literature reviews or meta-analyses, theoretical papers,
methodological papers, and case studies.
Here are the typical kinds of information found in most abstracts:
The context or background information for your research; the general topic under
study; the specific topic of your research the central questions or statement of the prob-
lem your research addresses what’s already known about this question, what previous
research has done or shown the main reason(s), the exigency, the rationale, the goals for
your research—Why is it important to address these questions? Are you, for example,
examining a new topic? Why is that topic worth examining? Are you filling a gap in
previous research? Applying new methods to take a fresh look at existing ideas or data?
Resolving a dispute within the literature in your field? . . . your research and/or ana-
lytical methods your main findings, results, or arguments the significance or implications
of your findings or arguments. Your abstract should be intelligible on its own, without
a reader’s having to read your entire paper. And in an abstract, you usually do not cite
references—most of your abstract will describe what you have studied in your research
and what you have found and what you argue in your paper. In the body of your paper,
you will cite the specific literature that informs your research.

iv
 Contents

CANDIDATE’S DECLARATION i

CERTIFICATE ii

ACKNOWLEDGEMENT iii

ABSTRACT iv

CONTENT vi

LIST OF TABLES vii

LIST OF FIGURES viii

LIST OF ABBREVIATIONS ix

ABBREVIATIONS ix

1 INTRODUCTION x
1.1 BACKGROUND . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
1.2 MOTIVATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
1.3 OBJECTIVE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
1.4 OVERVIEW OF TECHNOLOGIES . . . . . . . . . . . . . . . . . . . . . . xii
1.4.1 ”Ultra-fast Machine Learning Classifier Execution on IoT Devices
Without SRAM Consumption (Sudharsan et al. 2021)” . . . . . . . xii
1.4.2 ”Improving graph collaborative filtering with view explorer for so-
cial recommendation” . . . . . . . . . . . . . . . . . . . . . . . . . . xii
1.4.3 ”Real-time tracking of structural evolution in 2D MXenes using
theory-enhanced machine learning” . . . . . . . . . . . . . . . . . . xiii

v
 1.4.4 ”A review of applied research on low-carbon urban design: based
on scientific knowledge mapping” . . . . . . . . . . . . . . . . . . . xiii
1.4.5 ”Pruning Random Forests for Prediction on a Budget ” . . . . . . . xiii
1.4.6 ”Material Science with Machine Learning (Hollenbach et al. 2024):” xiv
1.4.7 ”Climate-Smart Forestry (Wang et al. 2025):” . . . . . . . . . . . . xiv

2 RELATED WORK 2

3 LITERATURE REVIEW 4
3.1 Personalized Energy-Saving Recommendations [1] . . . . . . . . . . . . . . 4
3.2 Climate-Smart Forestry [2] . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.3 AI and Climate Change Prediction [3] . . . . . . . . . . . . . . . . . . . . . 5
3.4 Low-Carbon Urban Design [4] . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.5 Real-time Tracking of Structural Evolution in 2D MXenes Using Theory-
Enhanced Machine Learning [5] . . . . . . . . . . . . . . . . . . . . . . . . 6
3.6 Climate-Smart Forestry: An AI-Enabled Sustainable Forest Management
Solution for Climate Change [6] . . . . . . . . . . . . . . . . . . . . . . . . 7
3.7 Ultra-fast Machine Learning Classifier Execution on IoT Devices Without
SRAM Consumption [7] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

4 RESULTS AND DISCUSSION 9

4.1 RESULT: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.2 DISCUSSION: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

5 CONCLUSION AND FUTURE SCOPE 11

5.1 CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.2 FUTURE SCOPE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

vi
 List of Tables

3.1 Literature Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

vii
 List of Figures

1.1 Analysis of Recommendation System . . . . . . . . . . . . . . . . . . . . . xii

1.2 Problem Statement Using Diagram . . . . . . . . . . . . . . . . . . . . . . 1

viii
 List of Abbreviations

AI Artificial Intelligence

ML Machine Learning

IOT Intetnet Of Things

SV M Support Vector Machine

DT Decision Tree

RL Reinforcement Learning

LCF LifeStyle Carbon FootPrint

CN N Convolutional Neural Network

CF Carbon FootPrint

ix
Chapter 1

INTRODUCTION

1.1 BACKGROUND
Several facets of climate change combining makes it multi-great crisis for the globe with
particular threats to ecosystems, economies, and public health. This effect is mainly
caused by increasing greenhouse gas emissions: it shakes the natural systems, further
worsens extreme weather conditions, and threatens biological diversity. Perhaps it is the
most demanding challenge for the current century. The traditional approach for dealing
with climate change has been through physical models and historical data that often fall
short of establishing the deep, dynamic interaction characteristic of the climate systems.
However, advancements in artificial intelligence (AI) now create transformational oppor-
tunities for enhanced understanding and response to climate change. AI is revolutionary
because it collects vast and diverse data into providing innovative tools for predicting en-
vironmental changes, anticipating effects, and shaping specialized strategies for mitigating
and adapting. The add-ons range from forecasting extreme weather events and improv-
ing agricultural productivity to empowering individuals and businesses with personalized
recommendations on carbon footprint reductions. Despite its usefulness, the integration
of AI in climate action needs to attend to ethical challenges such as data privacy, energy
use, and algorithmic biases. With interdisciplinary collaboration and responsible use,
however, AI can become one of the cornerstones of the global fight against climate change
and sustainability.

x
1.2 MOTIVATION
It’s motivating: The pressing requirement is to seek out new and scalable approaches
to climate change rather than resorting to old stereotypical ways. Large policies and
technologies would be part of the solution, but supporting the citizen in understanding
how he or she can make sensitive, environmentally related choices needs to be equally, if
not more, a part of the solution. Unfortunately, current methods of motivating personal
action on climate issues generally lack the kind of granularity and relevance that would
move people to sustained behavioral change. And here, it is where artificial intelligence
comes in as a game changer. AI could process huge, intricate piles of data and gener-
ate personalized recommendations, making it possible to bridge individual behavior and
global climate goals. By equipping people with actionable, context-aware insights, AI
doesn’t just engage them much deeper into sustainability practices but also multiplies
the collective impacts of personal actions in tackling climate change. Personalized, AI-
enabled climate action is that vision driving the exploration of its innovations for a more
sustainable future.

1.3 OBJECTIVE
AI can revolutionize climate change prediction and action by its precise predictions of
impacts and enabling meaningful and personalized actions. AI tools, especially machine
learning algorithms, can help improve the accuracy of climate forecasting, warn against
forthcoming natural disasters, and develop mitigation-adaptation strategies. In addition,
it can demonstrate that AI can empower people to take actions that cut carbon footprints
through recommendations unique to behavioral and lifestyle characteristics, based on
data. Addressing data privacy, algorithmic bias, and energy-intensiveness models, this
paper will address the importance of interdisciplinary collaborations and ethical use of
tools on the use of AI technology. The end objective is to show the function of AI as a
pillar of global sustainability efforts to building resilience and transform toward a carbon-
neutral world.

xi
 Figure 1.1: Analysis of Recommendation System

1.4 OVERVIEW OF TECHNOLOGIES

1.4.1 ”Ultra-fast Machine Learning Classifier Execution on IoT

Devices Without SRAM Consumption (Sudharsan et al.
2021)”

Models: Decision Trees (DT) and Random Forests (RF) Methodology: SRAM-optimized
models for deployment on MCUs; involves training models and then porting them to C
code for real-time deployment without consuming SRAM. Accuracy:The models main-
tain similar accuracy to their original versions, but specific numbers are not provided.
Performance is compared in terms of speedup (1-4x faster inference) Datasets Used:
The research focused on common machine learning datasets for classification tasks, but
specific datasets used in the paper were not explicitly listed. However, datasets typically
used for Decision Trees (DT) and Random Forests (RF) include: 1. Iris Dataset 2. Breast
Cancer Dataset

1.4.2 ”Improving graph collaborative filtering with view explorer

for social recommendation”

Name: Social View Explorer Collaborative Filtering (SVE-CF) Models: GNN-based

model for social recommendations • Methodology: o Correlates social and interaction

xii
behaviors of users. Creates follow, joint, and interaction views.
Uses tri-view joint learning for filtering noise and enhancing signal quality. Accuracy:
Outperforms seven state-of-the-art models in social recommendation tasks (specific met-
rics not directly provided in the snippet). Dataset: Three benchmark datasets in social
recommendations

1.4.3 ”Real-time tracking of structural evolution in 2D MXenes

using theory-enhanced machine learning”

Name: Theory-Enhanced Machine Learning Framework Models: Variational Autoen-

coders (VAE) combined with unsupervised feature extraction. Methodology: o Embeds
computationally simulated and experimental data into a unified latent space. o Uses un-
supervised learning for classification and feature extraction. Designed for real-time mon-
itoring and feedback in Transmission Electron Microscopy (TEM). Accuracy: Achieved
high classification accuracy and denoising effectiveness (quantitative accuracy metrics not
specified). Dataset: In situ EELS-SI dataset consisting of 570 spectra, supplemented with
computationally simulated datasets2.

1.4.4 ”A review of applied research on low-carbon urban design:

based on scientific knowledge mapping”

Authors: Gaixia Wang, Yunshan Wan, Chante Jian Ding, Xiaoqian Liu, Yuxin Jiang
• Year: 2023 Models: Framework for low-carbon urban design analysis (no specific
algorithmic model). • Methodology: o Employs tools like Citespace for mapping research
evolution in low-carbon cities. Focuses on themes like carbon sink accounting, spatial
systems, and multi-scale decarbonization. Accuracy: Not applicable as this is a review
and framework analysis. Dataset: Literature from the Web of Science core database; 319
documents analyzed.

1.4.5 ”Pruning Random Forests for Prediction on a Budget ”

Authors: F. Nan, J. Wang, V. Saligrama Models: Random Forests (RF) Method-

ology: Pruning techniques to reduce the complexity and memory requirements of RF
models for use in resource-constrained environments, maintaining prediction accuracy.

xiii
Accuracy: Maintains prediction accuracy even with pruning, but specific numbers are
not provided.
Datasets Used: This paper focuses on pruning Random Forests (RF) models to reduce
their size and computational cost. The datasets often used in Random Forest studies
include:
Iris Dataset
Adult Income Dataset (Census Income) – Used for predicting income based on demo-
graphic data.

1.4.6 ”Material Science with Machine Learning (Hollenbach et

al. 2024):”

Models: Variational Autoencoders (VAE) Methodology: Uses VAEs to model and pre-
dict structural changes in 2D MXenes in real time by integrating theoretical and exper-
imental data. Accuracy: The VAE model showed good prediction accuracy in tracking
structural changes, but specific metrics are not given. Dataset Used: This research
is focused on materials science, specifically 2D MXenes, and uses both theoretical and
experimental datasets. Common datasets in material science ML applications include:
Materials Project Database – Contains data on material properties and behaviors, typi-
cally used for prediction tasks.
Inorganic Crystal Structure Database (ICSD) – Used for predicting material properties
based on structure.
These datasets are used to train machine learning models that track material properties
and predict their behaviors in real-time.

1.4.7 ”Climate-Smart Forestry (Wang et al. 2025):”

Models: AI models for monitoring and managing forests Methodology: Focuses on

using AI, digital twins, and remote sensing for sustainable forest management, monitor-
ing carbon sequestration, and adapting to climate change Accuracy: The models are
designed for efficiency rather than classification accuracy, with performance evaluated
based on forest management outcomes. Dataset Used: This paper discusses the use

xiv
of AI in sustainable forestry management. Typical datasets for this application include:
Forest Inventory Data – Contains detailed information on tree species, height, and growth
metrics.
Climate Data – Includes temperature, precipitation, and soil data, useful for modeling
forest growth and carbon sequestration.

Figure 1.2: Problem Statement Using Diagram

1
Chapter 2

RELATED WORK

The related work section of the document highlights significant contributions to the in-
tersection of artificial intelligence (AI) and climate action. Various studies have explored
how AI can be used for personalized climate action, energy-saving recommendations, and
climate change prediction. For instance, Sardianos et al. (2020) developed the (EM)³
recommendation engine, which uses IoT sensors to provide real-time, personalized advice
for energy-saving in households and offices. However, this system is limited to energy
consumption and does not address broader aspects of individual carbon footprints, such
as transportation or dietary choices. Similarly, Wang et al. (2025) introduced the concept
of Climate-Smart Forestry (CSF), integrating AI and robotics to improve forest resilience
and carbon sequestration. Despite its innovative approach, the adoption of CSF technolo-
gies faces challenges, including the high costs of implementing advanced tools like LiDAR
and digital twins, as well as the long timeframes needed to validate the effectiveness of
AI interventions in forestry.
Further advancements in AI have been explored in climate change prediction. Satish et
al. (2023) focused on using AI algorithms, such as neural networks, to predict climate im-
pacts, including extreme weather events and sea-level rise. While AI models have proven
useful in improving disaster preparedness, they come with the drawback of significant
computational demands, which contribute to their environmental footprint. Additionally,
Wang et al. (2023) reviewed AI applications in low-carbon urban design, emphasizing the
role of AI in optimizing urban green spaces, integrating renewable energy, and enhancing
energy-efficient buildings. However, their study primarily focused on green spaces and
did not fully address other critical aspects of urban emissions, such as transportation and
industrial activities.

2
 Despite these promising developments, several limitations exist across these studies.
Many of the systems reviewed focus narrowly on specific aspects of climate action, such as
energy consumption or forestry management, leaving room for more holistic approaches
that incorporate diverse data sources, including transportation, dietary choices, and waste
management. Additionally, ethical concerns surrounding data privacy, biases in AI mod-
els, and the environmental impact of AI itself must be addressed for these technologies
to be implemented on a wider scale. There is also a need for cost-effective solutions to
make AI-driven climate action more accessible, particularly in resource-constrained en-
vironments, and for standardized metrics to ensure the reliability and comparability of
AI-driven models. These challenges underscore the importance of expanding the scope
of AI applications to provide more comprehensive and scalable solutions for personalized
climate action.

3
Chapter 3

LITERATURE REVIEW

The intersection of artificial intelligence (AI) and climate action has witnessed significant
advancements in recent years, particularly in the domains of personalized recommenda-
tions, climate impact prediction, and sustainable development. This section summarizes
key contributions from existing literature.

3.1 Personalized Energy-Saving Recommendations [1]

Existing Work: Sardianos et al. (2020) presented the (EM)³ recommendation engine,
designed to provide realtime, personalized energy-saving advice to household and office
users. By leveraging IoT sensors to monitor energy consumption and environmental
conditions, the system generates actionable recommendations tailored to user habits and
preferences. The integration of persuasive facts, such as projected savings, enhances user
compliance and engagement1.
Limitations:
The system’s reliance on IoT sensors and smart home devices limits its applicability in
regions or households without access to such technology.
The recommendation engine focuses primarily on energy consumption behaviors and does
not address broader aspects of individual carbon footprints, such as transportation or
dietary choices.

4
3.2 Climate-Smart Forestry [2]
Existing Work: Wang et al. (2025) introduced the concept of Climate-Smart Forestry
(CSF), which integrates AI, robotics, and digital twin technologies to improve forest re-
silience and carbon sequestration. The study emphasizes using AI for real-time monitoring
and predictive modeling to adapt forest management practices to climate change2.
Limitations:
The implementation of CSF is challenged by the high costs of adopting advanced tech-
nologies like LiDAR and digital twins, especially in resource- limited settings.
The long life cycles of trees make it difficult to validate the effectiveness of AI-driven
interventions in a short timeframe.

3.3 AI and Climate Change Prediction [3]

Existing Work: Satish et al. (2023) reviewed the role of AI in predicting climate change
impacts, such as extreme weather events and sea-level rise. AI algorithms like neural net-
works were used to analyze complex datasets, enabling precise predictions and improving
disaster preparedness. Additionally, AI-assisted modeling was applied to simulate climate
scenarios, guiding policymakers in crafting effective mitigation strategies3.
Limitations:
The computational demands of AI models contribute to a significant environmental foot-
print, which can counteract their intended benefits for climate action.
Ethical concerns around data privacy and bias in AI models remain unresolved, hindering
trust and widespread adoption.

3.4 Low-Carbon Urban Design [4]

Existing Work: Wang et al. (2023) conducted a comprehensive review of low-carbon
urban design, highlighting the role of AI in optimizing urban green spaces, integrating
renewable energy, and enhancing energy-efficient buildings. The study emphasizes the
need for a multi-scale approach to urban planning, leveraging AI to link carbon sink ac-
counting, spatial modeling, and design applications4.
Limitations:

5
The study primarily focuses on urban green spaces and does not address other critical
aspects of urban emissions, such as transportation and industrial activities. The lack of
standardized methods for carbon sink accounting in urban areas reduces the reliability of
AI-driven assessments.
Opportunities for Improvement
The limitations outlined in these works highlight key areas for future research: 1. Broader
Integration: Expanding the scope of personalized recommendations to include transporta-
tion, dietary choices, and waste management can provide a holistic approach to reducing
carbon footprints. 2. Ethical AI Development: Addressing concerns around privacy, data
security, and biases in AI models is essential for building trust and ensuring equitable de-
ployment. 3. Resource Accessibility: Developing cost-effective AI solutions can increase
accessibility and scalability, especially in resource-constrained environments. 4. Standard-
ized Metrics: Establishing global standards for carbon accounting in urban and forestry
contexts can enhance the reliability and comparability of AI-driven models. This synthe-
sis bridges individual, urban, and ecosystem-level solutions, emphasizing the potential of
AI to drive meaningful climate action while addressing current gaps and challenges.

3.5 Real-time Tracking of Structural Evolution in 2D

MXenes Using Theory- Enhanced Machine Learn-
ing [5]
Existing Work: This research develops a novel machine learning framework that inte-
grates theoretical and experimental data to track the atomic-scale structural evolution of
2D MXenes in real-time.
Variational autoencoders (VAEs) are used to embed multi-source data into a latent space
for accurate classification and dynamic structure prediction.
The framework supports accelerated material discovery and real-time insights into mate-
rial performance under varying conditions.
Limitations:
Dependence on large, labeled datasets for training limits scalability in new applications.
The computational expense of generating high-quality training data hinders real-time ap-

6
plicability in resource-constrained environments. Results may vary when applying the
framework to materials outside the initial study focus.

3.6 Climate-Smart Forestry: An AI-Enabled Sustain-

able Forest Management Solution for Climate Change
[6]
Existing Work:
This paper proposes integrating AI with sustainable forestry practices to enhance
resilience and carbon sequestration. Techniques such as digital twins, robotics, and real-
time data acquisition are applied to monitor and adapt forest management strategies. AI
models provide predictions for growth, climate resilience, and optimal harvest times to
align with climate adaptation and mitigation goals.
Limitations:
High costs of implementing advanced technologies like digital twins and robotics make
widespread adoption challenging. The time required to observe the long-term effects of AI-
guided strategies limits rapid feedback and improvements. Limited generalization of the
models to diverse forest types or regions with different climatic and ecological conditions.

3.7 Ultra-fast Machine Learning Classifier Execution

on IoT Devices Without SRAM Consumption [7]
Existing Work:
Introduces an SRAM-optimized approach for executing machine learning classifiers on
resource-constrained IoT devices, such as microcontroller units (MCUs). Focuses on de-
ploying models without altering their original structure, ensuring minimal computational
overhead while preserving accuracy. Demonstrates a 1-4x improvement in inference speed
compared to libraries like sklearnporter, m2cgen, and emlearn.
Limitations:
Scalability is limited as flash memory usage grows linearly with model complexity. The
method is focused solely on inference; real-time training and model updates are not sup-

7
ported. Application is limited primarily to tree-based classifiers (e.g., Decision Trees,
Random Forests) and does not extend to neural networks or other advanced models.
S.NO. REFERENCES PUBLICATION YEAR DATASET METHOD RESULT
1 [1] IEEE 2020 Iris Dataset Combines de- Achieves better
cision trees accuracy than
with ensemble traditional de-
methods to cision trees and
create compact, random forests
efficient models with a compact
that perform model. Specific
well on classifi- values not pro-
cation tasks. vided.
2 [2] ScienceDirect 2020 Forest Inventory Focuses on using The models are
Data AI, digital twins, designed for
and remote sens- efficiency rather
ing for sustain- than classifica-
able forest man- tion accuracy,
agement, moni- with perfor-
toring carbon se- mance evaluated
questration, and based on forest
adapting to cli- management
mate change outcomes.
3 [3] IEEE 2023 Three bench- Correlates social Outperforms
mark datasets and interaction seven state-of-
in social recom- behaviors of the-art models
mendations users. in social recom-
mendation tasks
(specific metrics
not directly
provided in the
snippet).

4 [4] IEEE 2012 Literature from String matching, Similarity

the Web of text comparison scores, match
Science core percentages
database
5 [5] Springer 2023 Literature from Employs tools Not applica-
the Web of like Citespace ble as this is
Science core for mapping re- a review and
database search evolution framework anal-
in low-carbon ysis.
cities
6 [6] ScientificReports 2024 In situ EELS-SI Embeds compu- Achieved high
dataset con- tationally simu- classification
sisting of 570 lated and experi- accuracy and
spectra, sup- mental data into denoising ef-
plemented with a unified latent fectiveness
computation- space (quantitative
ally simulated accuracy metrics
datasets2. not specified).
7 [7] IJERT 2013 14,807 Java Statistical anal- Cross entropy
projects, 352M ysis, N-gram (4.9 bits avg.),
LOC models, smooth- identifier
ing uniqueness,
predictability

Table 3.1: Literature Survey

8
Chapter 4

RESULTS AND DISCUSSION

4.1 RESULT:
The synthesis of all this diverse research brings out how massively advances have been
made in the use of artificial intelligence (AI) and machine learning (ML) across the multi-
ple fields of resource-constrained environments, social recommendation systems, materials
science, low-carbon urban design, and climate-smart forestry. For example, in its publi-
cation, herein Sudharsan et al. suggest SRAM-optimized models specifically designed for
IoT devices, which strongly enhance inference speed and memory efficiency during opera-
tion for Decision Trees and Random Forests while maintaining accuracy, thus making such
models possibly best suited to real-time deployment. In like manner, the same SVE-CF
model has designed and tested social recommendations through GNNs and tri-view joint
learning, and outperformed the best approach against relevant benchmark datasets.
In material science, Variational Autoencoders (VAEs) can be easily and efficiently
applied in tracking and forecasting structural changes in 2D MXenes in real time, synthe-
sizing theoretical and experimental data for unsupervised feature extraction and classifica-
tion. Thus, the study of low-carbon urban design being conducted by Wang et al. is quite
comprehensive; it utilizes from research mapping instruments like Citespace to underscore
the importance of spatial systems and carbon sink accounting in urban decarbonization.
With the aid of AI, digital twins, and remote sensing technologies, Climate-Smart Forestry
optimizes forest management, carbon sequestration, and resilience to climate change. Fi-
nally, pruning techniques for Random Forests, as discussed by Nan et al., enable their
efficient deployment in constrained environments while retaining predictive performance.
These works together underline the travelling transforming road of AI and ML into

9
the belly of truly complex interdisciplinary problems.

4.2 DISCUSSION:
These studies have shown the various applications of artificial intelligence (AI) and ma-
chine learning (ML) in diversifying areas ranging from climate action, resource-efficient
technologies to material science and forestry. IoT devices such as SRAM-optimized execu-
tion for Decision Trees (DT), Random Forests (RF) showcase efficient models of deploying
ML in resource-constrained environments with high speed and accuracy improvements
achieved. Further improvements are made in the pruning methods used in RF models to
solve computational problems that limit the deployment in low-resource settings without
a trade-off in performance predictions.
It is in a social recommendation system that the Social View Explorer-Collaborative
Filtering (SVE-CF) presents the graph-based model that supports tri-view learning, has
more user profile comparisons for noise reduction in recommendations and improvements
in recommendations for accuracy. This includes advancing in using variational autoen-
coders in both the fields of material science and urban design. For instance, in how
theory-enhanced machine learning combines theory and experimental data to perform
real-time monitoring of structure changes for 2D MXenes, variational autoencoders can
track changes in the material. Citationspace is another example of tools for urban area
supporting research synthesis in mapping themes, such as carbon sink accounting and
multi-scale strategies for decarbonization.
Forests and other sustainable managements are now utilizing AI for innovations in-
volving Climate-Smart Forestry; this involves digital twins and remote sensing to optimize
carbon sequestration and adapt to climate change. That is to say, several of these appli-
cations want to contribute to increasing resilience at the same time as predicting what
will happen with AI.

10
Chapter 5

CONCLUSION AND FUTURE SCOPE

5.1 CONCLUSION
Artificial integrating AI into personalized climate actions with resource-constrained sys-
tems as IoT devices would certainly turn a significant revolutionary dawn toward sus-
tainable development. Access to diversified data sources in conjunction with AI-enabled
insights will ensure individuals’ informed, eco-friendly decisions against large-scale climate
strategies into the individual behaviour; hence, it makes a vast impact on the global carbon
footprints. In line with the current, highly promising trend also in developing deployable
ML models on low-resource IoT devices: SRAM-optimized frameworks, real-time, energy-
efficient applications ranging from smart home systems to climate-smart forestry can now
possibly be achieved. However, even with all of these advancements, some challenges
persist, including data privacy, scalability, energy consumption, and applicability of AI
for all other types of models. Research in the future must, therefore, address these issues
by presenting highly adaptive, scalable, and resource-efficient AI solutions while seeking
their interdisciplinary application in areas such as healthcare, agriculture, and disaster
management. This makes a very efficient step toward addressing what could become crit-
ical gaps for AI in enabling a much fairer and more resilient and resource-efficient future,
thus linking personal contributions to global sustainability goals.

5.2 FUTURE SCOPE

Among them are very interesting avenues for future work for integrating artificial intel-
ligence (AI) into personalized climate action and lived environments with resource con-

11
straints. One is by adding more personalization to climate action regarding behavioral
data, such as eating habits, waste management, and social activitythus improving the ac-
curacy of carbon footprint estimations. Furthermore, adaptive recommendation through
reinforcement learning could result in higher student engagement and compliance. In
this sense, scalability of AI models for IoT devices also provides ground for research in
algorithms that balance growing complexity with energy and memory efficiency while ex-
tending SRAM-optimized methods to also cover deep learning models like convolutional
neural networks (CNNs) and recurrent neural networks (RNNs). Within an ecosystem,
real-time feedback systems through edge AI can help to develop dynamic recommendation
improvements from environmental and user behavioral changes in real-time. AI will still
be sustainable, aiming at lowering the carbon footprints of AI systems, energy-efficient
training and inference mechanisms, as well as green and low-cost hardware for edge ge-
ographies. Interdisciplinary applications-e.g., extending climate-smart forestry and IoT
innovations to healthcare, agriculture, and disaster management-could prove very lucra-
tive in terms of more inclusive social impact. Finally, using AI with blockchain would also
ensure transparency and secure data sharing within climate and IoT ecosystems. Over-
all, these require scalable, adaptive, and sustainable AI solutions that connect individual
action to global sustainability objectives.

12
 REFERENCES

[1] C. Sardianos, C. Chronis, I. Varlamis, G. Dimitrakopoulos, Y. Himeur, A. Alsalemi,

F. Bensaali, and A. Amira, “Real-time personalised energy saving recommendations,”
in 2020 International Conferences on Internet of Things (iThings) and IEEE Green
Computing and Communications (GreenCom) and IEEE Cyber, Physical and Social
Computing (CPSCom) and IEEE Smart Data (SmartData) and IEEE Congress on
Cybermatics (Cybermatics), 2020, pp. 366–371.

[2] P. Verkerk, R. Costanza, L. Hetemäki, I. Kubiszewski, P. Leskinen, G. Nabuurs,

J. Potočnik, and M. Palahı́, “Climate-smart forestry: the missing link,”
Forest Policy and Economics, vol. 115, p. 102164, 2020. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S1389934120300630

[3] M. Satish, Prakash, S. M. Babu, P. P. Kumar, S. Devi, and K. P. Reddy, “Artificial

intelligence (ai) and the prediction of climate change impacts,” in 2023 IEEE 5th In-
ternational Conference on Cybernetics, Cognition and Machine Learning Applications
(ICCCMLA), 2023, pp. 660–664.

[4] E. H. W. Chan, S. Conejos, and M. Wang, Low Carbon Urban Design: Potentials and
Opportunities. Cham: Springer International Publishing, 2017, pp. 75–88. [Online].
Available: https://doi.org/10.1007/978-3-319-49730-38

[5] P. C. J. H. e. a. Hollenbach, J.D., “Real-time tracking of structural evolution in 2d

mxenes using theory-enhanced machine learning.” in https://doi.org/10.1038/s41598-
024-66902-4, 2024.

[6] G. G. Wang, D. Lu, T. Gao, J. Zhang, Y. Sun, D. Teng, F. Yu, and J. Zhu,
“Climate-smart forestry: an AI-enabled sustainable forest management solution for
climate change adaptation and mitigation,” Journal of Forestry Research, vol. 36, no. 1,
p. 7, Nov. 2024.

13
[7] B. Sudharsan, P. Patel, J. G. Breslin, and M. I. Ali, “Ultra-fast machine learning classifier
execution on iot devices without sram consumption,” in 2021 IEEE International
Conference on Pervasive Computing and Communications Workshops and other Affiliated
Events (PerCom Workshops), 2021, pp. 316–319.

14