Digital Repository of Theses

Super Resolution (SR) Generative Adversarial Networks (GANs) for Feature Extraction on Different Datasets

Ajay Mishra — 2026-07-10

There have been tremendous advancements in using Generative Adversarial Network (GAN) models for computer vision. Despite new advancements in GANs for computer vision, super-resolution (SR) is still considered a challenging research topic in computer vision. Super-resolution has various challenges such as ill-posed inverse problem, complexity growth with increase in up-scaling factor and complexity in assessment of output quality. In recent years, there is a surge of interest in super-resolution methods using Generative Adversarial Networks to solve these challenges. Existing research has been mostly focused on few key techniques and single dataset for feature extraction and calculations. So far, there hasn’t been comprehensive research to study various GANs models for existing State-of-the-Art (SOTA) in super-resolution methods for feature extraction on multiple datasets.
The purpose of this research is to study and improve SOTA in super-resolution GANs through extending and modifying various existing architectures and try new architectures for feature extraction to evaluate results on multiple datasets. This paper focuses on VGG19, VGGFace2 and EfficientNet as pre-trained transfer learning backbones for feature extraction within super-resolution GANs on multiple large challenging face images datasets (CelebA, LFW and Simpson). The result will be evaluated using Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index (SSIM) metrics.

Cost of Vacancy (CoV) Reduction Through Employee Upskilling for Swiss Manufacturing in Industry 5.0

Anastasia Rachman — 2026-06-19

Based on the theories of Human Capital, Efficiency Wage, and Skill-Biased Technological Change, this doctoral dissertation examines the strategic role of employee upskilling in reducing turnover and the Cost of Vacancy (CoV) within Swiss manufacturing firms in the Industry 5.0 era. Industry 5.0 represents a new phase of industrial transformation that emphasises human–machine collaboration, resilience, and sustainable growth. Despite the proliferation of workforce development programs, many firms continue to treat them as costs rather than strategic investments, leaving limited empirical evidence on how upskilling impacts key operational outcomes, particularly vacancy cost reduction.
This study addresses this gap by developing and validating a CoV model that quantifies the economic impact of upskilling initiatives. It investigates how investments in technical and cognitive skills influence workforce stability, retention, and organisational competitiveness. Employing a mixed-methods approach, the research integrates quantitative data from 100 employees in a Swiss manufacturing firm with qualitative insights from managers, capturing both measurable outcomes and contextual organisational practices.
The results of this study indicate that structured upskilling programs reduce vacancy-related costs by approximately 35%, primarily through enhanced internal skill coverage, improved employee engagement, and reduced turnover intention. Regression
analysis demonstrates that upskilling accounts for 36% of the variance in Perceived Cost of Vacancy (PCoV), highlighting its substantial role in mitigating both perceived and monetized vacancy impacts. Employees who participated in multi-skilled, Industry 5.0– oriented training exhibited greater adaptability and role flexibility, enabling teams to maintain operational continuity during vacancy periods without relying on immediate external recruitment. Moreover, mediation analysis confirms that engagement partially explains this relationship (indirect effect β = -0.24), indicating that upskilling strengthens motivational and behavioural mechanisms that reduce turnover-related disruptions.
Collectively, these findings underscore the financial and strategic value of investing in employee development, positioning upskilling as a central driver of workforce resilience and cost efficiency in Swiss manufacturing firms operating in Industry 5.0 environments.
This dissertation contributes to the human capital and organisational management literature by providing a validated framework linking upskilling to operational efficiency and resilience. It offers actionable insights for manufacturing leaders seeking to leverage workforce development as a strategic tool to enhance competitiveness and sustain innovation in Industry 5.0. The findings underscore that upskilling is not merely a developmental activity but a critical driver of organisational stability, productivity, and long-term growth.
Keywords: Industry 5.0, upskilling, human capital, Cost of Vacancy (CoV), turnover, Swiss manufacturing, workforce development, employee retention, organisational resilience.

The Complexity Paradox: Why Digital Startups Simplification Efforts Generate Escalating Organisational Demands

Jose Martin Dip — 2026-06-19

Digital-native ventures face a paradox: tools meant to simplify operations generate complexity. Buffer's platform count grew from 12 to 47 in six years, and GitLab's from 8 to 127 over a decade, with platform-related work consuming 28-35% of engineering
capacity that should support product development (Buffer, 2020; GitLab, 2021).
Traditional organisational theory offers little explanation. Lifecycle models suggest that growing firms move toward stability; platform scholarship emphasises how digital tools reduce coordination costs; and institutional theorists argue that conformity reduces uncertainty. Yet digital startups experience precisely the opposite.
This research identifies three mechanisms driving recursive complexity. Digital artefact genealogy explains how each platform adoption spawns 2 to 4 descendant technologies across 5 to 7 generations. When Buffer adopted Salesforce, it required integration middleware, which needed governance protocols, which demanded monitoring tools.
Institutional logic cascading shows how conformity amplifies exposure. GitLab's SOC 2 certification attracted the GDPR authorities, then labour regulators, and finally tax agencies, each with incompatible requirements. Liquid-solid oscillation acceleration describes why restructuring intervals compress from 4 years to quarterly cycles.
Using digital process archaeology to reconstruct organisational evolution through versions, handbooks, code repositories, and timestamped documentation, the study tracked six digital ventures (GitLab, Buffer, Zapier, Notion, Airtable, Figma) over 6-11 years with temporal precision that retrospective interviews cannot provide.
These mechanisms compound synergistically. Complexity is embedded in approximately 40 platforms and 10 institutional domains, regardless of business model. The resulting framework enables founders and investors to diagnose genealogical risks, map institutional exposure, and implement temporal governance, transforming complexity from invisible tax into a manageable dynamic.

Artificial Intelligence Narratives in Earnings Calls and Their Effect on Market Dynamics: A Study of Volume and Volatility in Leading Tech Stocks

Akshay Digamber Kanade — 2026-06-19

The thesis titled "Artificial Intelligence Narratives in Earnings Calls and Their Effect on Market Dynamics: A Study of Volume and Volatility in Leading Tech Stocks," explores the short term financial impact of AI related corporate narratives presented during the quarterly earnings calls by the five leading NASDAQ listed technology companies (Google, Microsoft, Meta, Apple and Nvidia). Earning calls have become a tactical means to influence the expectations of investors using a purposefully crafted language as artificial intelligence has taken stage. This research centers on the impact of specific AI themed narratives ranging from optimism and uncertainty to investment and ethical risk on immediate market behavior, specifically, the fluctuations in trading volume and volatility.
The motivation behind this research comes from the recent recognition that qualitative disclosures now weigh equally, if not more compared to the traditional financial metrics in shaping the investor sentiment. Quantitative metrics such as earnings per share and financial ratios do remain core components of financial analysis. But the language used by executives -especially concerning emerging technologies like AI – can often serve as a signal of strategic intent, competitive positioning, and anticipated risks the business is surrounded by. Yet, existing financial analysis tools such as including sentiment lexicons, AI generated summaries, and predefined models, do not have the thematic specificity and transparency needed to capture the evolving vocabulary and contextual subtleties of AI related communication.
To fill this gap, the research suggests a systematic methodology that involves building a custom AI lexicon, thematically analyzing earnings call transcripts, and empirically connecting these narrative categories to short term market indicators. It uses text mining, winsorization, and regression analysis to evaluate the extent to which AI narratives correlate with abnormal trading activity and volatility changes in the short term. This includes metrics such as the VIX or VXN. The study also provides a replicable and interpretable approach to narrative finance by comparing custom lexicon to generic sentiment model.
This research interlinks multiple disciplines of financial verbiage, AI communication and movements in the financial market. The study focuses on some practical insights of analyzing the impact of AI narratives through the dissection of the language used on earnings calls which would help the investors interpret the cues into trading strategies. This research addresses a new gap in the literature and equips the financial wizards to better assess and respond to the narratives around Artificial Intelligence in the financial aspect.
Keywords: AI Narratives, Earnings Calls, Financial Linguistics, Market Dynamics, Custom Lexicon, NLP

The Role of Emerging Technologies in Business Transformation and Competitive Advantage

Tyler D. Stanley — 2026-06-19

Companies in fast-growing digital sectors have a lot of problems to deal with when it comes to using new technology to get a long-term competitive edge. This dissertation examines the strategic implementation of IoT, blockchain, AI, and cloud computing platforms within organizational contexts. It explores how organizations endeavor to embrace digital transformation and optimize the utilization of technological assets to achieve long-term market positioning objectives. The project is developing a comprehensive implementation model based on extensive mixed-methods research conducted on 487 firms across several sectors to investigate the organizational, environmental, and technological forces influencing the success of digital transformation. The proposed study seeks to utilize Rogers' Diffusion of Innovation Theory (Rogers, 1962) and Venkatesh's Unified Theory of the Acceptance and Use of Technology (Venkatesh, 2003) as foundational frameworks. It will employ structured survey data from C-level executives, interviews with technology leaders, and the analysis of case studies across various industry-related data types. For digital transformation to work, organizations need to have integrated technology plans instead of just deploying one piece of technology at a time This way, they have the technological integration skills and knowledge to adjust and align their strategies to get the most competitive edge. The study constructs the Strategic Digital Ecosystem Integration Model, which offers practitioners viable frameworks of technology implementation decisions, thereby tackling the barriers of implementation such as resource shortages, security exposure, talent shortages, and strategic misalignment with validated risk management and performance measurement protocols demonstrating performance and efficiency in various organizational settings and industry application.

Using Artificial Intelligence to Combat Social Media Scams in Thailand

Voravit Cheepphitaksakul — 2026-06-19

Social media scams have become a major threat in Thailand, targeting individuals and businesses through fraudulent messages, fake advertisements, and impersonation schemes. Traditional scam detection methods struggle to keep up with the rapid evolution of these deceptive tactics. This research proposes an AI-driven approach to detect and prevent social media scams in real-time. We develop a Thai-language scam detection model using WangchanBERTa, a BERT-based NLP model trained on a labeled dataset of scam-related content. The model is deployed as an API using FastAPI via A web dashboard. This AI-powered solution enhances cybersecurity efforts, providing an efficient and scalable defense against social media scams in Thailand

The Impact of Direct-to-Patient Pharmaceutical Digital Marketing on Patient Outcomes in Diabetes Management: A Survey-Based Analysis of Patient Adherence

Shrirang Ajvalia — 2026-06-19

In the US, the pharmaceutical industry is allowed to advertise prescription drugs directly to consumers through both traditional media and digital media outlets. These communications are intended to inform the consumer about their medications, promote safe and proper medication use to engage with their healthcare provider. This intention to assist consumers in making informed decisions and promoting appropriate use of medications creates a controversy on whether or not direct-to-patient (DTP) pharmaceutical marketing is to benefit patients or to simply serve as a means of increasing sales of the companies' products. Based on significant research, a patient benefit includes reminders to adhere to the medication’s protocol which improves their health outcomes.
A cross-sectional survey was conducted to examine the relationship between DTP pharmaceutical social media marketing and adherence to prescribed diabetes medications in people with type 2 diabetes in order to ascertain if there is a positive patient benefit. The survey included 300 adults with type 2 diabetes who reside in the US and excluded those with type 1 diabetes (due to lack of lifestyle modification impact on the disease). In addition to assessing self-reported exposures to pharmaceutical digital marketing and self-reported medication adherence, the survey also assessed trust in pharmaceutical digital advertising, demographics and socioeconomic status, such as age, income and insurance status. Descriptive statistics and cross-tabulations were used to determine patterns in medication adherence related to financial status and insurance status in response to DTP pharmaceutical social media marketing exposures.
Findings from the study indicate that increased levels of exposure to DTP pharmaceutical digital marketing are associated with increased self-reporting of medication adherence. Specifically, people who reported being exposed to DTP pharmaceutical digital marketing at least once per week had higher adherence rates than people who reported being exposed at least once every six months or less frequently.
The results from the study indicate that differences in adherence were present based upon financial and insurance status; specifically, privately insured and higher-income participants exhibited higher adherence rates at all levels of marketing exposure, whereas lower-income and uninsured participants exhibited lower adherence rates at all levels of marketing exposure. Participants who reported lower levels of trust in pharmaceutical digital marketing exhibited weaker adherence patterns regardless of the amount of marketing exposure they reported receiving.
Overall, findings from the current study suggest that DTP pharmaceutical digital marketing can function as an engagement mechanism to support medication adherence in people with diabetes if they have adequate access to care, however, it is less effective in overcoming financial and structural barriers to accessing treatments.

Evaluating Artificial Intelligence’s Impact on Leadership Agility in Sustainability-Driven SMEs in MENA Amid Market Challenges

Stephanie Ayoub — 2026-06-19

The MENA region has witnessed a growing adoption of Artificial Intelligence (AI) technologies among small and medium-sized enterprises (SMEs), particularly those driven by sustainability goals. Over the past decade, AI has become a central enabler of agility, innovation, and data-driven decision-making across industries. However, despite this technological advancement, many sustainability-oriented SMEs continue to face challenges in translating AI adoption into leadership agility and measurable performance improvement. One of the key reasons is the limited contextual understanding of how AI applications, such as automation, predictive analytics, and intelligent agents, interact with leadership behavior in resource-constrained environments. Current academic research remains fragmented, with most studies focusing on large corporations in digitally mature economies, leaving a significant gap in the MENA context. The study adopts a qualitative, capability-focused perspective, examining how AI-enabled tools shape leadership agility and sustainability-oriented decision-making, rather than empirically measuring organizational performance outcomes, which remain uneven and context-dependent across SMEs in the MENA region. This research project will investigate how AI technologies enhance leadership agility in sustainability-driven SMEs, aiming to develop a contextualized framework that links AI adoption to adaptive leadership, organizational resilience, and sustainable performance outcomes.

Exploring Transformational Leadership Skills in Senior Management Across Pakistan, UAE, and the USA

Nabeel K. Haq — 2026-06-19

Findings reveal that while transformational leadership principles are universally valued, their manifestation is deeply shaped by cultural norms, economic structures, and organizational dynamics. In Pakistan, leadership tends to blend traditional hierarchical values with modern participative practices, creating hybrid models that balance authority with empowerment. The UAE demonstrates a multicultural leadership landscape where cultural intelligence and inclusivity are essential for managing diverse teams. In contrast, the USA emphasizes autonomy, innovation, and employee empowerment, aligning with its low power-distance and individualistic culture.
Four overarching themes emerged from the data: Structured Leadership in Dynamic Cultures, Leading with Integrity and Trust, Empowering Employees for Innovation and Growth, and Culture of Human-Centered Leadership. Across all contexts, transformational leaders were found to enhance organizational performance by fostering trust, promoting learning, and aligning individual aspirations with collective goals.
The study contributes to cross-cultural leadership theory by developing a conceptual framework that integrates transformational leadership with cultural adaptability. It provides actionable insights for multinational organizations, policymakers, and educational institutions to design culturally responsive leadership development programs. Ultimately, the research underscores that effective global leadership requires not only vision and influence but also empathy, cultural awareness, and ethical integrity in navigating diverse organizational landscapes.

Optimizing AI and Machine Learning Technologies to Improve Last-Mile Delivery Efficiency in Quick-Commerce

Srinivasan K. P. — 2026-06-19

This study explores the optimization of Artificial Intelligence and Machine Learning (AI/ML) technologies to enhance efficiency in last-mile delivery within Quick- Commerce industries. Consumer expectations for fast and reliable delivery are increasing;
therefore, it is essential to assess the effectiveness of existing AI/ML technologies. This thesis addresses the key objectives such as assessing the performance of current algorithms, identifying biases in the existing tech stacks, and designing a hybrid AI/ML technology framework tailoring to the unique needs of last-mile delivery in Quick-Commerce. To explore algorithmic strengths and weaknesses, this study deploys a mixed-methods research approach which combines qualitative insights and quantitative analysis. The findings from this study specify the opportunities to optimize and address the biases in AI/ML algorithms that can affect performance and efficiency. The proposed hybrid model addresses the challenges within Quick-Commerce and lays the foundation for future innovation in last-mile delivery. This thesis bridges theoretical trends with practical applications and provides insights for industry professionals to optimize last-mile delivery processes and contributes to the academic discussion on AI/ML optimization in last-mile delivery of Quick-Commerce