University for Business and Technology
University of Business and Technology in Kosovo: UBT Knowledge Center CollectionsNot a member yet
8054 research outputs found
Sort by
Case Study of Strengthening an Existing Newly Constructed Reinforced Concrete Building in Prishtina
No full textThe strengthening of structures is a key process for their long-term maintenance and use, particularly when buildings face construction deficiencies, increased loads, damage from natural hazards, or new functional demands that exceed their original design capacity. Instead of replacing damaged elements, various strengthening methods can be applied to enhance loadbearing capacity and stability. These methods include the use of advanced fibers, epoxy injection in damaged members, and the addition of new structural elements to improve load distribution. This paper reviews the advantages and limitations of strengthening techniques, emphasizing their cost-effectiveness and applicability depending on materials and objectives. The focus is placed on reinforced concrete, which, despite its durability, is prone to degradation due to environmental exposure, continuous loading, or sudden impacts. A case study of a residential building is presented, where changes in the use of the lower floors required global and local reassessment of the structure. Structural verification was performed using engineering software, identifying elements failing to meet ultimate or serviceability limit states. Strengthening interventions included advanced fiber application and cross-sectional jacketing of members. The study highlights how tailored strengthening strategies ensure safety, extend service life, and adapt existing structures to evolving demands
Hydrogeological Research for Groundwater in the Nashec Spring
No full textThe Sozi spring is located in the northwestern part of the city of Prizren, south of the village of Nashec, and represents one of the most important sources of drinking water for the region. This karst spring originates from the stratified limestone formations of the Pashtrik massif, at an altitude of 312.0 meters above sea level. Traditionally, this spring has played a crucial role in the water supply of the area\u27s inhabitants since the establishment of settlements around it. Considering the karst character and the pronounced dependence on atmospheric precipitation, its capacity is estimated to be approximately 120 l/s, while the usable capacity for irrigation purposes and drinking water supply to local communities is 60-70 l/s. As part of efforts to ensure a sustainable water supply, a series of infrastructural and technical interventions have been carried out over the years. Among the most important ones are the construction of the collection chamber in 1977, the installation of drainage pipes in 2000 and 2010, the rehabilitation of the pumping station in 2020, and investments in water catchment in 2022. Currently, this source provides a stable water supply for 14 villages in the Has region and the western part of Prizren. Technical improvements have contributed to the stabilization of pressure in the distribution network and increased supply efficiency, guaranteeing better coverage and higher quality of drinking water for the residents of the area
Predicting Student Success Using Artificial Intelligence: DataDriven Approaches for Early Intervention
No full textArtificial Intelligence (AI) is increasingly being applied in the field of education to enhance student outcomes and support personalized learning. This research explores the use of AI algorithms to predict student success by analyzing diverse data sources, including academic performance, attendance, engagement metrics, and socio-demographic factors. Machine learning models, such as decision trees, neural networks, and ensemble methods, are employed to identify students who may be at risk of underperforming or require additional support. The study evaluates the accuracy and effectiveness of these models in forecasting academic performance and highlights the potential of AI-driven interventions to provide timely assistance, improve retention rates, and promote equitable learning opportunities. Furthermore, the research discusses challenges related to data privacy, ethical considerations, and model interpretability, emphasizing the importance of responsible AI implementation in educational settings. By leveraging predictive analytics, AI can empower educators to make informed decisions, tailor instructional strategies, and ultimately enhance student achievement
The Analysis and Implementation of the Algorithm for Gheg Albanian to English Translation
No full textMachine translation has demonstrated great success and efficiency in multilingual communication compared to traditional translation methods across high-resource language pairs. However, low-resource languages, such as the Gheg Albanian dialect, remain underexplored. Gheg Albanian plays an important role in Albanian identity and culture, especially in the northern Albanian-speaking regions, yet it lacks parallel data and translation systems with other languages. This paper tests the neural machine translation algorithm to perform one of the first experiments in this pipeline. A Gheg-English dataset containing 1.2k sentences was manually curated and standard pre-processing techniques were applied, including punctuation and whitespace normalization, outlier and duplicate removal, as well as clitics’ and case-sensitive orthographic normalization to distinguish names from simple nouns. Tokenization was performed utilizing the Marian Tokenizer based on Sentence-Piece sub-word tokenization. A pre-trained MarianMT model for Standard Albanian and English translation was evaluated in its base form and in six fine-tuned forms with varying numbers of epochs, learning rates, and data size. The models were evaluated with BLEU and chrF metrics. Initially, the base model scored poorly on the dataset and the finetuned models showed improvement when trained on the full dataset, however their performance was comparable in that they tended to struggle with more complex dialectal sentences. The performance also degraded with a smaller dataset (30%) due to overfitting, indicating the importance of having a significant amount of data available. While the current models are not feasible for a successful practical translating pipeline, the slight improvements with increased data are promising for future development in Gheg-English machine translation
Is Artificial General Intelligence possible?
No full textArtificial General Intelligence (AGI) aims to create a human like machine that can amalgamate, common sense reasoning, problem-solving, and the adaptability to combine knowledge over different domains. The possibility of AGI now is something plausible coming from the recent computational, empirical and theoretical progress. The underlying theories of Solomonoff’s inductive inference, Legg and Hutter’s universal intelligence and Hutter’s maximally intelligent agent put forward a precise mathematical initiation for comprehension of general learning and decision-making. Meantime, more practical progress in deep learning, reinforcement learning and Transfomers brought the emergent capabilities, which were only thought to be human like skills. The introduction of big models like GPT-3 and systems like AlphaGo, revealed that these systems have the capability to learn very complex reasoning and planning in a general domain rather than one task. The merging of scalable architecture, emergent behaviors and theoretical formalism advocate that the idea of AGI is an attainable pathway and not an abstract idea. The two crucial theories, Sutton’s “Bitter Lesson” and Silver et al.’s “Reward is Enough” showed that computation can scale and is able to learn for itself which then followed to better results when competing with human skills. Additionally, the studies done in different human abilities like creativity, curiosity-based learning or even in intrinsic motivation led on the idea that the universal principles of optimization can lead to the self-development and open-ended intelligence. As the rules of scaling continue to improve in one side and with the faster computing and big data, with the bigger models that can work in wider range of tasks in other side the line that separate narrow AI and general AI will start to disappear. The sum of all these great achievements advocate that AGI is not just a hypothetical idea but it is actually another imminent phase in AI revolution
From VPNs to Zero-Trust and Confidential Computing: A Comparative Study of AWS and Azure
No full textCloud computing has transformed modern computing infrastructure, yet security remains a central concern for enterprises and researchers alike. For decades, Virtual Private Networks (VPNs) have been a cornerstone of cloud access, providing encrypted tunnels between users and resources. However, in an era of distributed workforces, sophisticated cyber threats, and multi-tenant cloud environments, VPNs alone are no longer sufficient. This research explores emerging paradigms in cloud security—Zero-Trust Architecture and Confidential Computing—through a comparative lens across Amazon Web Services (AWS) and Microsoft Azure.Zero-Trust Architecture operates under the principle of “never trust, always verify,” enforcing strict identity, policy, and continuous validation across every request. In practice, this shifts organizations away from perimeter-based defenses toward granular access control, leveraging services such as AWS Identity and Access Management (IAM) or Azure Entra ID. Complementing this approach, Confidential Computing secures data-in-use by isolating workloads within Trusted Execution Environments (TEEs), preventing unauthorized access even from cloud providers themselves. AWS Nitro Enclaves and Azure Confidential VMs exemplify this new layer of defense, offering verifiable attestation and hardware-based isolation.By comparing AWS and Azure implementations, this research highlights the technical trade-offs, performance considerations, and strategic implications of adopting Zero-Trust and Confidential Computing. The session argues that while VPNs retain utility, they must evolve into broader, layered strategies. Together, Zero-Trust and Confidential Computing provide a path forward for robust, scalable, and future-ready cloud security. Expected Results. The study is expected to demonstrate that integrating Zero-Trust Architecture with Confidential Computing significantly enhances security resilience beyond traditional VPN models. Comparative evaluation across AWS and Azure should reveal differences in implementation maturity, performance overhead, and usability, offering organizations practical guidance for adoption. Ultimately, the findings aim to provide a framework for enterprises seeking to balance strong data protection, compliance, and operational efficiency in cloud environments
Traınıng and Development of Varıous Open-Source Artıfıcıal Intellıgence Models for Busıness Process Automatıon
No full textAI is reshaping traditional businesses into sustainable operations, so it can be used to optimize resource utilization, automate processes for improved efficiency This study aims to explore the potential of open-source artificial intelligence (AI) models for automating and streamlining business processes. Focusing on a company offering a wide range of digital services (websites, applications, ERP, CRM, marketing), the study develops and trains two different AI models, specific to the needs of the business. The main goal is to create an AI “agent” that can interpret natural language commands and activate pre-configured automations through the n8n platform. These automated operations include launching marketing campaigns, initial website setup, collecting marketing statistics, and automating content creation for blogs and knowledge bases, collaborating with other AI models accessible through APIs. The paper compares the performance and efficiency of the selected open-source models after fine-tuning with real business data, reducing manual and repetitive tasks, and increasing operational efficiency
Continuous Integration and Continuous Deployment (CI/CD) in Modern Software Engineering: A Comparative Study of Jenkins and GitHub Actions
No full textThe rapid evolution of software delivery models has shifted the focus from manual releases toward fully automated pipelines. Continuous Integration and Continuous Deployment (CI/CD) have become central pillars of modern software engineering, enabling faster iterations, early defect detection, and consistent delivery across environments. Among the most widely adopted tools for implementing CI/CD are Jenkins, an open-source automation server known for its extensibility, and GitHub Actions, a cloud-native CI/CD platform tightly integrated with GitHub repositories. This paper provides a comparative analysis of Jenkins and GitHub Actions in the context of DevOps practices, focusing on setup complexity, scalability, customization, integration flexibility, and cost-efficiency. The research highlights how Jenkins excels in flexibility and plugin support for hybrid infrastructures, while GitHub Actions simplifies workflow management through YAML-based declarative pipelines and seamless GitHub integration. The study emphasizes practical implementation scenarios in modern web projects, including Node.js and .NET Core applications.By evaluating both platforms through performance metrics and developer experience, the paper aims to guide software teams in choosing an optimal CI/CD solution aligned with their project scale and organizational maturity.Expected Results.The study is expected to demonstrate that while Jenkins offers superior customization and control for large-scale enterprise environments, GitHub Actions provides a more accessible, integrated, and cloud-native approach suitable for modern agile teams. Comparative results will underline trade-offs in setup effort, scalability, and automation efficiency, offering concrete recommendations for organizations adopting CI/CD pipelines in different contexts
The Use of Artificial Intelligence in Medicine: Benefits and Risks
No full textArtificial Intelligence (AI) is increasingly transforming the field of medicine by enabling rapid and precise diagnosis, personalized therapies, and advanced data management. From interpreting complex medical images to predicting chronic disease risks and tailoring treatments for individual patients, AI demonstrates unprecedented potential to enhance healthcare outcomes. However, alongside these opportunities emerge critical challenges and risks. Biased training datasets may lead to inaccurate diagnoses, while concerns over data privacy and ethical responsibility raise fundamental questions about the limits of automation in healthcare. This paper emphasizes that the integration of AI should not aim to replace physicians but rather to support them, ensuring that technology augments rather than diminishes the human dimension of medicine. Continuous expert oversight, inclusive data representation, and transparent use of patient information are essential to safeguard both accuracy and trust. Ultimately, AI in medicine must be guided not only by technical innovation but also by ethical responsibility, fostering a future where technology strengthens human care without compromising compassion and fairness
Comparative Analysis of Monolithic and Microservices Architectures in Web Development
No full textThis study examines the comparison between monolithic and microservices architectures in web application development, focusing on their impact on performance, scalability, maintainability, and complexity management. Through a review of the literature and practical analysis of real-world case studies, Netflix (representing monolithic architecture) and Etsy (representing microservices architecture) were evaluated using key performance and technical efficiency metrics, employing tools such as Google Lighthouse and HAR file analysis. Tests were conducted on both desktop and mobile platforms to assess architectural influence under varying conditions. Findings indicate that monolithic architecture provides stable performance and centralized control, particularly on desktop platforms, while microservices offer flexibility and modularity but require careful optimization to avoid unnecessary overhead, especially on mobile devices. The study emphasizes that the choice of architecture should be context-dependent, considering application requirements, scalability needs, and available technical resources, with hybrid approaches potentially offering the best balance in many scenarios