ACCEPTED MANUSCRIPT

Mastering supply chain’s decision-making establishing SDG’s goal: a social media analytics
study of the electronic devices in developing and developed countries
This Accepted Manuscript (AM) is a PDF file of the manuscript accepted for publication after peer review, when applicable, but
does not reflect post-acceptance improvements, or any corrections. Use of this AM is subject to the publisher's embargo period
and AM terms of use. Under no circumstances may this AM be shared or distributed under a Creative Commons or other form of
open access license, nor may it be reformatted or enhanced, whether by the Author or third parties. By using this AM (for
example, by accessing or downloading) you agree to abide by Springer Nature's terms of use for AM versions of subscription
articles: https://www.springernature.com/gp/open-research/policies/accepted-manuscript-terms

The Version of Record (VOR) of this article, as published and maintained by the publisher, is available online at:
https://doi.org/10.1007/s10479-024-06078-2. The VOR is the version of the article after copy-editing and typesetting, and
connected to open research data, open protocols, and open code where available. Any supplementary information can be found on
the journal website, connected to the VOR.

t
For research integrity purposes it is best practice to cite the published Version of Record (VOR), where available (for example,

ip
see ICMJE’s guidelines on overlapping publications). Where users do not have access to the VOR, any citation must clearly
indicate that the reference is to an Accepted Manuscript (AM) version.

cr
us
an
m
d
te
ep
cc
A
 ACCEPTED MANUSCRIPT

Mastering supply chain’s decision-making establishing SDG’s goal: a

social media analytics study of the electronic devices in developing and
developed countries

Authors and affiliations

Sajjad Shokouhyar, Australian Institute of Business, Department of Supply Chain and Operations
Management, 1 King William Street, Adelaide, SA 5000, Australia, s_shokouhyar@sbu.ac.ir

[Corresponding author] Mohammad Hossein Shahidzadeh, Management and Accounting Faculty,

t
Department of Industrial and Information Management, Shahid Beheshti University, Tehran, Iran,

ip
m_shahidzadeh@sbu.ac.ir, shahidzadeh@gmail.com
Abstract
This research proposed a multi-industry framework that aims to clarify reverse logistics decision-making

cr
through a social media analytics approach analyzing the electronics industry in developing compared to
developed countries in separated case studies. By leveraging deep learning on social media data, this
study intends to elucidate ideal product return policies and strategies aligned with sustainability, circular

us
economic principles, and sustainable development goals. Furthermore, this work outlines the potential of
social media analytics to align consumer expectations with supply chain decisions across geographic
contexts and identify the missing potential stream of material in supply chains. Overall, this study utilizes
social media data mining and sentiment analysis to shed light on optimal reverse logistics decision-
an
making for electronics firms in diverse national settings pertinent to circular economy ideals. The
juxtaposition of developing and developed nations provides an enhanced understanding of how optimal
return policies may differ given geographic and economic contexts. In summary, this research elucidates
data-driven reverse logistics insights via social media analytics to inform electronics sector circular
m
economy strategies tailored to national development levels.
Keywords: SDGs; Reverse logistics; Industry 5; Social media analytics; Deep learning
d
te

Mastering supply chain’s decision-making establishing SDG’s goal: a social media

analytics study of the electronic devices in developing and developed countries

ep

Abstract
cc

This research proposed a multi-industry framework that aims to clarify reverse logistics decision-making

through a social media analytics approach analyzing the electronics industry in developing compared to
A

developed countries in separated case studies. By leveraging deep learning on social media data, this study

intends to elucidate ideal product return policies and strategies aligned with sustainability, circular

economic principles, and sustainable development goals. Furthermore, this work outlines the potential of

social media analytics to align consumer expectations with supply chain decisions across geographic

1
 ACCEPTED MANUSCRIPT

contexts and identify the missing potential stream of material in supply chains. Overall, this study utilizes

social media data mining and sentiment analysis to shed light on optimal reverse logistics decision-making

for electronics firms in diverse national settings pertinent to circular economy ideals. The juxtaposition of

developing and developed nations provides an enhanced understanding of how optimal return policies may

differ given geographic and economic contexts. In summary, this research elucidates data-driven reverse

t
logistics insights via social media analytics to inform electronics sector circular economy strategies tailored

ip
to national development levels.

cr
Keywords: SDGs; Reverse logistics; Industry 5; Social media analytics; Deep learning

us
1. Introduction

Reverse logistics, managing the return of products from consumers to manufacturers, has become an
an
increasingly crucial issue in supply chain management and sustainability (Pourmehdi et al., 2022). The

disposition decisions around defective, damaged, or end-of-life products significantly affect costs,
m
efficiency, and the environment (Wilson and Goffnett, 2022). The electronics industry, in particular, faces

massive volumes of product returns and complex reverse logistics challenges due to short lifecycles and
d

frequent new product introductions (Ahmadi et al., 2024). At the same time, circular economy principles
te

have elevated the importance of reclaiming value from used electronics to reduce waste (Mishra et al.,
ep

2023). Effective supply chain decision-making is critical for firms seeking to balance profitability,

sustainability, and alignment with societal goals (Khan et al., 2021). Simultaneously, principles of
cc

sustainable development have become integral for responsible supply chain management, as evidenced by

the United Nations Sustainable Development Goals (SDGs), which call for ecological sustainability and
A

human well-being (Tseng et al., 2020). Electronics firms face particular decision-making complexity in

aligning smart supply chains with SDGs amidst pressures of short lifecycles, frequent new product releases,

and massive product returns (Nayal et al., 2022). While reverse logistics has garnered growing attention,

optimal decision-making remains unclear, especially given geographic and economic differences between

developed and developing nations (Chen et al., 2021). The objective of this research is to present a model

2
 ACCEPTED MANUSCRIPT

that investigates the reverse logistics policies of mass-production manufacturing industries by examining

consumer perspectives across various national contexts using social media analytics. The aim is to provide

insights and understanding into these policies. Specifically, we utilize data mining of social media posts

related to product returns to elucidate patterns, sentiments, and expectations around mass-production

manufacturing industries' reverse logistics in developed compared to developing countries. Furthermore,

t
this research aims to propose a framework for returning products to give insights and recommendations to

ip
mass-production companies to maximize economical, societal, and environmental profits, consumers’

cr
loyalty, and brand reputations while minimizing waste, returning products, and inventory levels, and bad

social/human effects align with ethical aspects and SDGs. We also uncover data-driven insights to master

us
sustainable electronics supply chain decision-making in developed versus developing nations in three

separate case studies. Specifically, we utilize social media analytics to examine consumer perspectives on
an
electronics reverse logistics and end-of-life management in detail. By leveraging the X (formerly Twitter),

Meta (Facebook), and Amazon API in near real-time, we collected nearly 700 million posts related to
m
returns and recycling of electronic devices over 15 months. We extracted sentiments, patterns, and

expectations around reverse logistics across geographic settings through machine learning and natural
d

language processing. Social media analytics offer a scalable and real-time means to capture consumer
te

opinions and preferences, providing a rich data source for supply chain decisions (Boone et al., 2019;
ep

Grover et al., 2022; Shahidzadeh and Shokouhyar, 2022b). Findings in case studies provide electronics

firms with tailored insights to enhance decision-making at the nexus of smart, sustainable supply chain
cc

management. Contributions include empirical insights into how firms can leverage social media data

mining to establish alignment with SDGs and responsible disposal amidst the circular economy. This study
A

informs proactive decision-making to competitively manage product returns and reclaim value from end-

of-life devices by mastering real-time understanding of consumer viewpoints. Overall, case studies results

aim to elucidate electronics supply chain decisions optimized for sustainability and societal goals based on

social media analytics across developed and emerging economies.

3
 ACCEPTED MANUSCRIPT

t
ip
cr
us
an
Figure 1. Conceptual framework of the research, including goals, strategies, achievements, and leading and lagging consequences
while identifying research gaps in the literature
m
The circular economy concept aligns with several Sustainable Development Goals because it focuses on

resource efficiency, waste reduction, and sustainable consumption and production. Here are some SDGs
d

that are particularly related to the circular economy and will be improved by applying the proposed model:
te

SDG 9 (Industry, Innovation, and Infrastructure) emphasizes the importance of sustainable industrialization
ep

and innovation, which are key aspects of the circular economy. It encourages the adoption of cleaner and

more resource-efficient technologies, eco-design, and sustainable manufacturing processes. SDG 11

cc

(Sustainable Cities and Communities) focuses on creating sustainable and resilient cities. The circular

economy approach can help cities reduce waste, improve resource management, and promote sustainable
A

consumption and production practices within urban areas. SDG 12 (Responsible Consumption and

Production) explicitly addresses sustainable consumption and production patterns, which are fundamental

to the circular economy. It promotes resource efficiency, waste reduction, recycling, and sustainable

management of natural resources. SDG 13 (Climate Action), The circular economy can significantly

contribute to climate action by minimizing greenhouse gas emissions, reducing energy consumption, and

4
 ACCEPTED MANUSCRIPT

promoting the use of renewable resources. It aligns with efforts to mitigate climate change and transition to

a low-carbon economy. SDG 14 (Life Below Water) and SDG 15 (Life on Land), The circular economy can

help reduce pollution and waste generation, contributing to preserving marine and terrestrial ecosystems.

Promoting sustainable practices, such as recycling and responsible resource management, supports

biodiversity conservation and ecosystem protection. SDG 17 (Partnerships for the Goals), The circular

t
economy requires collaboration and partnerships among governments, businesses, and civil society to

ip
achieve its objectives. SDG 17 emphasizes the importance of multi-stakeholder partnerships and knowledge

cr
sharing for sustainable development, including circular economy initiatives. While these SDGs are closely

related to the circular economy, it is essential to note that the circular economy concept spans multiple goals

us
and is inherently interlinked with various aspects of sustainable development. The circular economy

approach promotes a systemic shift towards a more sustainable and regenerative economic model.
an
1.1. Description of the problem, identification of research gap, and explanation of the motivation
m
behind the study.

Effective supply chain management is crucial for firms seeking to remain competitive while balancing
d

profitability, sustainability, and societal goals (Cai and Choi, 2020). Simultaneously, principles of
te

sustainable development have become integral, as evidenced by the UN Sustainable Development Goals

focus on ecological sustainability and human well-being (Tseng et al., 2020). These challenges are
ep

pronounced in the electronics industry, facing massive product returns and complex reverse logistics

decisions amidst short lifecycles and frequent new product releases (Agrawal and Singh, 2019; Shahidzadeh
cc

and Shokouhyar, 2022a). However, research on data-driven approaches to aligning smart supply chains
A

with sustainability across geographic settings remains limited. As Figure 1 shows, a key gap is

understanding how social media analytics can provide insights into optimal mass-production industries

reverse logistics decision-making consistent with SDGs and circular economy principles tailored to

developing versus developed economies. This research addresses these gaps by utilizing social media

mining to uncover actionable intelligence on consumer perspectives for electronics supply chain decisions

5
 ACCEPTED MANUSCRIPT

at the nexus of SDGs alignment. By leveraging the social media API to examine sentiments around product

returns and recycling, this study provides motivational insights into how firms can leverage consumer

viewpoints for proactive, sustainable decision-making. Findings aim to elucidate electronics supply chain

decisions optimized for SDGs principles based on social listening analytics across national contexts.

Overall, this research is motivated by the need for data-driven insights to master sustainable mass-

t
production industries' supply chain decision-making. This research aims to address the following key

ip
questions:

cr
Q1: How can social media analytics reveal consumer perspectives to proactively inform mass-production

industries manufacturers' strategic reverse logistics decision-making in line with SDGs?

us
Q2: What are optimal product return and recycling policies for electronic devices that balance consumer
an
preferences and zero waste strategies across developing vs. developed countries?

Q3: How can a holistic social media mining model provide tailored benchmarks to master sustainable
m
reverse logistics decision-making for electronics across geographic contexts?
d

The research questions focus on leveraging social data to uncover consumer insights, guide zero waste
te

approaches, and develop optimized reverse logistics decision frameworks across national settings - all to

master sustainable supply chain strategy aligned with SDGs. This research makes the following key
ep

contributions by answering the research question of mastering sustainable electronics reverse logistics

decision-making:
cc

Q1: The proposed social media analytics model aligns consumer perspectives with manufacturer
A

profitability and strategic decision-making to optimize reverse logistics policies while meeting SDGs and

circular economy concepts.

Q2: By considering disposal alternatives like reuse and remanufacturing, the model provides insights to

minimize returns and electronic waste in line with zero waste strategies across geographic settings. The

new concept of optimized material stream from one location to another location is proposed. The case study

6
 ACCEPTED MANUSCRIPT

of laptops, mobile phones, and tablets as electronics are applied to the proposed model, and results were

compiled to include some recommendations aligned with circular economy and SDGs.

Q3: The model generates tailored benchmarks for the electronics industry while remaining generalizable

across manufacturing sectors. The framework informs complex strategic decisions around product returns

and sustainability by leveraging consumer insights from social data.

t
ip
Overall, this study uniquely employs social listening and consumer sentiment mining to guide mass-

production industries in optimizing reverse logistics and recycling policies tailored to geographic contexts.

cr
The data-driven insights contribute to mastering supply chain decision-making at the nexus of sustainability

us
principles and alignment with societal goals. This research proposes a theoretical and practical social media

analytics model to aid mass-production supply chain decision-makers optimize reverse logistics strategies
an
aligned with circular economy ideals and sustainability principles (SDGs). By leveraging deep learning to

mine consumer perspectives on product quality and expectations, the model provides tailored benchmarks
m
to improve circular supply chain performance across geographic contexts. The designed framework is

applicable for mass-produced products because these productions are often modular, allowing for different
d

and optimized decisions regarding their components after the product is returned to the manufacturer. The
te

advantage of this framework becomes more evident when considering customers' varying preferences in

different geographical situations, such as developed and less developed countries, where some of the usable
ep

modules can be reused in new devices in different geographical locations. Specifically, the framework

connects crucial consumer insights from social media to inform mass-production manufacturers’ strategic
cc

and policy decisions around product returns and recycling. The model enables data-driven decision-making
A

consistent with zero-waste strategies by aligning reverse supply chain management with near real-time

monitoring of consumer opinions segmented by product features. Overall, this research introduces a

competitive advantage for mass-production firms like electronics, foods, and automobiles by employing

cutting-edge deep learning techniques for social media mining. The model contributes specifically to

7
 ACCEPTED MANUSCRIPT

mastering sustainable supply chain decision-making by elucidating electronics reverse logistics policies

optimized for SDGs based on a continuous understanding of consumer viewpoints mined from social data.

The remainder of this paper is structured as follows. Section 2 provides a review of relevant literature.

Section 3 discusses the research methodology. Section 4 applies the proposed model to the electronics

industry, specifically laptops, mobile phones, and tablets, through a comparative analysis of developed

t
versus developing countries. This section also presents data analysis, results, and policy recommendations.

ip
Section 5 offers a discussion of findings and additional analysis. Finally, section 6 details the theoretical

cr
and practical research implications. Furthermore, this section provides concluding remarks, highlights

unique contributions, acknowledges limitations, and suggests future research directions to advance the

us
knowledge in this emerging domain.

2. literature review an
Sustainable manufacturing and circular supply chains have become imperative, with firms facing pressure
m
to develop environmentally friendly goods and zero-waste strategies (Ghobakhloo et al., 2022; Leng et al.,

2022). Reverse logistics is critical for product-based companies to enact circular supply chain management,
d

though barriers exist in emerging economies (Nag et al., 2021). Consumer perspectives represent a key
te

obstacle in optimizing circular supply chains across geographic contexts (Ayati et al., 2022; González-
ep

Sánchez et al., 2020). Prior research has examined circular supply chain leadership and electronic waste

management, though insights into leveraging social data remain limited (Shahidzadeh and Shokouhyar,
cc

2022a). This research addresses gaps in employing social media analytics to uncover consumer insights for

tailored reverse logistics decision-making aligned with sustainability goals across developing and
A

developed nations. Findings will elucidate electronics supply chain decisions optimized for SDGs

principles based on mining consumer perspectives from social media data. A few studies have proposed

frameworks connecting manufacturer needs with sustainability while meeting consumer expectations,

promoting efficient investment decisions aligned with circular economy ideals (Govindan et al., 2021;

Majeed et al., 2021). However, research on leveraging social data to optimize reverse logistics policies

8
 ACCEPTED MANUSCRIPT

tailored to consumer preferences across geographic contexts remains limited. Social media analytics have

gained traction as consumer perspectives shared on social media provide inexpensive yet rich insights

shaping sustainability and reverse supply chain decisions (Govindan and Bouzon, 2018; Govindan and

Soleimani, 2017). Still, few studies employ user-generated data for holistic models balancing profitability,

sustainability, zero waste strategies, and consumer sentiment to inform context-specific reverse logistics

t
choices identified as 7Rs in the literature. As covered in the case study, this research gap is pronounced

ip
between developed and emerging economies, with barriers to reverse logistics implementation differing

cr
across national maturity levels (Moktadir et al., 2020b). However, no studies compare optimized reverse

logistics decision-making between developed and developing countries, leading to identifying an optimum

us
new material stream. This points to the need for social media mining models that can benchmark tailored

product return policies across geographic settings to master sustainable electronics supply chain

management.
an
m
2.1. Sustainable Development Goals (SDGs)

Sustainable development has become an integral focus for responsible supply chain management. The
d

United Nations Sustainable Development Goals provide a shared blueprint for peace and prosperity for
te

people and the planet by 2030 (United Nations, 2015). The 17 SDGs call for ending poverty, protecting the

environment, and ensuring human well-being. Supply chain and operations management research has
ep

sought to provide insights into how firms can align strategy with SDGs (Agrawal et al., 2022). Several

SDGs hold particular relevance for electronics reverse logistics and closed-loop supply chains. SDG 12
cc

promotes responsible consumption and production, including environmentally sound management of

A

chemicals and wastes (United Nations, 2015). SDG 9 emphasizes building resilient infrastructure and

sustainable industrialization. SDG 11 focuses on sustainable cities and communities, while SDG 6 aims to

ensure sustainable water usage in manufacturing. Integration of circular economy principles with sharing

economy and servitization business models can also accelerate progress across multiple SDGs (Ferasso et

al., 2020). Critically, the SDGs highlight the need for a triple bottom line approach balancing social,

9
 ACCEPTED MANUSCRIPT

environmental, and economic priorities (Neri et al., 2021). This requires firms to align competitiveness and

profitability with ecological sustainability and human well-being. While literature acknowledges the rising

importance of SDGs for supply chains, research remains limited to leveraging social media data to uncover

consumer perspectives for optimizing reverse logistics policies tailored to SDGs across geographic

contexts. This study addresses that gap by developing insights into electronics closed-loop supply chain

t
decision-making optimized for localized SDG alignment based on mining social media analytics. We

ip
scrutinized through very recent research on SDGs and circular economy in Q1 journals and summarized

cr
ten papers with the most relevancy in Table 1.

Table 1. Very recent research on SDGs and the circular economy

us
# References Journal Title Descriptions
A recent study employs Pythagorean fuzzy AHP and CODAS
methods to prioritize circular economy practices and align
sustainable development goals in an Indian manufacturing circular

1
Lahane S,
Kant R.
Waste
Management
(Lahane and
Kant, 2022)
an supply chain. The integrated model identifies government,
management, and economic initiatives as most influential for
adoption. Mitigating waste and enhancing environmental
sustainability emerged as the top realized SDG. The research
provides a robust decision framework to guide practitioners in
m
implementing circular economy practices to achieve sustainability
goals in emerging economy supply chains.
A recent study uses system dynamics modeling to assess the social
sustainability impacts of transitioning to a circular phosphorus
economy globally and regionally. While circular practices reduced
d

poverty and improved work conditions in some areas, results were

Science of The mixed on goals like gender equality, child labor, and nutrition
te

(El Wali et al.,

2 Total security. A paradoxical effect was found where circular strategies
2021)
Environment secured phosphorus supply but exacerbated social issues across
certain stakeholders. The research indicates a need for holistic
assessment and integration of economic, environmental, and social
ep

dimensions for an equitable and just circular economy transition

aligned with Agenda 2030 sustainable development goals.
A study conducted in the UAE examines the role of reverse
logistics in achieving circular economy objectives for major
cc

retailers. Interviews with reverse logistics specialists identify key

mechanisms, including product design, waste reduction, return
technologies, and data traceability. These capabilities promote
Production
(Butt et al., reuse, repair, and recycling while minimizing environmental
3 Planning &
2023) impact. The study offers practical guidance for practitioners to
A

Control
develop sustainable and socially responsible business models using
reverse logistics networks, aligning with relevant UN Sustainable
Development Goals. It highlights the potential of reverse logistics
advancements in driving circular economy goals across economic,
social, and environmental dimensions.
In this research, the focus is on investigating the closed-loop
supply chain within the framework of the circular economy. The
Business
(Zarbakhshnia study proposes an analytical approach for ranking third-party
4 Strategy and the
et al., 2023) logistics service providers (3PLSPs) based on their sustainability
Environment
performance. The proposed method combines fuzzy Decision-
Making Trial and Evaluation Laboratory (DEMATEL) and fuzzy

10

10
 ACCEPTED MANUSCRIPT

analytic network process (ANP) techniques, incorporating expert

opinions. A case study involving household appliances is
conducted to validate the model, providing essential criteria for
decision-making in the context of the circular economy and
contributing to promoting sustainable development.
While the circular economy is commonly linked with
sustainability, its contribution to achieving the SDGs, particularly
social aspects, has yet to be substantiated. Existing sustainability
assessment methods, derived from industrial ecology and supply
chain management, are suitable for evaluating the impact of CE
practices and meeting SDG targets, but they often overlook social
Sustainable
(Walker et al., inclusion. This research addressed this gap by interviewing leading
5 Production and

t
2021) companies actively involved in CE. The objective was to gain
Consumption

ip
insights into their perspectives on the social dimension, identify
barriers to social assessment, and learn from their experiences. The
findings indicate that these companies acknowledge the
significance of the social dimension; however, they encounter

cr
challenges in conducting social assessments due to the complexity
involved and the absence of standardized approaches.
This research explores the potential of combining Circular
Economy (CE) practices and industry 4.0 (I4.0) technologies to

us
achieve the Sustainable Development Goals (SDGs). A literature
Sustainable review of 50 articles highlights that the CE-I4.0 nexus directly
(Dantas et al.,
6 Production and contributes to SDGs 7, 8, 9, 11, 12, and 13. The study suggests that
2021)
Consumption this nexus is crucial for achieving the SDGs by integrating
an innovative technologies with circular production and business
models. Further research can explore quantitative impacts,
secondary effects, and case studies within the CE-I4.0 framework.
Food loss and waste pose significant challenges in the food
industry, impacting the entire food supply chain (FSC) and causing
m
environmental degradation, economic losses, and hunger. To
tackle these issues, this study examines adopting circular economy
(CE) principles in the FSC. Through a literature review and expert
discussions, 15 critical challenges are identified and analyzed
Business
(Kumar et al., using interpretive structural modeling (ISM) and grey decision-
d

7 Strategy and the

2023) making trial and evaluation laboratory (DEMATEL) techniques.
Environment
The study highlights the importance of government policies,
te

incentives, and environmental regulations as key challenges.

Addressing these challenges can facilitate CE adoption and foster
corporate social responsibility (CSR) practices. The study
recommends that practitioners embrace CE and achieve
ep

sustainable development goals.

Circular economy is essential for sustainable development goals,
but research on its relationship with sustainable development is
limited. The leather industry, known for its environmental impact,
has the potential for circular economy practices. A study proposes
cc

Sustainable
(Karuppiah et a framework to evaluate inhibitors to circular economy in leather.
8 Production and
al., 2021) Through a literature review, 25 inhibitors are identified, including
Consumption
uncertain consumer demand, lack of social awareness, stakeholder
agendas, technology gaps, and waste management challenges.
A

These findings guide industry and government in implementing

circular economy strategies in the leather sector.
During the 16th International Conference on Waste Management
and Technology, a special session discussed sustainable practices
and solutions in the circular economy. The session focused on
downstream materials and covered new strategies, innovative
Circular (Khajuria et al., technologies, and management methods. This article summarizes
9
Economy 2022) the key insights from the presentations, emphasizing the circular
economy's role in achieving sustainable development goals. The
findings highlight the circular economy's potential to maximize
resource value, minimize waste production, and promote the 3Rs
through governmental policies and public-private partnerships.

11

11
 ACCEPTED MANUSCRIPT

Furthermore, the circular economy positively impacts natural

resource utilization and overall sustainability.
Indonesia faces waste management challenges, such as collection,
transportation, and reliance on landfills. This study aims to
develop an innovative and sustainable waste management system
using industry 4.0 technologies. It employs a multidimensional
Journal of approach, assesses the waste management system's maturity, and
(Fatimah et al.,
10 Cleaner proposes a strategy to minimize problems. The proposed system
2020)
Production incorporates circular economy principles, separates municipal
waste, identifies waste characteristics, and implements sustainable
treatment technologies using IoT. It aligns with several SDGs and
can potentially deliver positive economic, social, and

t
environmental outcomes.

ip
2.2. Circular Economy and Circular Supply Chains

cr
Circular economy principles have gained increasing attention to enable more sustainable and resource-

us
efficient supply chains (Alamerew and Brissaud, 2020). CE seeks to transition from the traditional linear

"take-make-dispose" economic model to a closed-loop system that eliminates waste through product life

an
extension, materials recovery, and regeneration (Burke et al., 2023). Strategies like reuse, remanufacturing,

and recycling aim to circulate products, components, and materials to retain value (Moktadir et al., 2020a).
m
CE requires a fundamental rethinking of supply chain flows to be restorative and regenerative by

design(Julianelli et al., 2020). Reverse logistics enables circular supply chains (CSCs) to collect, transport,
d

sort, and direct used products into recovery processes. CSCs also entail collaboration across supply chain
te

stakeholders, including suppliers, manufacturers, retailers, and consumers (Mishra et al., 2023). Consumer

perspectives play a vital role in CSCs, as product returns initiate reverse network flows while expectations
ep

shape design, production, and end-of-life management (Shahidzadeh and Shokouhyar, 2022b). However,

research into leveraging consumer insights from social media analytics to optimize CSC decisions is
cc

limited. This study addresses that gap by developing a data-driven CSC model tailored to consumer

viewpoints on mass-production companies across geographic contexts. Findings will elucidate circular
A

strategies for electronics aligned with SDGs and circular economy principles based on mining consumer

sentiments from social media.

2.2. Social media and sustainability

12

12
 ACCEPTED MANUSCRIPT

Social media platforms have emerged as impactful tools for sustainability communication and motivating

sustainable behaviors (Moslehpour et al., 2023). Research indicates that international corporations leverage

social media for sustainability initiatives across sectors (Tseng et al., 2019). Social data also enables

quantifying public perspectives on sustainability issues like pollution (Shan et al., 2020). X, Meta, and

Amazon are extensively used in research, given their real-time insights and diverse functionality for

t
analyzing opinions (Ahmadi et al., 2020; Shahidzadeh et al., 2022; Singh et al., 2018). Mining social media

ip
data sentiment provides a valuable understanding of consumer viewpoints on corporate sustainability and

cr
product perceptions (Asghar et al., 2019). Such real-time social listening helps firms gauge reputational and

economic impacts of sustainability practices (Bangsa and Schlegelmilch, 2020). Consumer opinions shared

us
on social media ultimately influence returns, purchasing, and circular economy strategies - necessitating

analysis across geographic contexts. However, research into leveraging social data to inform closed-loop
an
supply chain decisions tailored to localized sustainability principles remains limited. This study addresses

this gap by developing a mass-production industry social media analytics model that benchmarks optimized
m
reverse logistics policies across developing and developed nations. Findings will elucidate how social

listening can enable nuanced SDGs, circular economy, and zero-waste strategies aligned with consumer
d

perspectives.
te

2.3. Deep Learning and Sentiment Analysis

ep

Deep learning has emerged as a revolutionary approach in artificial intelligence, enabling computational

models to learn and make predictions from vast data. Deep neural network architectures can capture
cc

complex patterns and feature representations for accurate predictive modeling and classification (Srinivasu
A

et al., 2021). In natural language processing, deep learning techniques like recurrent neural networks and

convolutional neural networks are well-suited for text analysis tasks (Lauriola et al., 2022). Sentiment

analysis uses text data to identify subjective opinions and emotions toward entities like products,

organizations, issues, or events (Nassif et al., 2021). Advancements in deep learning have enhanced

sentiment analysis capabilities to understand nuanced linguistic expressions and semantics in textual data

13

13
 ACCEPTED MANUSCRIPT

(Yadav and Vishwakarma, 2020). Key applications include product/service feedback monitoring, brand

monitoring, understanding customer satisfaction, and gauging public opinion. For supply chain

sustainability research, deep learning sentiment analysis enables extracting insights from consumer

perspectives shared on social media (Shahidzadeh et al., 2022). However, limited research employs these

techniques to optimize closed-loop supply chain decision-making tailored to geographic contexts while

t
meeting some SDGs’ pillars. This study addresses this gap by developing a deep learning and sentiment

ip
analysis methodology using social media data on mass-production companies to support reverse logistics

cr
policies aligned with circular economy ideals across developing and developed nations. Findings will

elucidate the value of deep learning-driven social media analytics for sustainable supply chains aligned

us
with SDGs’ pillars.

an
2.3.1. Application of deep learning techniques in sentiment analysis tasks

Sentiment analysis encompasses a range of tasks to extract subjective information from textual data,
m
including emotion detection, opinion mining, stance detection, and more (Nagamanjula and Pethalakshmi,

2020). Traditional machine learning approaches relied on feature engineering, but deep learning models
d

can automatically learn robust feature representations from text (Mohammad, 2022). Here, we review key
te

sentiment analysis tasks being advanced with deep learning:

ep

- Document-level sentiment classification - Determines overall sentiment of a document as

positive, negative, or neutral using convolutional and recurrent neural networks (Table 2) (Moraes
cc

et al., 2013).

- Sentence-level sentiment classification - involves the classification of individual statements

A

within a document (Zhang et al., 2019). However, it is important to note that not every statement

can be assumed to be an opinionated sentence. In sentence-level sentiment analysis, sentences are

categorized as negative, neutral, or positive (as shown in Table 2).

14

14
 ACCEPTED MANUSCRIPT

- Aspect-based sentiment analysis - Models like attention-based LSTM networks identify

sentiment towards specific aspects of text (Table 2) (Bensoltane and Zaki, 2023).

- Emotion detection - RNNs classify texts by emotions like joy, sadness, anger, etc. (Bokhare and

Kothari, 2023).

- Sarcasm detection - Crucial for interpreting sentiment, especially on social media; CNNs help

t
ip
detect sarcastic expressions (Govindan and Balakrishnan, 2022).

cr
- Stance detection - Determines perspective of text towards a target; memory networks show

promise for modeling stance (Alturayeif et al., 2023).

us
- Hate speech detection - Detecting abusive language is critical for analyzing online conversations;

an
CNNs effectively categorize hate speech (William et al., 2022).

These deep learning advances provide capabilities to extract nuanced consumer opinions from social data.
m
However, research applying these techniques to inform supply chain sustainability strategies is limited. This

study develops deep learning sentiment analysis tailored to electronics reverse logistics contexts across
d

geographic regions. Findings will demonstrate the value of deep neural networks for optimized, data-driven,
te

sustainable supply chain decision-making.

ep

Detection Emotions Sarcasm Stance/Attitude Hate speech

cc

Sentiment
analysis
A

Levels Document level Sentence level Aspect level

Figure 2. Illustration of the interconnection between sentiment analysis and the different levels of categorization and Detection of
the emotions

15

15
 ACCEPTED MANUSCRIPT

The key link between sentiment analysis and emotion detection is the use of textual cues like words,

phrases, and syntax to infer the underlying affective state. Both tasks rely on natural language processing

to parse text and extract semantic meaning. However, emotion detection requires more advanced techniques

like contextual modeling and deeper understanding of linguistic nuances to distinguish subtle emotional

shades. While sentiment analysis provides coarse categorization of positivity, negativity, and neutrality,

t
emotion detection offers finer-grained labeling of specific feelings. Multiclass emotion detection represents

ip
a more challenging machine-learning task than binary or three-class sentiment classification. Overall,

cr
progress in sentiment analysis capabilities provides a foundation for more nuanced emotion identification

from text (Figure 2).

us
Table 2. Recent research has focused on examining sentiment analysis from three distinct levels
Sentiment
No. References Description
classification
an
The article discusses sentiment classification, where sentiment attributes are learned
from annotated data to predict sentiment polarity in new text. Different domains
have varied expressions of feelings, affecting sentiment distribution. Lack of
(Lei and Li,
1 annotated data in specific domains requires utilizing existing corpus. The article
2021)
m
proposes an algorithm to extract sentiment sentences from product reviews and
enhances BERT for cross-domain sentiment classification. Experimental results
demonstrate the effectiveness of the proposed method.
Sentiment Analysis uses machine learning techniques to classify review texts into
positive, negative, or neutral sentiments. This paper explores traditional machine
d

learning and deep learning approaches for multiclass polarity classification on a

(Paramesha
2 cross-domain review dataset. Ensemble models with semantic and statistical features
et al., 2023)
perform well, and combining them with a sentiment-document model enhances
te

Document sentiment prediction. Deep learning models outperform traditional methods,

level achieving higher accuracy and F1 scores in multi-class polarity classification.
This paper introduces AttBiLSTM-2DCNN, a neural network model for sentiment
ep

analysis. It addresses challenges in modeling long texts and differentiating

(Mao et al., document features. The model combines bidirectional LSTM layers, a 2DCNN, and
3
2022) a two-layer attention mechanism to capture sentiment semantics and dependencies.
Evaluation of public review datasets demonstrates superior performance over state-
of-the-art models.
cc

This research paper presents MgJST, a solution for concurrent sentiment and topic
analysis in online reviews. In a probabilistic model, the approach combines
(F. Huang et sentence-level structural knowledge with latent Dirichlet allocation (LDA).
4
al., 2022) Evaluation of seven sentiment analysis datasets shows that MgJST outperforms
A

existing unsupervised approaches (WSTM and STSM) in sentiment detection

quality and topic extraction capability.
This research explores the correlation between Aspect-Based Sentiment Analysis
(ABSA) and Sentence-Based Sentiment Analysis (SBSA). A unified-trained deep
(Chiha et al.,
5 learning model is proposed for ABSA to improve performance. For SBSA, a hybrid
2022)
model combining deep learning and fuzzy logic is developed, achieving satisfactory
Sentence level performance compared to the Bert model.
This study addresses the limited contextual information in sentiment analysis of
(Altaf et al., Urdu text and the challenge of classifying sentiments in proverbs and idioms.
6
2023) Linguistic features of the Urdu language are leveraged for sentence-level sentiment
analysis, and classical machine learning techniques are employed for idioms and

16

16
 ACCEPTED MANUSCRIPT

proverbs classification. Experimental results show that the J48 classifier achieves
the highest performance with 90% accuracy and an F-measure of 88%.
This study evaluates language-specific sentiment analysis methods against a
baseline approach for social media content. Comparing sixteen English methods
(Araújo et with three language-specific methods on multiple datasets, the surprising finding is
7
al., 2020) that translating text to English and using the best English method outperforms the
language-specific approaches. The study releases codes, datasets, and the iFeel 3.0
system as a new multilingual sentence-level sentiment analysis baseline.
The paper introduces a neural network-based method, Convolution over a
Dependency Tree (CDT), for sentiment polarity determination of opinion words
related to specific sentence aspects. The model combines Bi-LSTM for sentence
(Sun et al.,
8 feature learning and a Graph Convolutional Network (GCN) operating on the
2019)

t
sentence's dependency tree. This integration improves representation learning by

ip
propagating contextual and dependency information, leading to state-of-the-art
performance in aspect-based sentiment classification.
This paper addresses aspect-level sentiment classification by introducing a new
Syntax- and Knowledge-based Graph Convolutional Network (SK-GCN) model.

cr
SK-GCN utilizes GCN to incorporate a syntactic dependency tree and commonsense
Aspect level (Zhou et al.,
9 knowledge. It combines two strategies, S-GCN and K-GCN, and incorporates pre-
2020)
trained BERT, achieving state-of-the-art performance. Experimental results on

us
benchmark datasets validate the approach's effectiveness in improving aspect-level
sentiment classification.
This paper discusses the importance of sentiment analysis at the aspect level,
specifically in analyzing customer reviews on online commercial websites. Existing
(Nandal et algorithms often overlook bipolar words that change polarity based on context. The
10
al., 2020) an
proposed approach focuses on item features and incorporates aspect-level sentiment
detection. It is implemented and tested using Amazon customer reviews, employing
preprocessing operations and assigning sentiment ranks for classification.
m
3. Research methodology

To ensure a systematic approach, this research first introduces the overall framework. Next, a novel human
d

and society-centric circular supply chain model is proposed utilizing deep learning techniques across four
te

key stages: social media analytics, feature extraction from mass-production products, sentiment analysis of
ep

user-generated data, and data-driven recommendations with industry benchmarks for product returns. The

study then details the deep learning methodology employed in the model, including discussion of advanced
cc

neural network architectures tailored for text mining and sentiment classification. By leveraging cutting-

edge deep learning, this research provides a robust social media analytics framework to optimize
A

decentralized consumer-focused decision-making for mass-production reverse logistics and waste

management aligned with sustainability goals (SDGs). The data-driven model informed by real-time social

listening thereby enables mastering circular supply chain strategies across developing and developed

markets.

3.1. The framework of the research

17

17
 ACCEPTED MANUSCRIPT

The research methodology is visually depicted in Figure 3 across four key phases. First, an extensive

literature review is conducted on pertinent concepts, including circular economy, SDGs, sustainability,

social media analytics, deep learning, and sentiment analysis. A critical gap has been identified in

leveraging social data to enable circular strategies, though some recent studies have explored different

reverse logistics approaches (Agrawal and Singh, 2019; Shahidzadeh and Shokouhyar, 2022a). The second

t
phase involves collecting large-scale social media user-generated data such as X, Meta, and Amazon

ip
datasets related to consumer mass-production manufacturing, such as electronics, through their APIs. Next,

cr
state-of-the-art deep learning techniques are leveraged to extract product features and analyze sentiments

from the textual data. Finally, optimized recommendations and benchmarks for electronics reverse logistics

us
policies are generated based on social media insights across developing and developed markets. This data-

driven approach enables mastering circular supply chain decision-making tailored to consumer perspectives

mined from social media data.

an
m
d
te
ep
cc
A

18

18
 ACCEPTED MANUSCRIPT

t
ip
cr
us
an
m
d
te
ep

Figure 3. The suggested framework for conducting the research methodology

The second phase involves developing a novel human-centric circular supply chain model based on some
cc

SDGs pillars related to reverse logistics leveraging social media analytics and sentiment mining of social

media posts (Shahidzadeh and Shokouhyar, 2022a). The model provides optimized recommendations for
A

product returns to supply chain decision-makers based on circular economy principles aligned with SDGs.

Alternative reverse logistics dispositions like reuse, recycling, and remanufacturing are suggested to

minimize waste. Industry benchmarks are also derived by applying the model across mass-production firms.

This benchmark allows managers to shape a new stream of materials and modules from developed nations

19

19
 ACCEPTED MANUSCRIPT

to less developed ones. In the third phase, a parallel case study of laptops, mobile phones, and tablets will

be conducted in developing versus developed markets to validate the model. Social media data is analyzed

to propose tailored best practices for reverse logistics for laptops, mobile phones, and tablets based on

consumer insights in each geographic context. The case findings establish electronics industry benchmarks

for circular economy strategies while meeting some SDGs pillars. Finally, the fourth phase discusses

t
theoretical and practical implications, highlights key contributions in leveraging social data analytics to

ip
advance sustainable supply chain decision-making, acknowledges limitations, and suggests future research

cr
directions. In summary, this data-driven approach enables mastering localized circular economy strategies

optimized for consumer preferences mined from social media data.

us
3.2. Proposed model/framework

an
Figure 4 illustrates the proposed model, which comprises two primary components. The first component

involves identifying keywords and hashtags associated with mass-production goods from social media posts
m
and manufacturer documentation through content analysis. This step enables extracting key product features

and attributes specific to each product model. The second stream gathers user-generated social media data
d

containing the identified terms. Analytical processing filters this content for descriptive analysis and
te

sentiment modeling using deep learning. Specifically, neural network architectures extract product aspects

from posts and classify sentiment as positive, negative, or neutral. This data-driven approach enables a
ep

human and social centric circular supply chain by connecting consumer perspectives to reverse logistics

decisions. As a case study, the model is applied to the electronics industry but can be generalized for circular
cc

economy benchmarking across mass-production manufacturing sectors. In summary, this advanced social
A

listening framework mines consumer opinions at scale to optimize sustainable supply chain strategies

tailored to user sentiments while meeting SDGs requirements. The proposed model is general enough for

mass-production manufacturing industries such as electronics, automobiles, foods, and textiles. After

applying the model to mass-production manufacturing goods, we ensured that SDG9, SDG11, SDG12,

SDG13, SDG14, SDG15, and SDG17 are satisfied in the circular economy context.

20

20
 ACCEPTED MANUSCRIPT

1) Social media data undergoes preprocessing, including stopword removal, tokenization, POS

tagging, and n-gram generation to improve deep learning performance for feature/model extraction

and sentiment analysis. Content analysis also informs feature extraction accuracy.

2) Aspect-based feature extraction leverages automated deep-learning techniques suited for

sentence-level social media posts. Experts validate extracted features to derive optimized

t
categorization based on relevant terms.

ip
3) State-of-the-art deep neural networks conduct granular sentiment analysis on each identified

cr
product feature. Categorizing sentiments as positive, neutral, or negative provides nuanced insights

us
to inform consumer-centric decisions.

4) After sufficient data collection, benchmarks are generated to optimize industry reverse logistics
an
strategies over time, barring technological disruptions. Benchmarks enable agile decision-making

that is aligned with continuously mined social data.

m

In summary, this advanced deep learning approach mines granular consumer perspectives from social media
d

to guide tailored, data-driven decision-making for mass-production manufacturing circular supply chains
te

meeting SDGs pillars.

3.2.1. Social media post gathering and filtering

ep

Unstructured social media data representing valuable consumer perspectives is gathered via APIs and stored
cc

alongside manufacturer documentation (Nguyen et al., 2018). Content analysis using diverse NLP

techniques is critical for extracting insights from informal textual data like posts containing hashtags, URLs,
A

and idioms (Shahidzadeh and Shokouhyar, 2022b). Preprocessing methods, including translation to

English, duplicate removal, emoji detection, tokenization, POS tagging, stopword removal, and n-gram

generation, transform the raw data into structured inputs for robust deep learning analysis (Avasthi et al.,

2021). These techniques overcome the variety of linguistic expressions in social media to yield semi-

structured data optimized for automated feature extraction and granular sentiment modeling. By applying

21

21
 ACCEPTED MANUSCRIPT

best practices in social media content analysis and preprocessing, this research enables high-performance

deep-learning text mining tailored to consumer mass-production manufacturing perspectives even across

geographic contexts. The resulting structured data powers optimized, data-driven circular supply chain

decision-making.

3.2.2. Feature and model identification

t
ip
Automated aspect extraction using deep learning is applied for sentence-level analysis of the typically short

social media posts (Ray and Chakrabarti, 2019). This ensures all product aspects are identified from the

cr
unstructured user-generated content. Experts then filter the extracted aspects based on industry relevance

us
to derive key product features. Ranking aspects by frequency provides initial prioritization, though rankings

may shift over time. While some aspects may have semantic synonymity, a fine-grained approach is taken
an
here with expert validation of each aspect. This captures nuanced consumer opinions versus aggregating

similar aspects. The resulting list of hashtags and keywords categorizes the social data for subsequent
m
feature-based sentiment modeling. The model utilizes customized aspect extraction techniques specifically

designed for the mass-production manufacturing industry. This allows for reliable supervised learning
d

methods to be applied, enabling the extraction of consumer perspectives from social data on a large scale.
te

3.2.3. Sentiment analysis

ep

The preprocessed social media data is fed into advanced deep-learning techniques for detailed sentiment

analysis. Aspect-based classification from the prior stage focuses on modeling sentence-level expressions.
cc

Subjectivity detection further filters out objective sentences before classifying subjective sentiments as

positive, neutral, or negative (Tang et al., 2009). Deep learning provides robust supervised sentiment
A

modeling tailored to the consumer mass-production manufacturing context. Visualizations are generated to

represent the sentiment modeling results after benchmarking and descriptive analysis to summarize the data

statistically. Descriptive analytics are commonly used in supply chain intelligence to create executive

dashboards to monitor SDGs’ level of realization (Govindan et al., 2015; Shahidzadeh and Shokouhyar,

22

22
 ACCEPTED MANUSCRIPT

2022b). This enables data-driven decision-making based on a graphical representation of mined consumer

perspectives from social media.

3.2.4. Recommendations and industry benchmark

The granular sentiment analysis of each product feature enables descriptive analytics to guide managers in

human and society-centric circular supply chain strategies. These data-driven recommendations are

t
ip
aggregated from industry benchmarks documented internally and across organizations (Park and Lee,

2021). Industry and sub-industry benchmarks empower rapid, evidence-based decision-making. As product

cr
lifecycles near end-of-life, benchmarks must be re-analyzed. However, the proposed model's real-time

us
nature allows continuous integration of the latest consumer insights for agile decision optimization. Overall,

this social media analytics approach provides tailored, up-to-date benchmarks to master sustainable reverse
an
logistics tailored to mined consumer perspectives across developing and developed mass-production

manufacturing markets as a case study.

m
d
te
ep
cc
A

Figure 4. The proposed circular supply chain management framework prioritizes the well-being of humans and society. It
leverages social media analytics and deep learning methods to gather insights. This framework is designed to align with the
principles of Sustainable Development Goals.

3.3. Implementation of the proposed model/framework

23

23
 ACCEPTED MANUSCRIPT

While machine learning has been widely used, deep learning has become preferable for text mining due to

increased computing power enabling massively parallel processing on GPUs (Kamiş and Goularas, 2019).

Though complex and time-intensive to train, deep neural networks outperform traditional feature extraction

and sentiment analysis methods. Recursive CNNs and GRUs exhibit comparable performance to CNNs and

LSTMs, respectively. Ensembling CNN and LSTM architectures has shown additional accuracy gains

t
(Rayhan Ahmed et al., 2023). Recent research has focused on combining diverse neural network

ip
configurations and word embedding strategies to achieve state-of-the-art results surpassing individual deep

cr
learning models. This study leverages the most advanced deep learning techniques tailored to the social

media text analysis tasks needed for optimizing consumer-focused mass-production manufacturing reverse

us
logistics decisions. The complex deep learning approach provides the power to extract nuanced insights

from unstructured big social data.

an
3.4. Findings of the case study - a comparison between developing and developed nations
m
As depicted in Figure 4, the proposed model is designed to be universally applicable in various

manufacturing sectors. This study utilizes the model to analyze consumer electronics in both developing
d

and developed countries, aiming to uncover unique insights in each context during Phase IV of the research.
te

Comparative analysis of the sentiment modeling results reveals nuances between consumer perspectives in

emerging versus mature markets. This demonstrates how the holistic social media analytics model can be
ep

leveraged to tailor circular economy strategies based on localized consumer opinions mined from social

data. The framework provides electronics firms with optimized reverse logistics decision-making
cc

benchmarks tailored to geographic and economic contexts. Overall, the model enables mastering
A

sustainable supply chain management by generating data-driven insights customized for developing versus

developed markets.

4. The examination of a case study and the obtained outcomes and discoveries

24

24
 ACCEPTED MANUSCRIPT

Due to its broad applicability to mass-production manufacturing industries, we opt to utilize the proposed

model specifically in the field of electronics. This choice is driven by the higher generation of waste in this

industry and its significant impact on people's lifestyles. As sub-industry, we applied the proposed model

to the laptop, mobile phone, and tablet in parallel. Their feature is very similar to each other. Any social

media-supporting API can be used for data extraction. Considering the widespread usage of X, Meta, and

t
Amazon, we have chosen them as our sources to gather user-generated data. We innovatively designed case

ip
study. As social media data enabled us to identify the users' geolocations, we classified posts on a country

cr
basis. It allows us to filter out features/models by country separately. By running the model on a specific

industry, we identified a stream of turning back product modules from unsatisfied/unhappy countries or

us
regions to satisfied/happy (different geo-location usually). These recommendations give the strategic

manager of the manufacturer insight into reusing their products or a module of them. This also satisfies
an
SDG goals and circular economy criteria. Hence, the same social media data was leveraged for comparative

analysis between developing and developed countries. The social media API and geolocation attributes
m
enabled filtering posts by country development levels per the UN's WESP 2020 report (United Nations,

2020). From January 2022 to March 2023, posts were separated into developing and developed groups. The
d

model was applied to each country dataset to uncover differentiated insights and refine the robustness of
te

the framework. As depicted in Figure 4, while social media collection and feature extraction were
ep

consistent, sentiment analysis and benchmarking results differed. Comparative findings provide tailored

suggestions to policymakers on optimizing reverse logistics based on consumer perspectives in emerging

cc

versus mature markets. Sensitivity analysis across periods further validated the model's effectiveness in

supporting localized circular economy decisions. This approach demonstrates how the same social listening
A

model can be leveraged to master sustainable supply chain strategies customized to geographic and

economic contexts based on mined consumer sentiments. The human and social-centric circular supply

chain is achieved using the proposed model, and SDG9, SDG11, SDG12, SDG13, SDG14, SDG15, and

SDG17 are satisfied. The identified relationship between SDGs and the proposed model is discussed in the

introduction.

25

25
 ACCEPTED MANUSCRIPT

4.1. Social media data

The first phase involved collecting laptop, mobile phone, and tablet-related posts via X, Meta, and Amazon

search API using required authorization keys. Posts were retrieved in JSON format containing unique IDs,

authors, timestamps, geolocation, if available, and other metadata. This raw social data was then

preprocessed and analyzed using the associated keyword queried across 34 languages supported on X,

t
Meta, and Amazon (Shahidzadeh et al., 2022). Leveraging the social media API enabled large-scale

ip
gathering of consumer perspectives related to electronic devices across global contexts. Preprocessing

cr
transformed the unstructured JSON post data into inputs for subsequent deep learning textual analysis to

uncover localized insights around circular economy strategies.

us
4.1.1. Content analysis and descriptive analysis
an
Posts were analyzed by parsing key components, including author name/handle, main text, hashtags, links,

images, engagement metrics, emoticons, and geolocation. Since deep learning performance improves on
m
English text, main post text was translated from other languages. Duplicate post contents were then

removed. As shown in Table 3, this descriptive analysis extracted relevant metadata from the raw JSON
d

post data to prepare inputs for feature extraction and sentiment modeling. Preprocessing unstructured social
te

data enabled more robust deep learning analysis by overcoming language diversity. Isolating emojis also
ep

provide valuable sentiment signals (training sets) for subsequent aspect-based sentiment modeling. This

content analysis transformed informal, multilingual posts into optimized English-language inputs for
cc

mining consumer perspectives.

Table 3. Summarization of descriptive posts’ data analysis

A

Mobile phone Laptop Tablet Total

Amazon

Amazon

Amazon
Meta

Meta

Meta
X

26

26
 ACCEPTED MANUSCRIPT

199,842,859

111,853,003

670,786,820
94,396,687

31,107,054

81,524,092

39,186,550

59,411,130

37,437,770

16,027,675
All collected posts

23,981,143

13,932,951

13,690,808

13,206,903

88,274,290
4,106,131

4,279,171

6,986,749

5,705,516

2,384,918
After pre-processing

t
14,925,284

55,304,943
8,862,765

2,280,343

7,606,516

8,616,569

2,763,161

5,128,569

3,680,593

1,441,144

ip
Developing countries

cr 30,954,024
8,576,236

4,651,790

1,784,456

5,809,734

4,192,001

1,428,484

1,716,639

1,906,632

888,050
Developed countries

us
2,015,323
479,623

418,396

274,558

398,333

141,541

118,291
41,332

87,527

55,723
Location is not identified
an
m
As summarized in Table 3, the preprocessed post data was separated into developing and developed country

corpora for comparative analysis. This enabled downstream application of the social media analytics model
d

to uncover localized insights relative to geographic and economic contexts. Separating the multilingual post
te

inputs by country development levels provides the foundation for tailored circular economy

recommendations based on consumer perspectives mined from each national X, Meta, and Amazon dataset.
ep

The comparative approach highlights how the same model can be leveraged to optimize reverse logistics

strategies for electronics firms based on social listening analytics customized to developing versus
cc

developed markets.
A

4.1.2. Pre-processing

Optional text preprocessing was applied to improve deep learning accuracy on the informal post data.

Techniques including POS tagging, stemming, stopword removal and word normalization transformed the

unstructured, complex posts into more simplistic inputs optimized for sentiment modeling (Ahmadi et al.,

27

27
 ACCEPTED MANUSCRIPT

2020). Removing extraneous terms ensures that deep neural networks focus on sentiment-laden vocabulary.

As depicted in Figure 5, the resulting structured posts better expose key consumer opinions versus the

original raw social data. By tailoring preprocessing to the unique linguistic properties of social media, this

research enabled state-of-the-art deep-learning text mining tailored to informal electronics-related posts for

optimized circular economy decision support.

t
ip
Post data Raw post
warehouse
Amazon
X

cr
1. Filtering hashtags and
keyword
2. Store in data
Facebook warehouse

us
1. Separating texts,
hyperlinks, images,
Text cleaning
Tokenize
an
emojis Remove stopwords and Feature/Model extraction
m
2. Extracting meta data punctuations
Final Analytical Process
(user, like, quote, repost, POS-tagging
issued date, location, ...) Stemming
Sentiment analysis
3. Translating non-English N-grams
texts into English Remove duplicates
d
te

Training sets Deep learning

ep

Figure 5. Mainstream of data from social media to the final stage of decision-making

4.2. Feature extraction

cc

Feature extraction identified the most discussed laptops, mobile phones, and tablets attributes in the posts
A

and other documents like PDFs and webpages. Supply chain experts filtered the extracted aspects to derive

the most relevant product features. A hybrid CNN-LSTM deep learning architecture was leveraged for

optimized aspect-based categorization per prior research. Table 4 shows the expert-validated laptop, mobile

phone, and tablet features extracted from the textual data. This approach provided robust aspect-based

analysis of the unstructured social media and text documents related to electronic devices. Expert validation

28

28
 ACCEPTED MANUSCRIPT

combined with deep learning feature extraction tailored to informal textual data enabled accurate modeling

of consumer perspectives towards key product attributes for subsequent sentiment analysis.

Table 4. Feature extraction of laptops, mobile phones, and tablets with deep learning technique of aspect extraction and
categorizations, along with the number of posts separated by laptops, mobile phones, and tablets in developed and developing
countries

Laptop

Mobile

Tablet
phone
Extracted features Related keywords and Hashtags Relation to the RL

t
ip
DingC **

DingC **

DingC **
DedC *

DedC *

DedC *
cr 2,209,342

3,221,470

2,581,307

2,608,793

1,468,416
942,830
Cost, Affordable, Expensive, Discount,

us
1 Price Deal, Value, Overpriced, Pricey, Sale, INDIRECT
Purchase

Space, Capacity, Memory, Gigabytes

(GB), Terabytes (TB), Hard drive, Cloud

2,898,318

1,050,211

5,008,213

1,831,944
471,048

426,033
2 Storage an
storage, External storage, Internal
storage, Expandable storage, Storage
solutions, Solid State Drive (SSD), Hard
disk
DIRECT

Screen, Monitor, Clarity, High-definition

m
(HD), Retina display, OLED, LCD,

2,064,822

1,263,016

3,393,246

1,653,988
800,365

487,160
Touchscreen, Bezel-less, Refresh rate,
3 Display Viewing angles, LCD, Retina, DIRECT
Brightness, Graphics, GPU, Picture,
Backlight, Colour, LED, True Tone,
d

Image Quality, Pixels, Resolution 1,529,361

1,417,784

3,715,216
te

904,580

349,162

886,622
Style, Visuals, Creativity, Artistic,
4 Design Colors, Minimalist, Modern, Eleganting DIRECT
angles, Size, Weight, Beauty
ep

Protection, Cybersecurity, Data security,

1,515,275

1,662,640

2,269,007

2,545,981

631,641

607,428
Encryption, Password, Security breach,
5 Security Online safety, Hackers, Malware, Two- INDIRECT
factor authentication (2FA), Privacy,
Authentication, Secure
cc

Long-lasting, Rechargeable, Battery

1,402,724

2,927,452

1,746,666

3,694,744

1,483,091
338,974

drain, Low battery, Power, Battery Life,

6 Battery Lithium-polymer, Adapter, Fast Charge, DIRECT
Lifespan, Battery Performance, Energy-
A

Saving, Battery Health

Port, Cable, Adapter, Plug, Jack, Type-
1,000,457

1,641,208

1,691,392

2,633,453

202,829

589,781

C, Audio connector, Video connector,

7 Connector Charging port, Headphone jack, Docking DIRECT
station, USB, HDMI, Lightning, Data
Transfer, USB-C, Headphone

29

29
 ACCEPTED MANUSCRIPT

Guarantee, Warranty period, Coverage,

Service, Replacement, Extended

2,080,117

2,509,450
723,994

883,556

868,737

434,859
warranty, Warranty claim, Manufacturer
8 Warranty warranty, Customer support, Warranty INDIRECT
policy, Warranty registration, Warranty
terms, Warranty expiration, Warranty
service center, Repair, Refurbished
Software, Platform, Interface, User-
friendly, Compatibility, Updates,

1,046,311

2,316,980

1,585,190

1,294,175
483,650

263,955
Stability, Performance, Mobile OS,
9 Operating system INDIRECT
Desktop OS, Operating system version,
App store, OS, OS Performance, Speed,

t
Apps

ip
As shown in Table 4, the extracted laptop, mobile phone, and tablet features were categorized as having

cr
direct or indirect relationships with reverse logistics. Direct features like storage, display, design, battery,

us
and connector represent physical components that could be reused through circular economy strategies.

Indirect features such as price, security, warranty, and operating system reflect consumer perspectives that
an
influence reverse logistics decision-making. By classifying the key aspects into nine categories - five direct

and four indirect - this aspect-based analysis structured the unstructured social data to align with reuse
m
considerations. The granular feature extraction enabled targeted sentiment modeling in subsequent phases

to inform component-level reuse, remanufacturing, and recycling policies based on mined consumer
d

opinions.
te

4.3. Sentiment analysis

ep

A hybrid CNN-LSTM deep learning architecture was implemented for sentence-level sentiment analysis of

the posts, achieving 93.78% accuracy based on prior research (Feizollah et al., 2019). The short, limited
cc

nature of posts makes sentence-level modeling most appropriate. Training data consisted of posts with

happy and sad emoticons (Peacock and Khan, 2019). As shown in Table 5, granular sentiment analysis was
A

performed on each feature category to uncover nuanced consumer opinions. To gain additional insights,

developing and developed country posts were analyzed separately. This comparative modeling revealed

contrasts in consumer perspectives based on geographic and economic contexts. By leveraging cutting-

30

30
 ACCEPTED MANUSCRIPT

edge deep learning tailored to informal textual data, this social media analytics approach enabled granular,

accurate, and localized sentiment modeling to inform targeted circular economy strategies.

Table 5. The final results of the sentence level of the extracted posts are separated by Happy, Neutral, and Unhappy (Laptop (a),
Mobile Phone (b), and Tablet (c) case study)
(a) Laptop
Features Number of posts Final SA Results
Happy Neutral Unhappy
Price

t
DedC* 354,489 98,537 1,756,317 Unhappy
DingC** 1,933 88,107 3,131,430 Unhappy

ip
Storage
DedC 363,061 17,806 90,182 Happy
DingC 132,163 45,214 2,720,941 Unhappy

cr
Display
DedC 585,627 7,563 207,175 Happy
DingC 1,629,970 12,079 422,772 Happy

us
Design
DedC 671,108 9,408 224,064 Happy
DingC 1,051,971 71,345 406,045 Happy
Security
DedC
DingC
183,045
1,046,216 17,375 an
30,533 1,301,697

Battery
599,049
Unhappy
Happy

DedC 1,041,452 53,724 307,547 Happy

DingC 66,746 127,930 2,732,777 Unhappy
m
Connector
DedC 274,075 39,618 686,764 Unhappy
DingC 1,150,569 57,935 432,705 Happy
Warranty
d

DedC 1,757,386 43,370 279,360 Happy

DingC 24,761 5,828 693,405 Unhappy
te

Operating system
DedC 904,484 15,485 126,342 Happy
DingC 152,805 1,506 2,162,669 Unhappy
ep

(b) Mobile phone

Features Number of posts Final SA Results
Happy Neutral Unhappy
cc

Price
DedC 2,052,397 19,102 509,808 Happy
DingC 394,450 40,175 2,174,168 Unhappy
Storage
A

DedC 878,921 50,515 120,774 Happy

DingC 777,525 151,498 4,079,190 Unhappy
Display
DedC 886,132 28,607 348,277 Happy
DingC 1,068,363 143,704 2,181,178 Unhappy
Design
DedC 834,366 9,074 574,344 Happy
DingC 1,126,639 66,317 2,522,260 Unhappy
Security
DedC 530,721 106,416 1,631,870 Unhappy
DingC 1,919,924 112,914 513,143 Happy

31

31
 ACCEPTED MANUSCRIPT

Battery
DedC 1,073,850 53,448 619,368 Happy
DingC 1,175,298 58,377 2,461,069 Unhappy
Connector
DedC 513,929 61,482 1,115,980 Unhappy
DingC 831,644 27,388 1,774,421 Unhappy
Warranty
DedC 2,086,357 78,169 344,924 Happy
DingC 153,164 21,338 709,054 Unhappy
Operating system
DedC 333,235 97 150,318 Happy

t
DingC 1,069,210 23,857 492,122 Happy

ip
(c) Tablet
Features Number of posts Final SA Results
Happy Neutral Unhappy

cr
Price
DedC 349,908 42,899 550,023 Unhappy
DingC 280,247 35,315 1,152,854 Unhappy

us
Storage
DedC 124,593 10,203 291,236 Happy
DingC 280,379 87,659 1,463,907 Unhappy
Display
DedC
DingC
181,638 9,914
an 295,609
410,603 54,995 1,188,391
Design
Happy
Happy

DedC 109,235 5,953 233,973 Happy

DingC 393,527 10,772 482,323 Happy
m
Security
DedC 212,200 21,128 398,313 Unhappy
DingC 185,569 2,551 419,308 Happy
Battery
d

DedC 56,965 6,491 275,518 Happy

DingC 655,526 57,247 770,318 Happy
te

Connector
DedC 99,751 1,440 101,638 Happy
DingC 252,426 29,194 308,161 Happy
Warranty
ep

DedC 51,342 18,504 798,891 Happy

DingC 150,983 15,264 268,613 Unhappy
Operating system
DedC 62,689 11,588 189,678 Happy
cc

DingC 346,127 41,478 906,570 Happy

A

As shown in Table 5, granular sentiment analysis results for each feature were calculated separately for

developing and developed country laptop, mobile phone, and tablet posts. A low correlation existed between

the two contexts, though storage, security, battery, warranty, and OS satisfaction were higher in developed

markets. While display and design drove positive opinions in both regions, price garnered negative

sentiments. Figure 6 visually depicts the relative happiness/unhappiness levels per the following formula

32

32
 ACCEPTED MANUSCRIPT

for each feature for laptop, mobile phone, and tablet. In general, by identifying the gap between customer

satisfaction in developed and developing countries, a new high-value-added stream can be identified for

mass producers. This identification enables supply chain managers to achieve sustainable supply chain

benefits in economic, social, and environmental dimensions by creating product flows. Additionally,

placing humans and society at the center helps achieve the Sustainable Development Goals and contributes

t
to creating a safer community for the betterment of all humanity. In the upcoming section, we will delve

ip
into the value generated. This comparative sentiment modeling revealed that while some perspectives

cr
aligned, nuances existed based on geographic and economic maturity levels. By uncovering these contrasts,

this social media analytics approach provides tailored insights to guide localized circular economy

us
strategies, leveraging positive sentiments and addressing negative opinions mined from consumer

viewpoints in each market type.

an
𝑋: 𝐿𝑒𝑣𝑒𝑙 𝑜𝑓 ℎ𝑎𝑝𝑝𝑖𝑛𝑒𝑠𝑠
𝛿: 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 ℎ𝑎𝑝𝑝𝑦 𝑝𝑜𝑠𝑡𝑠
m
𝛿 ′ : 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑢𝑛ℎ𝑎𝑝𝑝𝑦 𝑝𝑜𝑠𝑡𝑠
𝛿
𝑋=
𝛿 + 𝛿′
The relative happiness score was calculated on a scale of 0 to 1 using the ratio of positive sentiment posts
d

to the total number of posts deducted from neutral ones for each product feature. This standardized metric
te

enables comparison of consumer perspectives across features and geographic regions. The granular
ep

happiness levels derived from the deep learning sentiment analysis provide electronics firms with tailored

insights into prioritizing circular economy strategies by component based on social listening analytics.
cc
A

33

33
 ACCEPTED MANUSCRIPT

Laptop
Developing countries Developed countries

Price
1.000
Operating system 0.800 Storage
0.600
0.400

t
ip
Warranty 0.200 Display
0.000

cr
Connector Design

us
Battery Security

an
(a) laptop

Mobile phone
m
Developing countries Developed countries

Price
1.000
d

Operating system 0.800 Storage

te

0.600
0.400
Warranty 0.200 Display
ep

0.000
cc

Connector Design

Battery Security
A

(b) mobile phone

34

34
 ACCEPTED MANUSCRIPT

Tablet
Developing countries Developed countries

Price
1.000
Operating system 0.800 Storage
0.600
0.400

t
ip
Warranty 0.200 Display
0.000

cr
Connector Design

us
Battery Security

an
(c) tablet
Figure 6. The level of customer satisfaction in developing countries versus developed countries (a) laptop, (b) mobile phone, and
(c) tablet
m
In Figure 6, features closer to the spider graph perimeter indicate greater happiness, while proximity to the

center represents unhappiness. As shown, sentiment analysis revealed different results across most features
d

except warranty, which saw much higher satisfaction in developed countries attributable to localized service
te

infrastructure. Price drove the most dissatisfaction in both contexts. A detailed discussion of the results will
ep

be presented separately in the following section. Beyond aggregate sentiments, the granular sentence-level

classification gives managers detailed insights into consumer opinions. The comparative analysis enables
cc

tailored strategies to leverage shared perspectives. By mining nuanced and localized sentiments from

unstructured social data, this approach empowers targeted circular economy and sustainability decision-
A

making optimized for consumer viewpoints in each geographic and economic context.

4.3.1. sentiment analysis results on developing and developed countries (common features)

Developing countries have considerably larger populations compared to developed countries. Additionally,

in certain developing nations with relatively improved economic conditions, electronic device adoption

35

35
 ACCEPTED MANUSCRIPT

rates surpass the average rate observed in developed countries (Nnorom and Osibanjo, 2008). Therefore, it

is recommended that more investment be made in developing countries due to the larger population. Based

on implementing the model on laptops, mobile phones, and tablets, the larger the gap between consumer

satisfaction and dissatisfaction between developed and developing countries, the more the proposed model

emphasizes investing in recovering the product (or its module) that caused the dissatisfaction. After

t
collecting the product with less satisfaction, the model suggests moving that product in the supply chain

ip
towards a higher-satisfaction location. This policy will develop a more sustainable supply chain that meets

cr
other requirements, such as ethics while utilizing the SDGs and circular economy requirements. In the three

case studies conducted on laptops, mobile phones, and tablets, the following common points were found:

us
- Price: This feature indirectly affects the supply chain. The recommendations provided in this regard help

an
increase the company's profit. Dissatisfaction is a common aspect of this feature. However, by examining

opinions that consider good infrastructure for product collection and providing new products at a discount
m
in exchange for receiving older products, higher satisfaction with mobile phone prices has been created in

developed countries. It seems that providing new products at a discount in exchange for receiving older
d

products in the first step in developed countries will increase supply chain efficiency and better compliance
te

with SDG requirements. This action can also be recommended in developing countries in the second step.

- Storage: This feature directly affects the supply chain. By identifying the satisfaction gap, shifting
ep

products from dissatisfaction to higher satisfaction will help increase supply chain sustainability and

achieve SDG requirements. The common cause of dissatisfaction with this feature in developing countries
cc

is due to the lack of development of internet infrastructure and use of cloud spaces for data storage.
A

Therefore, it is suggested that device manufacturing companies use storage devices of electronic devices in

developed countries more than in developing countries.

- Security: While indirectly affecting the supply chain, this feature has jointly caused dissatisfaction in

developed countries and relative satisfaction in developing countries. By examining users' opinions and

increasing users' awareness of information leakage risks, their expectations from manufacturing companies

36

36
 ACCEPTED MANUSCRIPT

have risen much higher. In addition, several reports on hacking user information and digital exchanges have

undermined users' trust. Manufacturing companies can partly regain this lost confidence through new

methods and culture-building.

- Battery: This feature can profoundly impact achieving SDGs and reducing waste, directly impacting the

supply chain. Due to elements such as lithium and lead, batteries can have the most negative impact on the

t
environment. There are two reasons for good satisfaction with this feature in developed countries. First, the

ip
development of electric power infrastructure has made electricity available everywhere. Second, the

cr
development of discount proposals for new products in exchange for receiving used products has reduced

the time mobile phone use in the hands of users. The impact of the duration of use of electronic equipment

us
on battery life is significant. Therefore, it is recommended that the satisfaction gap be identified and a flow

an
from developed to developing countries be created. However, due to tablets' prominence and the possibility

of more extended device use, greater relative satisfaction was found in developing countries.
m
- Warranty: As a characteristic, it indirectly impacts the supply chain; however, investing in it can enhance

the direct effects of other characteristics. Due to the lack of implementation of product return acceptance
d

programs and the development of defective product collection centers or post-sales service centers in
te

developing countries, there is considerable dissatisfaction.

ep

Since there are more retail outlets for products in developed countries, overall warranty satisfaction is much

higher in developed countries compared to developing countries.

cc

- Operating system: The operating system is a characteristic that indirectly impacts the supply chain,

primarily influencing product support. Due to increased customer interaction, the proliferation of multiple
A

operating system versions on electronic devices, and the ability to quickly address issues through patches,

there is greater satisfaction with this characteristic among all users. However, the results indicate that this

has led to dissatisfaction in developing countries, such as on laptops with no operating system available

(except for Apple laptops with their specific macOS operating system).

37

37
 ACCEPTED MANUSCRIPT

The noteworthy findings between developed and developing countries are diverse and necessitate an

individual examination on a case-by-case basis, as specified in section 4.3.2.

4.3.2. Results of sentiment analysis on developing and developed countries (non-common features)

Developing countries have a larger population compared to developed countries; however, the penetration

rate of the Internet and the adoption of electronic devices is lower in these developing nations (Myovella et

t
ip
al., 2020). This causes manufacturers of electronic devices to be advised to focus more on developed

countries. The specific results of developing countries are as follows:

cr
- Display and design: While closely related, these two features are separated in use by users. Upgrading

us
and improving these features has a direct impact on the supply chain. Users are relatively satisfied with the

display except for mobile phones. In mobile phones, dissatisfaction has been created due to its small size.
an
Some have also criticized the high sensitivity of the display to impact. A similar situation has been formed

regarding product design.

m

Therefore, in display and design, it is recommended that, given the relative satisfaction in developed
d

countries, the flow of goods in the supply chain should be from developing countries to developed countries.
te

- Connector: This feature has a direct impact on the supply chain. Due to the widespread changes and

differences between Apple and non-Apple phones as well as the registration of different types of ports in
ep

phones, there is relative dissatisfaction in developing countries. This satisfaction gap causes the flow of

mobile phone products to move from developing to developed countries.

cc

This comparative analysis highlights key differences in consumer opinions based on geographic and
A

economic maturity factors. By uncovering dissatisfaction with warranty infrastructure accessibility, results

provide targeted insights to improve circular economy strategies in developing markets. The model enables

optimized decision-making by revealing contrasts in sentiment across regions, empowering electronics

firms to tailor sustainability initiatives to localized consumer viewpoints mined from social data.

38

38
 ACCEPTED MANUSCRIPT

4.4. Deriving industry benchmark for fast and reliable decision-making

Beyond sentiment analysis, distinct benchmarks were derived for developing and developed markets to

empower rapid reverse logistics decisions considering direct reuse-related and indirect consumer

perspective attributes. For indirect features like price and security, posts insights provided suggestions for

future product design or enhancements tailored to each region. For direct physical components, benchmarks

t
optimized circular economy and SDGs for product returns and recycling based on localized consumer

ip
needs. As a novel innovation in this article, the gap between satisfaction and dissatisfaction with a module

cr
is utilized between developing and developed countries to fulfill the enumerated requirements of SDGs and

achieve a sustainable supply chain. By generating tailored benchmarks, this social listening approach

us
enabled targeted decision-making to close the loop in alignment with consumer viewpoints mined from

an
social media data. Companies can leverage positive opinions on reuse-amenable components while

addressing negative sentiments through region-specific benchmarks. This demonstrates the value of
m
granular, location-based social media analytics in providing electronics firms with optimized, sustainable

reverse logistics policies customized for developing versus developed markets.

d
te
ep
cc
A

Figure 7. The visual depiction of the forward flow in the supply chain and the reverse flow in the supply chain (Shahidzadeh and
Shokouhyar, 2022a)

39

39
 ACCEPTED MANUSCRIPT

The happiness spectrum in Figure 8 was mapped to three reverse logistics disposition options to translate

sentiment analysis insights into zero-waste strategies, as shown in Figure 7. Specifically, happiness levels

were aligned with reuse, refurbishment, or remanufacture decisions rather than waste recommendations. It

should be noted that in the proposed model, recycling is recommended as the optimal solution if the

satisfaction score falls below 0.33 in both developed and developing countries. This provided optimized

t
circular economy strategies tailored to consumer opinions mined from social media data. By connecting

ip
granular sentiment modeling to tailored product return policies, companies can leverage positive

cr
perspectives on components amenable to reuse while avoiding waste based on social listening benchmarks.

Overall, this approach enabled data-driven decision optimization for sustainable electronics reverse

us
logistics management grounded in localized consumer viewpoints.

an
m
d

Figure 8. The process of mapping and aligning consumer satisfaction and decision-making tendencies
te

The mapping between consumer happiness levels from sentiment analysis and reverse logistics decisions
ep

enabled an electronics industry benchmark for laptops grounded in circular economy principles and zero-

waste strategies. As shown in Table 6, this data-driven benchmark guides managers in sustainable policies
cc

aligned with mined consumer perspectives. This industry-specific benchmark provides actionable

intelligence to enhance electronics circular supply chain performance by leveraging social listening insights
A

to derive optimized disposition recommendations. The model empowers firms to integrate consumer

opinions into waste minimization initiatives tailored to product components based on granular, localized

social media analytics across developing and developed markets.

Table 6. The derived benchmark, which is a subsection within the electronics industry (a) laptop, (b) mobile phone, and (c) tablet
(a) laptop

40

40
 ACCEPTED MANUSCRIPT

The best this position decision to meet SDGs and circular economy
Features Sentiment analysis
requirements
Developing Developed
countries countries
Moving from developing countries to developed countries to be
Storage 0.046 0.801
refurbished
Local refurbishing without moving out of developed or developing
Display 0.794 0.739
countries' zones
Local refurbishing without moving out of developed or developing
Design 0.722 0.750
countries' zones
Moving from developing countries to developed countries to be
Battery 0.024 0.772
refurbished

t
Connect Moving from developed countries to developing countries to be
0.727 0.285

ip
or refurbished

(b) mobile phone

cr
The best this position decision to meet SDGs and circular economy
Features Sentiment analysis
requirements
Developing Developed

us
countries countries
Moving from developing countries to developed countries to be
Storage 0.160 0.879
refurbished
Moving from developing countries to developed countries to be
Display 0.329 0.718
refurbished
Design

Battery
0.309

0.323
0.592

0.634
an
Moving from developing countries to developed countries to be repaired
and reused
Moving from developing countries to developed countries to be repaired
and reused
m
Connect
0.319 0.315 Local recycling
or

(c) tablet
d

Feature The best this position decision to meet SDGs and circular economy
Sentiment analysis
s requirements
Developing Developed
te

countries countries
Moving from developing countries to developed countries to be
Storage 0.161 0.700
refurbished
ep

Local refurbishing in developing countries and local repair and reuse in

Display 0.743 0.619
developed countries
Design 0.551 0.682 Local repair and reuse in both developing and developed countries
Local repair and reuse in developing countries and local refurbishing in
Battery 0.540 0.829
developed countries
cc

Connect
0.550 0.505 Local repair and reuse in both developed and developing countries
or
A

As shown in Table 6, mapping sentiment scores to disposition decisions enabled comparative laptop, mobile

phone, and tablet benchmarks for developed and developing markets. Instead of focusing solely on local

disposition decisions or providing general recommendations for a specific module within a device, our

contribution lies in addressing the broader and more comprehensive requirements of SDGs and circular

41

41
 ACCEPTED MANUSCRIPT

economy. This is achieved by identifying the flow of materials between developed and developing

countries, from developed to developing countries or vice versa. By uncovering nuanced contrasts in

component-level sentiment, this social listening approach provides tailored circular economy benchmarks

intra-region and extra-region to fill the consumer satisfaction gap. The granular, data-driven benchmarks

empower optimized electronics reverse logistics decision-making to minimize waste based on consumer

t
perspectives. The derived benchmark represents a subset of the broader electronics industry. Focusing

ip
specifically on laptops, mobile phones, and tablets, this research generates targeted reverse logistics and

cr
sustainability insights for a major segment within the electronics sector. While the methodology and

findings may apply to other electronics categories, the laptop, mobile phone, and tablet benchmark offers

us
tailored decision support for this critical device category. Using social media analytics, the laptop, mobile

phone, and tablet-specific benchmark encapsulates optimized SDGs and circular economy strategies based
an
on mined consumer perspectives. This enables electronics manufacturers to leverage the benchmark for

evidence-based and data-driven decision-making around sustainable product returns, reuse, and recycling.
m
Overall, the laptop, mobile phone, and tablet benchmark exemplifies how this research approach can elicit

tailored industry sub-segment insights to inform closed-loop supply chain management within the
d

electronics domain align with SDG9, SDG11, SDG12, SDG13, SDG14, SDG15, and SDG17.
te

5. Practical and theoretical implications of the research findings

ep

This study introduces a circular supply chain model that prioritizes human and societal well-being by

addressing crucial aspects such as raw material requirements, waste management, consumer satisfaction,
cc

and sustainability. The model operates within a circular economy ecosystem and aligns with the SDGs. By
A

comparing consumer behaviors in developing and developed markets, we can observe how these

differentiated behaviors can be incorporated into customized industry benchmarks. Moreover, identifying

a new material stream between these markets will greatly facilitate efforts to establish a sustainable and

human/society-centric supply chain more efficiently than ever before. While developing countries have

pursued circular strategies, benchmark differences were insignificant, indicating room for improvement.

42

42
 ACCEPTED MANUSCRIPT

Due to larger populations and product usage, accelerated adoption of the proposed model is recommended

in developing markets to enhance circular supply chain performance rapidly. This addresses gaps in holistic

circular supply chain execution frameworks integrating consumer perspectives into managerial decision-

making (Goyal et al., 2021). Advanced deep learning enabled optimized text mining to uncover nuanced

consumer insights from social data. The model provides electronics firms with tailored benchmarks to

t
master sustainable reverse logistics management aligned with localized consumer opinions mined from X,

ip
Meta, and Amazon. This empowers competitive circular economy strategies that balance business

cr
objectives with societal needs.

us
6. Concluding remarks

Driven by increased environmental awareness and unsustainable linear consumption models, organizations
an
globally are pursuing circular supply chain strategies by applying SDGs and circular economy requirements

to utilize resources and meet customer needs efficiently (Al-Saidi et al., 2021). However, due to regulatory
m
and financial barriers, CSC implementation remains challenging, especially in developing countries (Cantú

et al., 2021). Social media analytics provides an affordable and expedient method to convert supply chains
d

into consumer-centric circular supply chains, which serves as the initial step towards meeting SDGs
te

requirements (Agrawal et al., 2023). Despite the promise, research on generalizable social media models

for circular supply chains is limited (Lahane et al., 2020). Meeting consumer needs is critical not only for
ep

waste reduction but also for guiding managerial reverse logistics decisions. Reverse logistics

implementation is vital in developing countries, particularly electronics, facing significant sustainability

cc

challenges (Agrawal and Singh, 2019). This research helps address gaps in holistic circular supply chain
A

social media analytics tailored to the electronics industry across geographic contexts. Findings provide

managers with data-driven insights to optimize circular strategies based on mining consumer perspectives

in developing versus developed markets. This study contributes actionable intelligence to advance circular

supply chains globally by leveraging social data for sustainable electronics supply chain management.

In summary, this research makes the following key contributions:

43

43
 ACCEPTED MANUSCRIPT

- The proposed model aligns real-time consumer perspectives mined from social media with manufacturer

profitability and strategic reverse logistics decision-making.

- Deep learning techniques provide data-driven insights on product returns to enable zero-waste strategies

based on reuse, refurbishing, and remanufacturing.

- Tailored industry benchmarks empower circular supply chain management across manufacturing sectors.

t
ip
- By leveraging consumer sentiments from social data, the model addresses the challenge of balancing

cr
consumer expectations with sustainable returns management.

- A novel material stream is identified by conducting sentiment analysis on social media posts, enabling

us
managers to enhance sustainability efforts aligned with SDGs and circular economy principles.

an
Overall, this work uniquely employs advanced social listening techniques to uncover actionable intelligence

from consumer opinions for optimized, localized, and human-centric reverse logistics policies. The data-
m
driven approach guides electronics firms in mastering circular economy supply chain decisions across

developing and developed markets.

d

6.1. Unique contributions

te

This research proposes a holistic deep learning framework integrating consumer perspectives into supply
ep

chain decision-making for circular economy strategies meeting SDG9, SDG11, SDG12, SDG13, SDG14,

SDG15, and SDG17. By extracting insights from social media, the model aligns stakeholder needs across
cc

sustainability dimensions, zero-waste approaches, and reverse logistics theory. An electronics case study

generates tailored laptop, mobile phone, and tablet benchmarks to optimize developing versus developed
A

market returns policies. The comparative analysis demonstrates applicability across sectors for location-

specific circular supply chain management grounded in consumer sentiments. A new material stream is

identified from a level of consumer satisfaction. State-of-the-art natural language processing increases the

precision of data-driven decision optimization to minimize waste (Shahidzadeh and Shokouhyar, 2022a).

44

44
 ACCEPTED MANUSCRIPT

By mapping mined opinions to product disposition options, this approach enables fact-based alignment of

business growth, ecological sustainability, and climate change mitigation align with SDGs. Overall, the

social listening model equips managers and policymakers with actionable intelligence to master circular

supply chains customized to consumer perspectives. This advances competitive, socially responsible

strategies for electronics firms across geographic contexts.

t
The significant contributions of this research are:

ip
- A theoretical and practical social media analytics model to enable circular supply chain decisions aligned

cr
with consumer perspectives and SDGs.

us
- Deep learning techniques provide real-time consumer insights rather than traditional surveys to optimize

reverse logistics.
an
- Industry benchmarks inform circular economy strategies across manufacturing sectors.
m
- Aligning mined opinions with capabilities improves circular supply chain performance.

- Optimized reverse logistics recommendations consider sustainability and consumer needs.

d
te

- The granular analysis of sentiments by product features provides targeted insights.

- Real-time monitoring of consumer viewpoints by model/feature enables agile decisions.

ep

- Minimum inventory, returns, and waste alongside maximum value recovery from returns and profitability.
cc

In summary, this work uniquely leverages AI-driven social analytics to uncover real-time consumer

perspectives that master sustainable, circular supply chain strategies and SDGs tailored to developing and
A

developed markets.

6.2. Potential future directions and limitations of the study

While the proposed circular supply chain model is generalized for mass production industries, future work

could extend it to lower volume contexts with distinct consumer needs. Advanced deep-learning approaches

45

45
 ACCEPTED MANUSCRIPT

require substantial data and computing power for accurate text mining. The integrated feature extraction

methodology relies on expert input, warranting careful implementation. This research analyzes the

framework in electronics using comparative developing versus developed market benchmarks. Additional

sector benchmarks, such as automotive manufacturing, could be derived. The model could also inform the

development of reverse logistics roadmaps to transition “as-is” to optimal “to-be” states based on data-

t
driven insights. In summary, the proposed social media analytics model provides a robust foundation for

ip
unlocking circular economy potential across industries grounded in consumer perspectives. Future work

cr
should enhance deep learning techniques, expand sector contexts, and leverage findings in strategic road-

mapping to master sustainable data-driven supply chain decision-making.

us
Declaration an
Conflict of interest
m

No conflict of interest has to be declared.

d
te

References
Adel, A., 2022. Future of industry 5.0 in society: human-centric solutions, challenges and
ep

prospective research areas. Journal of Cloud Computing 2022 11:1 11, 1–15.
https://doi.org/10.1186/S13677-022-00314-5
Agrawal, R., Majumdar, A., Majumdar, K., Raut, R.D., Narkhede, B.E., 2022. Attaining sustainable
development goals (SDGs) through supply chain practices and business strategies: A systematic
cc

review with bibliometric and network analyses. Bus Strategy Environ 31, 3669–3687.
https://doi.org/10.1002/BSE.3057
Agrawal, S., Agrawal, R., Kumar, A., Luthra, S., Garza-Reyes, J.A., 2023. Can industry 5.0 technologies
A

overcome supply chain disruptions?—a perspective study on pandemics, war, and climate change
issues. Operations Management Research 1–16. https://doi.org/10.1007/S12063-023-00410-
Y/TABLES/5
Agrawal, S., Singh, R.K., 2019. Analyzing disposition decisions for sustainable reverse logistics: Triple
Bottom Line approach. Resour Conserv Recycl. https://doi.org/10.1016/j.resconrec.2019.104448
Ahmadi, S., Shokouhyar, S., Amerioun, M., Salehi Tabrizi, N., 2024. A social media analytics-based
approach to customer-centric reverse logistics management of electronic devices: A case study on
notebooks. Journal of Retailing and Consumer Services 76, 103540.
https://doi.org/10.1016/J.JRETCONSER.2023.103540

46

46
 ACCEPTED MANUSCRIPT

Ahmadi, S., Shokouhyar, S., Shahidzadeh, M.H., Elpiniki Papageorgiou, I., 2020. The bright side of
consumers’ opinions of improving reverse logistics decisions: a social media analytic framework.
International Journal of Logistics Research and Applications 0, 1–34.
https://doi.org/10.1080/13675567.2020.1846693
Alamerew, Y.A., Brissaud, D., 2020. Modelling reverse supply chain through system dynamics for realizing
the transition towards the circular economy: A case study on electric vehicle batteries. J Clean Prod
254, 120025. https://doi.org/10.1016/J.JCLEPRO.2020.120025
Al-Saidi, M., Das, P., Saadaoui, I., 2021. Circular Economy in Basic Supply: Framing the Approach for the
Water and Food Sectors of the Gulf Cooperation Council Countries. Sustain Prod Consum 27, 1273–
1285. https://doi.org/10.1016/J.SPC.2021.03.004

t
Altaf, A., Anwar, M.W., Jamal, M.H., Bajwa, U.I., 2023. Exploiting Linguistic Features for Effective

ip
Sentence-Level Sentiment Analysis in Urdu Language. Multimed Tools Appl 1–27.
https://doi.org/10.1007/S11042-023-15216-0/METRICS

cr
Alturayeif, N., Luqman, H., Ahmed, M., 2023. A systematic review of machine learning techniques for
stance detection and its applications. Neural Comput Appl 35, 5113–5144.
https://doi.org/10.1007/S00521-023-08285-7/FIGURES/4

us
Araújo, M., Pereira, A., Benevenuto, F., 2020. A comparative study of machine translation for multilingual
sentence-level sentiment analysis. Inf Sci (N Y) 512, 1078–1102.
https://doi.org/10.1016/J.INS.2019.10.031
Asghar, Z., Ali, T., Ahmad, I., Tharanidharan, S., Nazar, S.K.A., Kamal, S., 2019. Sentiment Analysis on
an
Automobile Brands Using Twitter Data, in: Communications in Computer and Information Science.
https://doi.org/10.1007/978-981-13-6052-7_7
Avasthi, S., Chauhan, R., Acharjya, D.P., 2021. Processing Large Text Corpus Using N-Gram Language
m
Modeling and Smoothing, in: Lecture Notes in Networks and Systems. https://doi.org/10.1007/978-
981-15-9689-6_3
Ayati, S.M., Shekarian, E., Majava, J., Wæhrens, B.V., 2022. Toward a circular supply chain: Understanding
barriers from the perspective of recovery approaches. J Clean Prod 359, 131775.
d

https://doi.org/10.1016/J.JCLEPRO.2022.131775
Bangsa, A.B., Schlegelmilch, B.B., 2020. Linking sustainable product attributes and consumer decision-
te

making: Insights from a systematic review. J Clean Prod.

https://doi.org/10.1016/j.jclepro.2019.118902
Bensoltane, R., Zaki, T., 2023. Aspect-based sentiment analysis: an overview in the use of Arabic
ep

language. Artif Intell Rev 56, 2325–2363. https://doi.org/10.1007/S10462-022-10215-3/METRICS

Bokhare, A., Kothari, T., 2023. Emotion Detection-Based Video Recommendation System Using Machine
Learning and Deep Learning Framework. SN Comput Sci 4, 1–6. https://doi.org/10.1007/S42979-
022-01619-7/METRICS
cc

Boone, T., Ganeshan, R., Jain, A., Sanders, N.R., 2019. Forecasting sales in the supply chain: Consumer
analytics in the big data era. Int J Forecast 35, 170–180.
https://doi.org/10.1016/J.IJFORECAST.2018.09.003
A

Burke, H., Zhang, A., Wang, J.X., 2023. Integrating product design and supply chain management for a
circular economy. Production Planning & Control 34, 1097–1113.
https://doi.org/10.1080/09537287.2021.1983063
Butt, A.S., Ali, I., Govindan, K., 2023. The role of reverse logistics in a circular economy for achieving
sustainable development goals: a multiple case study of retail firms. Production Planning & Control.
https://doi.org/10.1080/09537287.2023.2197851
Cai, Y.J., Choi, T.M., 2020. A United Nations’ Sustainable Development Goals perspective for sustainable
textile and apparel supply chain management. Transp Res E Logist Transp Rev 141, 102010.
https://doi.org/10.1016/J.TRE.2020.102010

47

47
 ACCEPTED MANUSCRIPT

Cantú, A., Aguiñaga, E., Scheel, C., 2021. Learning from Failure and Success: The Challenges for Circular
Economy Implementation in SMEs in an Emerging Economy. Sustainability 2021, Vol. 13, Page 1529
13, 1529. https://doi.org/10.3390/SU13031529
Chen, Z.S., Zhang, X., Govindan, K., Wang, X.J., Chin, K.S., 2021. Third-party reverse logistics provider
selection: A computational semantic analysis-based multi-perspective multi-attribute decision-
making approach. Expert Syst Appl 166, 114051. https://doi.org/10.1016/J.ESWA.2020.114051
Chiha, R., Ayed, M. Ben, Pereira, C. da C., 2022. A complete framework for aspect-level and sentence-
level sentiment analysis. Applied Intelligence 52, 17845–17863. https://doi.org/10.1007/S10489-
022-03279-9/METRICS
Dantas, T.E.T., de-Souza, E.D., Destro, I.R., Hammes, G., Rodriguez, C.M.T., Soares, S.R., 2021. How the

t
combination of Circular Economy and Industry 4.0 can contribute towards achieving the

ip
Sustainable Development Goals. Sustain Prod Consum 26, 213–227.
https://doi.org/10.1016/J.SPC.2020.10.005

cr
El Wali, M., Golroudbary, S.R., Kraslawski, A., 2021. Circular economy for phosphorus supply chain and its
impact on social sustainable development goals. Science of The Total Environment 777, 146060.
https://doi.org/10.1016/J.SCITOTENV.2021.146060

us
Fatimah, Y.A., Govindan, K., Murniningsih, R., Setiawan, A., 2020. Industry 4.0 based sustainable circular
economy approach for smart waste management system to achieve sustainable development
goals: A case study of Indonesia. J Clean Prod 269, 122263.
https://doi.org/10.1016/J.JCLEPRO.2020.122263
an
Feizollah, A., Ainin, S., Anuar, N.B., Abdullah, N.A.B., Hazim, M., 2019. Halal Products on Twitter: Data
Extraction and Sentiment Analysis Using Stack of Deep Learning Algorithms. IEEE Access 7.
https://doi.org/10.1109/ACCESS.2019.2923275
m
Ferasso, M., Beliaeva, T., Kraus, S., Clauss, T., Ribeiro-Soriano, D., 2020. Circular economy business
models: The state of research and avenues ahead. Bus Strategy Environ 29, 3006–3024.
https://doi.org/10.1002/BSE.2554
Ghobakhloo, M., Iranmanesh, M., Foroughi, B., Babaee Tirkolaee, E., Asadi, S., Amran, A., 2023. Industry
d

5.0 implications for inclusive sustainable manufacturing: An evidence-knowledge-based strategic

roadmap. J Clean Prod 417, 138023. https://doi.org/10.1016/J.JCLEPRO.2023.138023
te

Ghobakhloo, M., Iranmanesh, M., Mubarak, M.F., Mubarik, M., Rejeb, A., Nilashi, M., 2022. Identifying
industry 5.0 contributions to sustainable development: A strategy roadmap for delivering
sustainability values. Sustain Prod Consum 33, 716–737. https://doi.org/10.1016/J.SPC.2022.08.003
ep

González-Sánchez, R., Settembre-Blundo, D., Ferrari, A.M., García-Muiña, F.E., 2020. Main Dimensions in
the Building of the Circular Supply Chain: A Literature Review. Sustainability 2020, Vol. 12, Page
2459 12, 2459. https://doi.org/10.3390/SU12062459
Govindan, K., Bouzon, M., 2018. From a literature review to a multi-perspective framework for reverse
cc

logistics barriers and drivers. J Clean Prod 187, 318–337.

https://doi.org/10.1016/j.jclepro.2018.03.040
Govindan, K., Shaw, M., Majumdar, A., 2021. Social sustainability tensions in multi-tier supply chain: A
A

systematic literature review towards conceptual framework development. J Clean Prod 279,
123075. https://doi.org/10.1016/J.JCLEPRO.2020.123075
Govindan, K., Soleimani, H., 2017. A review of reverse logistics and closed-loop supply chains: a Journal
of Cleaner Production focus. J Clean Prod 142, 371–384.
https://doi.org/10.1016/j.jclepro.2016.03.126
Govindan, K., Soleimani, H., Kannan, D., 2015. Reverse logistics and closed-loop supply chain: A
comprehensive review to explore the future. Eur J Oper Res.
https://doi.org/10.1016/j.ejor.2014.07.012

48

48
 ACCEPTED MANUSCRIPT

Govindan, V., Balakrishnan, V., 2022. A machine learning approach in analysing the effect of hyperboles
using negative sentiment tweets for sarcasm detection. Journal of King Saud University - Computer
and Information Sciences 34, 5110–5120. https://doi.org/10.1016/J.JKSUCI.2022.01.008
Goyal, S., Garg, D., Luthra, S., 2021. Analyzing critical success factors to adopt sustainable consumption
and production linked with circular economy. Environ Dev Sustain. https://doi.org/10.1007/s10668-
021-01655-y
Grover, P., Kar, A.K., Dwivedi, Y.K., 2022. Understanding artificial intelligence adoption in operations
management: insights from the review of academic literature and social media discussions. Ann
Oper Res 308, 177–213. https://doi.org/10.1007/S10479-020-03683-9/METRICS
Huang, F., Yuan, C., Bi, Y., Lu, J., Lu, L., Wang, X., 2022. Multi-granular document-level sentiment topic

t
analysis for online reviews. Applied Intelligence 52, 7723–7733. https://doi.org/10.1007/S10489-

ip
021-02817-1/METRICS
Huang, S., Wang, B., Li, X., Zheng, P., Mourtzis, D., Wang, L., 2022. Industry 5.0 and Society 5.0—

cr
Comparison, complementation and co-evolution. J Manuf Syst 64, 424–428.
https://doi.org/10.1016/J.JMSY.2022.07.010
Ivanov, D., 2023. The Industry 5.0 framework: viability-based integration of the resilience, sustainability,

us
and human-centricity perspectives. Int J Prod Res 61, 1683–1695.
https://doi.org/10.1080/00207543.2022.2118892
Jafari, N., Azarian, M., Yu, H., 2022. Moving from Industry 4.0 to Industry 5.0: What Are the Implications
for Smart Logistics? Logistics 2022, Vol. 6, Page 26 6, 26.
https://doi.org/10.3390/LOGISTICS6020026 an
Julianelli, V., Caiado, R.G.G., Scavarda, L.F., Cruz, S.P. de M.F., 2020. Interplay between reverse logistics
and circular economy: Critical success factors-based taxonomy and framework. Resour Conserv
m
Recycl 158, 104784. https://doi.org/10.1016/j.resconrec.2020.104784
Kamiş, S., Goularas, D., 2019. Evaluation of Deep Learning Techniques in Sentiment Analysis from Twitter
Data, in: Proceedings - 2019 International Conference on Deep Learning and Machine Learning in
Emerging Applications, Deep-ML 2019. https://doi.org/10.1109/Deep-ML.2019.00011
d

Karuppiah, K., Sankaranarayanan, B., Ali, S.M., Jabbour, C.J.C., Bhalaji, R.K.A., 2021. Inhibitors to circular
economy practices in the leather industry using an integrated approach: Implications for
te

sustainable development goals in emerging economies. Sustain Prod Consum 27, 1554–1568.
https://doi.org/10.1016/J.SPC.2021.03.015
Khajuria, A., Atienza, V.A., Chavanich, S., Henning, W., Islam, I., Kral, U., Liu, M., Liu, X., Murthy, I.K.,
ep

Oyedotun, T.D.T., Verma, P., Xu, G., Zeng, X., Li, J., 2022. Accelerating circular economy solutions to
achieve the 2030 agenda for sustainable development goals. Circular Economy 1, 100001.
https://doi.org/10.1016/J.CEC.2022.100001
Khan, S.A.R., Yu, Z., Golpira, H., Sharif, A., Mardani, A., 2021. A state-of-the-art review and meta-analysis
cc

on sustainable supply chain management: Future research directions. J Clean Prod 278, 123357.
https://doi.org/10.1016/J.JCLEPRO.2020.123357
Kumar, M., Raut, R.D., Jagtap, S., Choubey, V.K., 2023. Circular economy adoption challenges in the food
A

supply chain for sustainable development. Bus Strategy Environ 32, 1334–1356.
https://doi.org/10.1002/BSE.3191
Kumar, P., Singh, R.K., Kumar, V., 2021. Managing supply chains for sustainable operations in the era of
industry 4.0 and circular economy: Analysis of barriers. Resour Conserv Recycl 164.
https://doi.org/10.1016/j.resconrec.2020.105215
Lahane, S., Kant, R., 2022. Investigating the sustainable development goals derived due to adoption of
circular economy practices. Waste Management 143, 1–14.
https://doi.org/10.1016/J.WASMAN.2022.02.016

49

49
 ACCEPTED MANUSCRIPT

Lahane, S., Kant, R., Shankar, R., 2020. Circular supply chain management: A state-of-art review and
future opportunities. J Clean Prod. https://doi.org/10.1016/j.jclepro.2020.120859
Lauriola, I., Lavelli, A., Aiolli, F., 2022. An introduction to Deep Learning in Natural Language Processing:
Models, techniques, and tools. Neurocomputing 470, 443–456.
https://doi.org/10.1016/J.NEUCOM.2021.05.103
Lei, Y., Li, Y., 2021. A novel scheme of domain transfer in document-level cross-domain sentiment
classification. https://doi.org/10.1177/01655515211012329 49, 567–581.
https://doi.org/10.1177/01655515211012329
Leng, J., Sha, W., Wang, B., Zheng, P., Zhuang, C., Liu, Q., Wuest, T., Mourtzis, D., Wang, L., 2022. Industry
5.0: Prospect and retrospect. J Manuf Syst 65, 279–295.

t
https://doi.org/10.1016/J.JMSY.2022.09.017

ip
Maddikunta, P.K.R., Pham, Q.V., B, P., Deepa, N., Dev, K., Gadekallu, T.R., Ruby, R., Liyanage, M., 2022.
Industry 5.0: A survey on enabling technologies and potential applications. J Ind Inf Integr 26,

cr
100257. https://doi.org/10.1016/J.JII.2021.100257
Majeed, A., Zhang, Y., Ren, S., Lv, J., Peng, T., Waqar, S., Yin, E., 2021. A big data-driven framework for
sustainable and smart additive manufacturing. Robot Comput Integr Manuf 67, 102026.

us
https://doi.org/10.1016/J.RCIM.2020.102026
Mao, Y., Zhang, Y., Jiao, L., Zhang, H., 2022. Document-Level Sentiment Analysis Using Attention-Based Bi-
Directional Long Short-Term Memory Network and Two-Dimensional Convolutional Neural
Network. Electronics 2022, Vol. 11, Page 1906 11, 1906.
https://doi.org/10.3390/ELECTRONICS11121906 an
Mishra, A., Dutta, P., Jayasankar, S., Jain, P., Mathiyazhagan, K., 2023. A review of reverse logistics and
closed-loop supply chains in the perspective of circular economy. Benchmarking 30, 975–1020.
m
https://doi.org/10.1108/BIJ-11-2021-0669/FULL/XML
Mohammad, S.M., 2022. Ethics Sheet for Automatic Emotion Recognition and Sentiment Analysis.
Computational Linguistics 48, 239–278. https://doi.org/10.1162/COLI_A_00433
Moktadir, M.A., Kumar, A., Ali, S.M., Paul, S.K., Sultana, R., Rezaei, J., 2020a. Critical success factors for a
d

circular economy: Implications for business strategy and the environment. Bus Strategy Environ 29.
https://doi.org/10.1002/bse.2600
te

Moktadir, M.A., Rahman, T., Ali, S.M., Nahar, N., Paul, S.K., 2020b. Examining barriers to reverse logistics
practices in the leather footwear industry. Ann Oper Res 293, 715–746.
https://doi.org/10.1007/S10479-019-03449-Y/TABLES/16
ep

Moraes, R., Valiati, J.F., Gavião Neto, W.P., 2013. Document-level sentiment classification: An empirical
comparison between SVM and ANN. Expert Syst Appl 40, 621–633.
https://doi.org/10.1016/J.ESWA.2012.07.059
Moslehpour, D., Ekowati, R., Qiu, S., Xie, S., Rasool Madni, G., 2023. Impact of Social Media on Young
cc

Generation’s Green Consumption Behavior through Subjective Norms and Perceived Green Value.
Sustainability 2023, Vol. 15, Page 3739 15, 3739. https://doi.org/10.3390/SU15043739
Myovella, G., Karacuka, M., Haucap, J., 2020. Digitalization and economic growth: A comparative analysis
A

of Sub-Saharan Africa and OECD economies. Telecomm Policy 44, 101856.

https://doi.org/10.1016/J.TELPOL.2019.101856
Nag, U., Sharma, S.K., Govindan, K., 2021. Investigating drivers of circular supply chain with product-
service system in automotive firms of an emerging economy. J Clean Prod 319, 128629.
https://doi.org/10.1016/J.JCLEPRO.2021.128629
Nagamanjula, R., Pethalakshmi, A., 2020. A novel framework based on bi-objective optimization and
LAN2FIS for Twitter sentiment analysis. Soc Netw Anal Min 10, 1–16.
https://doi.org/10.1007/S13278-020-00648-5/METRICS

50

50
 ACCEPTED MANUSCRIPT

Nandal, N., Tanwar, R., Pruthi, J., 2020. Machine learning based aspect level sentiment analysis for
Amazon products. Spatial Information Research 28, 601–607. https://doi.org/10.1007/S41324-020-
00320-2/METRICS
Nassif, A.B., Elnagar, A., Shahin, I., Henno, S., 2021. Deep learning for Arabic subjective sentiment
analysis: Challenges and research opportunities. Appl Soft Comput.
https://doi.org/10.1016/j.asoc.2020.106836
Nayal, K., Kumar, S., Raut, R.D., Queiroz, M.M., Priyadarshinee, P., Narkhede, B.E., 2022. Supply chain
firm performance in circular economy and digital era to achieve sustainable development goals.
Bus Strategy Environ 31, 1058–1073. https://doi.org/10.1002/BSE.2935
Neri, A., Cagno, E., Lepri, M., Trianni, A., 2021. A triple bottom line balanced set of key performance

t
indicators to measure the sustainability performance of industrial supply chains. Sustain Prod

ip
Consum 26, 648–691. https://doi.org/10.1016/J.SPC.2020.12.018
Nnorom, I.C., Osibanjo, O., 2008. Electronic waste (e-waste): Material flows and management practices

cr
in Nigeria. Waste Management 28, 1472–1479. https://doi.org/10.1016/J.WASMAN.2007.06.012
Ozkan-Ozen, Y.D., Kazancoglu, Y., Kumar Mangla, S., 2020. SYNCHRONIZED BARRIERS FOR CIRCULAR
SUPPLY CHAINS IN INDUSTRY 3.5/INDUSTRY 4.0 TRANSITION FOR SUSTAINABLE RESOURCE

us
MANAGEMENT. Resour Conserv Recycl 161, 104986.
https://doi.org/10.1016/J.RESCONREC.2020.104986
Paramesha, K., Gururaj, H.L., Nayyar, A., Ravishankar, K.C., 2023. Sentiment analysis on cross-domain
textual data using classical and deep learning approaches. Multimed Tools Appl 82, 30759–30782.
an
https://doi.org/10.1007/S11042-023-14427-9/METRICS
Park, J., Lee, B.K., 2021. An opinion-driven decision-support framework for benchmarking hotel service.
Omega (United Kingdom). https://doi.org/10.1016/j.omega.2021.102415
m
Peacock, D.C., Khan, H.U., 2019. Effectiveness of social media sentiment analysis tools with the support
of emoticon/emoji, in: Advances in Intelligent Systems and Computing.
https://doi.org/10.1007/978-3-030-14070-0_68
Pourmehdi, M., Paydar, M.M., Ghadimi, P., Azadnia, A.H., 2022. Analysis and evaluation of challenges in
d

the integration of Industry 4.0 and sustainable steel reverse logistics network. Comput Ind Eng 163,
107808. https://doi.org/10.1016/J.CIE.2021.107808
te

Ray, P., Chakrabarti, A., 2019. A Mixed approach of Deep Learning method and Rule-Based method to
improve Aspect Level Sentiment Analysis. Applied Computing and Informatics.
https://doi.org/10.1016/j.aci.2019.02.002
ep

Rayhan Ahmed, M., Islam, S., Muzahidul Islam, A.K.M., Shatabda, S., 2023. An ensemble 1D-CNN-LSTM-
GRU model with data augmentation for speech emotion recognition. Expert Syst Appl 218, 119633.
https://doi.org/10.1016/J.ESWA.2023.119633
Shahidzadeh, M.H., Shokouhyar, S., 2022a. Shedding light on the reverse logistics’ decision-making: a
cc

social-media analytics study of the electronics industry in developing vs developed countries.

https://doi.org/10.1080/19397038.2022.2101706 15, 163–178.
https://doi.org/10.1080/19397038.2022.2101706
A

Shahidzadeh, M.H., Shokouhyar, S., 2022b. Toward the closed-loop sustainability development model: a
reverse logistics multi-criteria decision-making analysis. Environment, Development and
Sustainability 2022 1–93. https://doi.org/10.1007/S10668-022-02216-7
Shahidzadeh, M.H., Shokouhyar, S., Javadi, F., Shokoohyar, S., 2022. Unscramble social media power for
waste management: A multilayer deep learning approach. J Clean Prod 377, 134350.
https://doi.org/10.1016/J.JCLEPRO.2022.134350
Shan, S., Peng, J., Wei, Y., 2020. Environmental Sustainability assessment 2.0: The value of social media
data for determining the emotional responses of people to river pollution—A case study of Weibo
(Chinese Twitter). Socioecon Plann Sci. https://doi.org/10.1016/j.seps.2020.100868

51

51
 ACCEPTED MANUSCRIPT

Singh, A., Shukla, N., Mishra, N., 2018. Social media data analytics to improve supply chain management
in food industries. Transp Res E Logist Transp Rev 114, 398–415.
https://doi.org/10.1016/j.tre.2017.05.008
Srinivasu, P.N., Sivasai, J.G., Ijaz, M.F., Bhoi, A.K., Kim, W., Kang, J.J., 2021. Classification of Skin Disease
Using Deep Learning Neural Networks with MobileNet V2 and LSTM. Sensors 2021, Vol. 21, Page
2852 21, 2852. https://doi.org/10.3390/S21082852
Sun, K., Zhang, R., Mensah, S., Mao, Y., Liu, X., 2019. Aspect-Level Sentiment Analysis Via Convolution
over Dependency Tree. EMNLP-IJCNLP 2019 - 2019 Conference on Empirical Methods in Natural
Language Processing and 9th International Joint Conference on Natural Language Processing,
Proceedings of the Conference 5679–5688. https://doi.org/10.18653/V1/D19-1569

t
Tang, H., Tan, S., Cheng, X., 2009. A survey on sentiment detection of reviews. Expert Syst Appl 36,

ip
10760–10773. https://doi.org/10.1016/J.ESWA.2009.02.063
Tseng, M.L., Lim, M.K., Wu, K.J., Peng, W.W., 2019. Improving sustainable supply chain capabilities using

cr
social media in a decision-making model. J Clean Prod.
https://doi.org/10.1016/j.jclepro.2019.04.202
Tseng, T.W.J., Robinson, B.E., Bellemare, M.F., BenYishay, A., Blackman, A., Boucher, T., Childress, M.,

us
Holland, M.B., Kroeger, T., Linkow, B., Diop, M., Naughton, L., Rudel, T., Sanjak, J., Shyamsundar, P.,
Veit, P., Sunderlin, W., Zhang, W., Masuda, Y.J., 2020. Influence of land tenure interventions on
human well-being and environmental outcomes. Nature Sustainability 2020 4:3 4, 242–251.
https://doi.org/10.1038/s41893-020-00648-5
an
Walker, A.M., Opferkuch, K., Roos Lindgreen, E., Simboli, A., Vermeulen, W.J.V., Raggi, A., 2021. Assessing
the social sustainability of circular economy practices: Industry perspectives from Italy and the
Netherlands. Sustain Prod Consum 27, 831–844. https://doi.org/10.1016/J.SPC.2021.01.030
m
William, P., Gade, R., Chaudhari, R.E., Pawar, A.B., Jawale, M.A., 2022. Machine Learning based
Automatic Hate Speech Recognition System. International Conference on Sustainable Computing
and Data Communication Systems, ICSCDS 2022 - Proceedings 315–318.
https://doi.org/10.1109/ICSCDS53736.2022.9760959
d

Wilson, M., Goffnett, S., 2022. Reverse logistics: Understanding end-of-life product management. Bus
Horiz 65, 643–655. https://doi.org/10.1016/J.BUSHOR.2021.10.005
te

Yadav, A., Vishwakarma, D.K., 2020. Sentiment analysis using deep learning architectures: a review. Artif
Intell Rev 53. https://doi.org/10.1007/s10462-019-09794-5
Zarbakhshnia, N., Govindan, K., Kannan, D., Goh, M., 2023. Outsourcing logistics operations in circular
ep

economy towards to sustainable development goals. Bus Strategy Environ 32, 134–162.
https://doi.org/10.1002/BSE.3122
Zhang, Y., Zhang, Z., Miao, D., Wang, J., 2019. Three-way enhanced convolutional neural networks for
sentence-level sentiment classification. Information Sciences 477.
cc

https://doi.org/10.1016/j.ins.2018.10.030
Zhou, J., Huang, J.X., Hu, Q.V., He, L., 2020. SK-GCN: Modeling Syntax and Knowledge via Graph
Convolutional Network for aspect-level sentiment classification. Knowl Based Syst 205, 106292.
A

https://doi.org/10.1016/J.KNOSYS.2020.106292

52

52