ABSTRACT. As a consequence of the explosive rise of Internet of Things (IoT), big data processing and analysis became necessary, which further necessitated efficient data offloading to perform tasks effectively and efficiently. As IoT devices come with limited computing, memory and energy, the task/data offloading to edge, fog, cloud and/or cooperative nodes has gained importance as a way to enhance performance. This review provides a brief overview of recent developments in the context of offloading for IoT environment. In this context, basic definitions and architecture designs, along with control processes and classifications are discussed. The major challenges in this field such as scalability, energy efficiency, latency, mobility, resource allocations, load balancing and security, and potential solutions to these problems are also explored through recent developments. Major applications are also highlighted while discussing some emerging trends such as AI-enabled optimization, IoT and 6G network integration, edge intelligence, blockchain-enabled security, serverless offloading, cooperative offloading, and green offloading. Lastly, some future challenges and open issues have been mentioned. The review has taken into account the recent literature available between 2020 and 2026.

ABSTRACT. This study examines how Scrum methodology contributes to IT project success in Yemeni IT companies and investigates whether project complexity explains part of this relationship. A cross-sectional survey was conducted among Scrum project team members, and the data were analyzed using regression-based mediation analysis. The findings show that Scrum methodology has a positive effect on project success. Scrum teams, Scrum events, and Scrum artifacts all contribute to better project outcomes, with Scrum artifacts showing the strongest practical role. The results also indicate that project complexity partially mediates the relationship between Scrum implementation and project success. Structural, technological, and organizational complexity were found to support this mediating role, while dynamic complexity did not show a meaningful mediating effect. These findings suggest that Scrum improves project success not only through direct implementation but also by helping teams manage specific forms of project complexity. The study contributes an actionable Scrum complexity framework for IT managers working in resource-constrained and unstable project environments.

ABSTRACT. Reactive and schedule-based maintenance policies often create unexpected downtime, unnecessary interventions, redundant data collection, and inefficient resource use. This paper proposes a Lean-AI-IoT framework for predictive maintenance (PdM) that links sensor-network data acquisition, artificial intelligence (AI) analytics, Lean waste filtering, decision support, and continuous feedback in one conceptual workflow. The framework specifies how vibration, temperature, current, pressure, and acoustic signals can be processed through edge/cloud analytics; how anomaly detection, classification, and remaining useful life (RUL) estimation can support maintenance decisions; and how Lean criteria can prioritize actions that reduce downtime, cost, response time, and overmaintenance. To address the conceptual nature of the contribution, the paper adds a validation roadmap based on a digital-twin pilot scenario and defines measurable key performance indicators (KPIs), including downtime reduction, maintenance-cost reduction, RUL error, false-alarm rate, response time, asset availability, and resource utilization. The proposed contribution is intended as a theoretical and implementation-oriented basis for future empirical studies, simulation experiments, and industrial pilot deployments.

ABSTRACT. Lean Six Sigma (LSS) is a widely used methodology for reducing waste, variation, and quality losses, while the Internet of Things (IoT) enables real-time sensing, connectivity, and data-driven decision support in smart manufacturing. However, literature still lacks a concise DMAIC-centered view of how IoT capabilities can support LSS improvement activities from problem definition to process control. This paper presents a conceptual, literature-based review and proposes an IoT-enabled DMAIC framework for LSS in smart manufacturing. The framework connects realtime monitoring, intelligent quality assurance, strategic KPI alignment, and human-centric continuous improvement with the Define, Measure, Analyze, Improve, and Control phases. To address the absence of empirical validation, the paper provides an illustrative CNC machining scenario, KPI-based baselinetarget evaluation logic, and a future validation plan involving expert assessment, digital twin simulation, pilot implementation, and statistical comparison. The contribution is conceptual rather than empirically proven, and the reported KPI values are illustrative planning benchmarks for future validation.

ABSTRACT. Abstract—Classifying and processing data, especially images, presents significant challenges and is of paramount importance, particularly in the medical field for early disease diagnosis and disease prevention. Therefore, accurate image classification is crucial. In this work, we present an innovative approach to classifying medical images using deep learning. We apply a deep complex neural network (DCNN) model built from scratch. The model was trained on a medical image dataset. The objective was to accurately identify the main patterns and features present in the images. Using performance metrics, the proposed model demonstrated an accuracy of 98.62, a precision of 98.4, a recall of 99.7, and an F1-score of 99.1 in an independent test set. These results demonstrate the effectiveness of the model in classifying images rather than relying on transferring learning from previous models, which requires configuration. Furthermore, the proposed method opens avenues for further research and applications in image analysis. By working with medical images to detect and diagnose diseases accurately and rapidly, this paper is a valuable contribution to both machine learning and the medical field, opening new horizons in disease detection and broad applications in artificial intelligence.

ABSTRACT. This paper compares the performance of Software-Defined Networking (SDN) utilizing OpenFlow controller with POX -Vs- traditional networking (OSPF-based), through experimental analysis. We conducted experiments using Mininet simulation environment with a standardized four routers ring topology supporting two endpoint hosts. The performance analysis encompassed six relevant metrics: round-trip time (RRT), network convergence time, delay, jitter, packet loss rate, and bandwidth utilization. SDN demonstrated statistically significant performance improvements across all metrics p<0.001 for all comparison. Specifically, RTT improved by 13.5% (95% CI: 10.2-16.8%), network convergence time decreased by 28% (95% CI: 22.4-33.6%), delay reduced by 20.3% (95% CI: 16.7-23.9%), jitter decreased by 19.5% (95% CI: 15.2-23.8%), packet loss diminished by 30.9% (95% CI: 26.1-35.7%), and bandwidth utilization increased by 20.4% under high-load conditions (95% CI:17.2-23.6%).

ABSTRACT. Industrial Internet of Things (IIoT) systems face challenges due to excessive communication overhead and energy consumption from periodic transmission of raw sensor data. This paper introduces a semantic event-driven communication framework for IIoT, utilizing intelligent edge nodes to extract and transmit only significant semantic industrial events. The approach employs lightweight intelligence at the edge for context-aware event abstraction and transmission, and utilizes federated learning to personalize edge models across diverse industrial settings without raw data sharing.

The proposed framework emphasizes communicating essential information rather than all data, achieving a 98.7% reduction in communication bandwidth compared to traditional methods while preserving high detection performance. Theoretical analysis illustrates the reduction in communication overhead and improvement in latency for event-driven semantic transmission. Extensive experiments with real-world industrial time-series datasets confirm the framework's effectiveness.

ABSTRACT. Information Technology Service Management (ITSM) plays an important role in enhancing the quality and efficiency of IT services in educational institutions. Initial observations at University of Aden identified several challenges in IT service documentation and incident management. These issues require a structured assessment using the ITIL framework to determine the current maturity level and identify areas for Improvement. This paper evaluates current ITSM practices at University of Aden and proposes enhancements aligned with the Information Technology Infrastructure Library (ITIL) framework. The study focuses on key ITIL processes, including Incident Management, Service Desk, and Change Management. The results of this research show that implementing ITIL best practices can significantly improve service quality, response time, and user satisfaction.

ABSTRACT. Arabic Dialect Identification (ADI) is a critical component of equitable and socially sustainable Natural Language Processing (NLP) systems in the Arab world. However, existing Arabic language models frequently exhibit systematic biases due to data imbalance across dialects and demographic groups, particularly gender, leading to degraded performance for under-represented communities. This paper introduces S-ArFair, a fairness-aware framework for Arabic Dialect Identification that explicitly integrates Counterfactual Data Augmentation (CDA) with fairness-constrained fine-tuning of transformer-based language models. The proposed framework incorporates a demographic parity regularization term directly into the optimization objective, encouraging equitable predictive behavior across dialectal and gender groups while preserving linguistic discrimination. Experiments conducted on the MADAR Arabic Dialect Corpus demonstrate that S-ArFair achieves a substantial reduction in gender bias (up to 81%) and significantly improves inter-dialect parity, as measured by Demographic Parity Difference and Theil Index, without compromising classification performance. The model attains a Macro-F1 score of 0.86 and maintains high overall accuracy compared to standard fine-tuned AraBERT and MARBERT baselines. These findings highlight that fairness-aware optimization can be effectively operationalized in Arabic NLP, positioning ADI as a fairness-critical task and contributing toward more inclusive and socially sustainable digital transformation in the Arab world.

ABSTRACT. This paper advances symbolic AI for legal drafting by applying constrained answer set programming to encode and test machine-interpretable statutes at authoring time within a Rules-as-Code workflow. The method integrates deductive and abductive reasoning, higher-order defeasibility handling, and automatic natural-language justifications to surface ambiguities, contradictions, and under-specifications in normative text. A testdriven pipeline iteratively encodes provisions, executes suites of compliance scenarios, and proposes minimal textual amendments when explanatory traces reveal unintended effects. In a real statutory case study, the approach identified a material misspecification and, after a targeted amendment, converted prior failures to passes across the test suite, demonstrating the value of explainable reasoning for policy robustness. Implementation highlights include model-returning queries, abducibles for underspecified relations, and a lightweight defeasibility scheme that preserves section-to-code alignment for maintainability. Overall, justified answer set programming is positioned as an AI instrument for verifiable, transparent, and draft-ready legal rules, with noted trade-offs in syntax ergonomics and runtime on complex encodings.

ABSTRACT. The appearance of online shopping and package delivery among university students has shown that there are limitations to the traditional manual handling of packages. InstaStation is a package tracking dashboard that aims to improve the package handling service of Universiti Malaysia Pahang Al-Sultan Abdullah (UMPSA). It is a real-time monitoring and interaction system between students and package handling staff through a mobile-friendly web application. The application was developed using Laravel, React.js, Inertia.js, Tailwind, and MySQL as a database management system. The main functionality of this system involves real-time package tracking, auto-verification through a QR system, safe handling, and reporting. The development of this system followed a methodology called Rapid Application Development (RAD). RAD focuses on developing prototypes of a system to ensure that it has the highest functionality and reliability. The future development of the system will consider the integration of Internet of Things (IoT) technology and artificial intelligence (AI) analytics for the support of predictive analytics in package handling.

ABSTRACT. When digital transformation has embedded in the businesses nowadays, entrepreneurial education needs to improve to develop capabilities that are connected to economy based on artificial intelligence. Comprehending that the usage of artificial intelligence is necessary for companies to face their competitors in this digital era. This study aims to examine the relationship between entrepreneurial education and entrepreneurial capabilities and mediated by AI adoption and entrepreneurial intention. Using structural equation modelling with partial least squares, this study analyses data from 175 participants to test proposed model where entrepreneurial education dimensions, artificial adoption factors, entrepreneurial intention, and entrepreneurial capabilities profiles. Results of this study confirm the enhancement of AI adoption to the effects of entrepreneurial education to entrepreneurial capabilities. Meanwhile, entrepreneurial intention also mediates the relationship between entrepreneurial education and entrepreneurial capabilities. Our proposed model also signifies the effect of entrepreneurial education to both strategic thinking and opportunity recognition when mediated by both entrepreneurial intention and AI adoption. These findings contribute to the entrepreneurial education literatures by displaying digital innovation enhances previous approaches in education.

ABSTRACT. As more students engage with social media, many publications have begun investigating the adoption of social media apps and their motivating factors in educational settings. The study has been conducted among undergraduate students at Al-Idrisi University College, Al-Anbar, Iraq. The data was collected using questionnaires, and 123 valid surveys were analyzed by "The partial least squares-structural equation modeling (PLS-SEM)". The results indicated that the student's experience and self-efficacy in social media significantly impact “Perceived Usefulness” (PU) and “Perceived Ease of Use” (PEOU). Similarly, PU and PEOU significantly impact students' intention to use social media for educational purposes. Lastly, PEOU has a substantial effect on PU. The findings of the present study can help clarify the significant aspects of the TAM for social media within the higher education system. The results have significant implications for higher education system (HES) sector practitioners regarding the use of social media platforms in the education system.

ABSTRACT. Medical data systems are rapidly expanding and highly adaptive. Heart disease remains a major global health threat but can often be prevented in its early stages, reducing severe complications such as dementia and heart failure. Accurate and timely identification of cardiac conditions is critical to improving patient survival. Machine learning techniques have shown strong potential in disease diagnosis compared to traditional approaches. This study applies machine learning and deep learning methods for early heart disease detection. Ten classifiers were implemented to enhance predictive performance. Findings indicate that the Random Forest model achieved superior results on the Heart Dataset, with an accuracy of 0.9878 and an AUC of 0.9994, highlighting its effectiveness for clinical decision support and improved healthcare outcomes.

ABSTRACT. This study examines how Social Media Marketing (SMM) influences sales performance of Food and Beverage in Klang Valley and Kuala Selangor. It focuses on four SMM components: customer feedback, communication, content distribution, and customer relationship management. Using quantitative surveys of 50 SMEs selected through stratified random sampling, the findings show that effective SMM improves brand visibility, and customer engagement. However, challenges such as negative feedback and poor content quality can hinder marketing effectiveness. The study recommends training SMM personnel, and increasing government support for digital transformation. Statistical analysis using SPSS (correlation, t-test, and ANOVA) shows significant relationships (p<0.05) between SMM practices and sales performance. The research aligns with SDGs 8, 9, and 12 but suggests further studies with larger samples.

ABSTRACT. Traditional credit approval processes in many banking institutions remain heavily dependent on expert committees and policy-based judgment, resulting in inconsistencies across heterogeneous economic sectors and limited scalability. Although machine learning has demonstrated strong predictive capability in credit risk assessment, practical adoption remains constrained by explainability, governance compatibility, and institutional trust requirements. This study proposes a Hybrid Sector-Aware Explainable AI framework designed to support rather than replace committee-based lending decisions. The framework combines independent sector-specific models with a unified sector-aware learning architecture to evaluate the trade-off between sector specialization and portfolio-level robustness. Explainability is incorporated through SHAP-based global and local interpretations, while a Human–AI Alignment analysis quantifies consistency between committee decisions and model risk assessments. Experiments conducted on real-world banking credit data across four sectors (Agriculture, Fisheries, Investment, and Consumer lending) show that the proposed Hybrid Sector-Aware Random Forest achieves the highest predictive performance (AUC = 0.763, PR-AUC = 0.457, Accuracy = 0.830) compared with unified and sector-independent baselines. The framework also demonstrates 73.4% Human–AI alignment, while identifying 26.3% of approved loans as potentially high-risk, highlighting latent institutional exposure. Moreover, reduced cross-sector performance dispersion indicates improved robustness under heterogeneous portfolio conditions. The findings suggest that integrating sector awareness, explainability, and governance-oriented evaluation can transform credit risk modeling from purely predictive systems toward transparent, accountable, and institutionally compatible decision-support frameworks suitable for regulated banking environments.

ABSTRACT. This study identifies key determinants influencing the decision to use E-Wallets among Generation Z in Vietnam, focusing on five explanatory variables: Technological Attributes (TEA), Perceived Ease of Use (PEOU), Privacy Security (PSE), Social Effects (SOE), and Individual Innovativeness (INI). Data were collected using a non-probability convenience sampling method, resulting in a valid sample size of N=341, with the dependent variable coded in binary format (Yes/No). The analytical procedure included descriptive statistics, scale reliability testing via Cronbach’s alpha, and hypothesis testing using Binary Logistic Regression. The results indicate that the model possesses high statistical significance and explanatory power (Chi-square = 239.128; p = 0.000; Nagelkerke R² = 0.818), achieving a classification accuracy of 94.4%. All hypotheses (H1–H5) were supported (p < 0.01). Notably, Social Effects (SOE) emerged as the strongest predictor (OR = 7.704), followed by Technological Attributes (TEA) (OR = 5.449), Perceived Ease of Use (PEOU) (OR = 3.656), Privacy Security (PSE) (OR = 3.092), and Individual Innovativeness (INI) (OR = 2.273). The study contributes empirical evidence regarding the binary decision-making process in the Vietnamese context and prioritizes key drivers for Gen Z adoption. Policy implications emphasize expanding network effects (encouraging referrals and ecosystem linkages), enhancing technical quality and UI/UX, and strengthening security standards and data transparency to reinforce user trust.

ABSTRACT. Whole slide image analysis in digital pathology usually relies on the extraction of dense grid-based patches following background removal, which often tends to result in high redundancy and increased computational complexity. Although this approach is highly effective for removing nonhistochemical regions, it does not specifically target morphologically complex regions in the tumor. In this study, we propose an effective morphological preprocessing strategy for adaptive patch selection in melanoma slide images. This approach relies on lightweight morphological feature computation, such as tissue ratio, entropy, texture variation, and edge density, to rank patches and select the top 30% at the slide level before feature extraction. Tested on 66 slides from the CMB-MEL dataset, our method achieved an approximate 70% reduction in the number of extracted patches compared to the conventional grid-based extraction method, while enhancing the quality of morphological information. To further assess representation quality, an unsupervised analysis was conducted using ResNet50 features. The morphology-based selection reduced within-slide variance by approximately 21% compared to random selection, while marginally increasing between-slide variance. Additionally, PCA showed lower dispersion and fewer outliers. The proposed strategy provides an efficient, interpretable, and representationally consistent preprocessing solution, which makes it suitable for large-scale melanoma analysis systems.

ABSTRACT. The rapid deployment of 5G wireless systems requires small-form-factor, high-efficiency antennas operating at millimeter-wave frequencies to enable high data rates and low latency. This paper presents the design and simulation of a high-efficiency microstrip antenna operating at 83 GHz for E-band 5G wireless communication applications. The proposed antenna is fabricated on a Rogers RT 5880 substrate with a dielectric constant of 2.2, a thickness of 0.203 mm, and a loss tangent of 0.0009. The antenna has a compact size of 3.478 × 5.21 × 0.203 mm3 and employs a simple rectangular patch geometry to achieve strong impedance matching and stable radiation performance. Simulation results obtained using CST Microwave Studio demonstrate that the proposed antenna resonates accurately at 83 GHz, with a return loss of −76.48 dB and a voltage standing wave ratio (VSWR) of 1.0003. The -10dB impedance bandwidth is 4.842 GHz, extending from 80.628 GHz to 85.47 GHz. The peak gain is 10.61 dBi, and the radiation efficiency is 99.08%. These results indicate that the proposed antenna is a promising candidate for high-speed millimeter-wave 5G systems, particularly in compact wireless communication devices requiring high gain, wide bandwidth, and excellent efficiency.

ABSTRACT. This research paper describes a communicative architecture based on multi-stream communication using the TCP/TLS protocol with application-layer message fragmentation distributed among multiple independent transport sessions, as opposed to single-stream transmission models. The proposed architecture distributes message fragments across parallel TCP/TLS connections; therefore, each transport stream has its own independent transport behaviour, sequence space, and timing characteristics. An experimental evaluation of the proposed communication architecture was conducted in a controlled localised environment using test applications developed with Python and analysed using a Wireshark traffic analysis tool. The experimental analysis provides evidence of decentralised transmission behaviour with multiple independent TCP/TLS streams and fragmented payload visibility through separate communication channels. The proposed architecture does not provide a replacement for existing cryptographic protections, such as the TLS protocol; however, it introduces an enhancement to telecommunications at the application layer designed to create more decentralised elements of transmission and therefore add complexity to attempts at passive correlation and reconstruction of traffic. This study provides preliminary proof of concept for the distributed application-layer communications architectures and will discuss how these types of architectures may affect the future design of secure communication systems