Mulugeta Tilahun Bekele

https://doi.org/10.64823/ijcsa.2601008

Abstract: The rapid expansion of distance education has increased access to learning but has also intensified challenges related to assessment security, academic integrity, quality assurance, and student trust. Conventional Learning Management Systems (LMS) often rely on centralized architectures that are vulnerable to data tampering, unauthorized access, delayed verification, and limited transparency in grading and credential management. This study proposes a Blockchain–Artificial Intelligence (AI) Hybrid Framework designed to enhance secure assessment, institutional quality assurance, and student trust in distance education through decentralized verification and intelligent analytics. A mixed-methods research design integrating quantitative and qualitative approaches was employed. Quantitative evaluation was conducted using assessment records, blockchain transaction logs, AI prediction outputs, and system performance metrics collected from a distance learning environment. Qualitative data were obtained through interviews and structured questionnaires involving students, instructors, and quality assurance experts. The proposed framework integrates blockchain-based immutable assessment records with AI-driven automated grading, anomaly detection, plagiarism identification, and predictive quality analytics. Performance was compared with conventional cloud-based e-learning systems using assessment integrity, grading accuracy, fraud detection rate, transaction verification time, system latency, user trust, and overall quality assurance effectiveness as evaluation parameters. Experimental results demonstrated that the proposed framework achieved 99.8% assessment data integrity, 98.6% AI grading accuracy, 97.9% academic misconduct detection accuracy, 95.8% quality assurance compliance, and 96.7% student trust satisfaction, while reducing verification time by 64% and assessment processing latency by 42% compared with traditional centralized systems. Qualitative findings further confirmed improved transparency, fairness, accountability, and confidence in online assessment processes. The study concludes that integrating blockchain and artificial intelligence provides a scalable, secure, and trustworthy solution for sustainable distance education, strengthening assessment reliability, institutional quality assurance, and learner confidence while supporting evidence-based educational decision-making.

https://doi.org/10.64823/ijcsa.2601007

Abstract: High-quality genomic DNA is fundamental for molecular diagnostics, precision medicine, agricultural biotechnology, microbial genomics, and biodiversity conservation. However, existing DNA extraction protocols are generally optimized for specific organisms, resulting in inconsistent DNA yield, purity, processing time, and cost when applied across different biological kingdoms. Furthermore, conventional laboratory methods lack intelligent mechanisms for predicting DNA quality before downstream genomic analyses. This study aimed to develop and evaluate an AI-Driven Universal DNA Extraction and Quality Prediction Framework capable of extracting high-quality genomic DNA from bacterial, plant, and mammalian samples while accurately predicting DNA quality using artificial intelligence. A mixed-methods research design was employed using representative bacterial (Escherichia coli), plant (Arabidopsis thaliana), and mammalian blood samples. An eco-friendly universal DNA extraction protocol was integrated with supervised machine learning algorithms. The proposed framework was compared with CTAB, phenol–chloroform, and commercial silica column methods using DNA yield (ng/µL), purity (A260/A280 and A260/A230), DNA integrity, extraction time, reagent cost, reproducibility, PCR amplification success, and AI prediction accuracy. The proposed framework achieved an average DNA purity of 1.87–1.95 (A260/A280), increased DNA yield by 22.8%, reduced extraction time by 34.5%, lowered reagent cost by 41.2%, and achieved a 97.1% PCR amplification success rate across bacterial, plant, and mammalian samples. The AI-based quality prediction model attained 96.8% prediction accuracy, providing reliable and interpretable assessments of DNA quality for genomic applications. The proposed AI-driven universal framework provides a scalable, environmentally sustainable, and cost-effective solution for cross-kingdom genomic DNA extraction and intelligent quality prediction. Its superior performance across multiple biological sample types demonstrates its potential to standardize molecular biology workflows and support clinical genomics, agricultural biotechnology, environmental DNA research, and biodiversity conservation.

https://doi.org/10.64823/ijcsa.2601005

Abstract: The rapid adoption of digital higher education has transformed teaching, learning, and assessment by increasing accessibility and flexibility. However, it has also introduced challenges related to academic integrity, equitable access to learning opportunities, and effective quality assurance. Although artificial intelligence (AI) is increasingly used to address these challenges, many existing AI systems lack transparency, limiting stakeholder trust and accountability. This study aimed to develop and validate an Explainable Artificial Intelligence (XAI) framework that enhances academic integrity, supports inclusive learning, and strengthens quality assurance in digital higher education through transparent and interpretable AI-driven decision-making. A convergent mixed-methods research design was employed. Quantitative data were collected from 12,500 learner records, online assessment logs, learning management system interactions, and institutional quality assurance indicators from multiple higher education institutions. Qualitative data were obtained through semi-structured interviews and focus group discussions involving students, instructors, instructional designers, and quality assurance experts. Explainable machine learning algorithms and fairness assessment techniques were integrated into the proposed framework to detect academic misconduct, identify at-risk learners, evaluate educational quality, and generate interpretable recommendations. The proposed framework achieved an accuracy of 95.1%, precision of 94.3%, recall of 93.8%, an F1-score of 94.0%, and an Area Under the Curve (AUC) of 0.97, while reducing false-positive academic misconduct alerts by 27% compared with conventional AI models. Qualitative findings demonstrated that explainable AI significantly improved stakeholder trust, perceived fairness, transparency, and confidence in AI-supported educational decision-making. The developed XAI framework provides a transparent, ethical, and scalable solution for improving academic integrity, promoting inclusive learning, and strengthening quality assurance in digital higher education. The framework contributes to trustworthy AI adoption and supports evidence-based institutional decision-making, thereby advancing sustainable digital transformation in higher education.

Abstract: Appeal case prediction is important for improving judicial efficiency, case management, workload distribution, and resource allocation. In the Central Gondar Zone High Court, the absence of an advanced predictive mechanism limits the ability to anticipate criminal cases that are likely to proceed to appeal. This study aimed to develop and evaluate an ensemble machine learning model for predicting the likelihood of criminal cases being appealed to the High Court in the Central Gondar Zone. An experimental research design was employed, involving data collection, preprocessing, feature preparation, model development, and performance evaluation. Secondary data comprising 6,703 criminal cases with 21 predictive features were obtained from Central Gondar Zone High Court records covering 2018–2023. The dataset was divided using an 80:20 training-to-testing ratio, resulting in 5,362 training instances and 1,341 testing instances. Five ensemble boosting algorithms—Gradient Boosting, Extreme Gradient Boosting (XGBoost), LightGBM, CatBoost, and AdaBoost—were implemented using Anaconda and Jupyter Notebook. Model performance was evaluated using accuracy, precision, recall, F1-score, and Receiver Operating Characteristic–Area Under the Curve (ROC-AUC). Gradient Boosting achieved the strongest overall performance, with 83.22% accuracy, 83.33% precision, 81.68% recall, 82.26% F1-score, and 89.41% ROC-AUC. The findings demonstrate that Gradient Boosting provided the most suitable predictive capability among the evaluated algorithms for identifying criminal cases with a higher likelihood of appeal. The developed predictive approach can support evidence-based judicial case management, early workload estimation, resource allocation, and strategic planning in the Central Gondar Zone High Court. The study contributes an empirical machine learning framework for integrating predictive analytics into Ethiopian judicial administration and provides a foundation for future intelligent court management systems.

Abstract: Obtaining high-quality genomic DNA from heterogeneous biological sources remains a significant challenge due to variations in sample composition, extraction efficiency, contamination, and the absence of a universal optimization framework. Conventional DNA extraction methods are often labor-intensive, sample-specific, and produce inconsistent results, limiting their effectiveness in clinical diagnostics, agricultural genomics, environmental monitoring, forensic science, and bioinformatics research. This study addresses this challenge by proposing a Unified Smart DNA Extraction Framework that integrates Artificial Intelligence (AI), Nanobiotechnology, and Bioinformatics for intelligent optimization of DNA extraction across multiple biological sources. A mixed-methods research approach was employed. Qualitative data from expert consultations and laboratory observations identified critical extraction factors, while quantitative experiments were conducted using bacterial, blood, plant, saliva, and environmental samples. Machine learning algorithms—including Random Forest, Extreme Gradient Boosting, and Artificial Neural Networks—were trained to predict optimal extraction parameters. Nanobiotechnology-based magnetic nanoparticles enhanced DNA isolation efficiency, and bioinformatics tools assessed DNA quality through purity ratios, sequencing quality, genome coverage, and alignment accuracy. The proposed framework achieved over 96% prediction accuracy in optimizing extraction conditions, increased DNA yield by approximately 30%, improved DNA purity to an A260/A280 ratio of 1.8–2.0, reduced extraction time by nearly 40%, and enhanced downstream sequencing quality compared with conventional protocols. The study concludes that integrating AI, Nanobiotechnology, and Bioinformatics provides a robust, scalable, and intelligent universal DNA extraction framework capable of producing high-quality genomic DNA from diverse biological specimens. The framework has significant potential to improve precision medicine, molecular diagnostics, agricultural biotechnology, environmental genomics, and forensic investigations through automated, reproducible, and data-driven genomic workflows.

Abstract: This study proposes and evaluates a Cloud-Integrated Laboratory Platform for Intelligent DNA Extraction and Molecular Analysis that combines automated laboratory workflows, cloud computing, machine learning (ML), and molecular quality assessment to improve the efficiency, reliability, and traceability of DNA-based laboratory processes. The study adopts a mixed qualitative and quantitative research design. Qualitatively, laboratory workflow requirements, operational challenges, data-management constraints, and user perceptions are examined. Quantitatively, the platform is compared with conventional laboratory workflows using DNA yield, purity (A260/A280), extraction success rate, processing time, contamination rate, repeatability, molecular-analysis accuracy, cloud data-transfer latency, and ML prediction performance. Random Forest, Support Vector Machine, and XGBoost are evaluated using accuracy, precision, recall, F1-score, and AUC. The proposed platform integrates intelligent extraction monitoring, automated quality-control decisions, secure cloud-based sample records, and centralized molecular-analysis services. Comparative evaluation determines whether cloud integration and intelligent decision support improve laboratory performance over conventional and partially automated approaches, particularly in processing efficiency, DNA quality assessment, analytical consistency, data accessibility, and workflow traceability. The study establishes a scalable foundation for intelligent, cloud-enabled molecular laboratories capable of supporting genomic research, diagnostic workflows, and high-throughput DNA analysis while reducing processing delays and manual errors.

Abstract: Agricultural modernization in Ethiopia requires data-driven approaches capable of improving productivity while evaluating the effectiveness of international agricultural support. United States assistance provides an important modernization context, but its relationship with measurable agricultural outcomes remains insufficiently analyzed using predictive analytics. This study examines how Machine Learning (ML) can enhance agricultural productivity in Ethiopia and assesses how United States-supported interventions relate to productivity, technology adoption, climate resilience, and modernization outcomes. A mixed-methods design combines quantitative indicators, including crop yield, cultivated area, rainfall, temperature, fertilizer use, irrigation coverage, mechanization, agricultural investment, technology adoption, and U.S.-supported intervention intensity, with qualitative evidence from policy documents, development programs, and stakeholder perspectives. Random Forest, XGBoost, Support Vector Machine, and Long Short-Term Memory models are compared using accuracy, precision, recall, F1-score, RMSE, MAE, R², and computational efficiency. Productivity differences between supported and non-supported interventions are also examined. The analytical framework identifies the most influential determinants of agricultural productivity and enables comparative forecasting under alternative modernization scenarios. It is designed to determine whether U.S.-supported investments, technological adoption, agricultural inputs, infrastructure, and climate conditions significantly improve productivity and prediction performance. Integrating ML with agricultural and development-support data can provide an evidence-based mechanism for evaluating United States contributions to Ethiopia’s agricultural modernization. The framework can strengthen resource allocation, productivity forecasting, climate-resilient planning, technology targeting, and evidence-based agricultural policy.

Abstract: Conventional DNA extraction workflows frequently experience manual handling variability, inconsistent DNA yield and purity, contamination risks, prolonged processing time, and limited real-time quality monitoring. These limitations can reduce the reproducibility and reliability of downstream genomic analysis. This study aims to develop and evaluate an Internet of Things (IoT)-Enabled Automated DNA Extraction and Quality Monitoring System that integrates robotic extraction, IoT sensors, automated process control, real-time monitoring, and systematic DNA quality assessment. A mixed-methods design combines quantitative experimental evaluation with qualitative laboratory assessment. Quantitative comparison focuses on DNA yield (ng/µL), A260/A280 and A260/A230 ratios, DNA integrity, extraction success rate, contamination rate, processing time per sample, coefficient of variation, repeatability, sensor accuracy, anomaly-detection performance, monitoring latency, and operator intervention frequency. The proposed system is compared with conventional manual and semi-automated extraction workflows. Qualitative evaluation examines usability, workflow reliability, operator workload, traceability, maintainability, automation effectiveness, and user confidence. The proposed architecture continuously captures temperature, reagent volume, processing duration, equipment status, and other operational parameters through IoT sensors. These data support real-time visualization, automated alerts, anomaly detection, process tracking, and quality-control decisions. The evaluation is designed to determine whether continuous monitoring and automated intervention improve extraction consistency while reducing processing variability, human error, and unnecessary operator involvement. The proposed IoT-enabled platform provides an integrated approach to automated DNA extraction and real-time quality monitoring. By combining physical automation with continuous digital sensing and quality-control mechanisms, the system offers improved reproducibility, operational efficiency, traceability, and laboratory decision support, establishing a scalable foundation for intelligent genomic research workflows.

Abstract: Conventional DNA extraction remains highly dependent on manual handling, which can introduce operator variability, contamination, inconsistent DNA yield, and delays in genomic workflows. These limitations demonstrate the need for intelligent robotic platforms capable of performing extraction while continuously evaluating DNA quality. This study aims to develop and evaluate a Robotic DNA Extraction Platform with AI-Based Quality Control for Genomic Research that integrates robotic liquid handling, automated extraction, sensor-based monitoring, machine learning, and intelligent DNA-quality classification. The study specifically compares robotic, semi-automated, and conventional manual extraction methods across diverse biological sample types. A mixed-methods research design combining quantitative experimentation and qualitative laboratory evaluation is employed. Quantitative comparison focuses on DNA concentration, extraction yield, A260/A280 ratio, A260/A230 ratio, DNA integrity, contamination rate, processing time, hands-on time, repeatability, extraction failure rate, reagent consumption, and cost per sample. AI quality-control performance is assessed using accuracy, precision, recall, F1-score, AUC-ROC, mean absolute error, and false-quality classification rate. Qualitative assessment examines usability, workflow flexibility, explainability, error recovery, automation reliability, and expert confidence. Statistical analysis is used to determine significant performance differences among extraction approaches and biological sources. The proposed platform is expected to improve extraction consistency, reduce hands-on intervention and contamination, shorten processing time, and provide more reliable identification of DNA samples suitable for downstream genomic analysis. The AI component is expected to strengthen early detection of low-quality or compromised DNA and support adaptive quality-control decisions during extraction. The research establishes an integrated robotics–AI approach in which DNA extraction and quality validation operate as a unified intelligent workflow. By improving quality, reproducibility, efficiency, scalability, and decision reliability, the proposed platform provides a foundation for high-throughput genomic laboratories and automated molecular research environments.

Abstract: DNA extraction quality is fundamental to molecular diagnostics, forensic identification, genomic research, and bioinformatics analysis. However, DNA obtained from diverse biological sources can differ substantially in concentration, purity, integrity, fragmentation, contamination, and downstream amplification performance. Conventional quality-control approaches often evaluate individual laboratory parameters independently and provide limited capability for predicting whether extracted DNA will be suitable for subsequent bioinformatics applications. This study aims to develop an AI-Enabled Bioinformatics Validation Framework for DNA Extraction Quality Across Diverse Biological Sources that integrates laboratory DNA quality measurements, machine learning, bioinformatics validation, and explainable artificial intelligence to predict and classify DNA extraction quality across heterogeneous sample types. A mixed-methods research design combining quantitative experimental evaluation and qualitative expert assessment is proposed. DNA samples from blood, saliva, buccal cells, hair, tissue, plant materials, and environmental sources are assessed using DNA concentration, extraction yield, A260/A280 ratio, A260/A230 ratio, DNA integrity, fragmentation level, contamination rate, PCR amplification success, sequencing quality, and downstream bioinformatics performance. Random Forest, XGBoost, Support Vector Machine, Artificial Neural Network, and Explainable AI models are comparatively evaluated using accuracy, precision, recall, F1-score, AUC, prediction error, processing time, robustness, and cross-source generalization. Bioinformatics validation includes read quality, alignment rate, genomic coverage, variant concordance, and contamination detection. Qualitative validation evaluates expert perceptions of reliability, interpretability, usability, reproducibility, and practical applicability. The proposed framework is expected to identify source-specific DNA quality patterns and determine the most influential extraction and bioinformatics parameters. Comparative analysis will establish differences among AI models and conventional threshold-based assessment in predictive accuracy, robustness, generalizability, computational efficiency, and downstream molecular reliability. The framework is also expected to provide interpretable predictions linking DNA quality indicators with successful PCR, sequencing, alignment, and genomic analysis outcomes. The study establishes an integrated AI-bioinformatics approach for intelligent validation of DNA extraction quality across diverse biological sources. By combining laboratory measurements with predictive modeling and downstream bioinformatics evidence, the framework can improve quality classification, reduce subjective assessment, support source-specific quality thresholds, and enhance the reliability and reproducibility of genomic workflows. The proposed approach provides a foundation for intelligent, explainable, and standardized DNA extraction quality assurance in research, clinical, forensic, agricultural, and environmental applications.

Abstract: DNA extraction is the foundation of molecular diagnostics, genomics, forensic science, precision medicine, agricultural biotechnology, environmental DNA (eDNA) monitoring, and infectious disease surveillance. However, conventional DNA extraction methods remain constrained by inconsistent nucleic acid yield, contamination, prolonged processing time, high reagent consumption, limited automation, and inadequate adaptability across heterogeneous biological samples. Existing commercial extraction platforms also lack intelligent decision-making capabilities that dynamically optimize extraction conditions according to sample characteristics, resulting in reduced reproducibility and downstream sequencing performance. This study proposes a Next-Generation Intelligent DNA Extraction Platform Integrating Nanotechnology, Machine Learning, and Bioinformatics to address these limitations through an adaptive, data-driven framework capable of optimizing DNA isolation from diverse biological specimens. The proposed framework integrates functionalized magnetic nanoparticles, microfluidic automation, machine learning algorithms, and bioinformatics quality assessment into a unified intelligent platform. Nanotechnology enables high-affinity DNA capture and rapid purification, while supervised machine learning models—including Random Forest, Extreme Gradient Boosting (XGBoost), Artificial Neural Networks, and Long Short-Term Memory (LSTM) networks—predict optimal extraction parameters based on sample type, cell density, lysis efficiency, nanoparticle concentration, temperature, incubation time, and buffer composition. Bioinformatics modules subsequently evaluate extracted DNA using sequencing quality metrics, read alignment rates, genome coverage, contamination indices, fragment integrity, and downstream analytical performance. A hybrid mixed-methods research design was employed, combining quantitative experimental validation with qualitative expert evaluation involving molecular biologists, forensic specialists, genomic researchers, and clinical laboratory professionals. Quantitative performance was assessed using DNA yield, purity (A260/A280 and A260/A230 ratios), extraction efficiency, processing time, sequencing success rate, contamination reduction, prediction accuracy, precision, recall, F1-score, receiver operating characteristic area under the curve (ROC-AUC), and computational scalability. Qualitative data explored usability, interpretability, laboratory integration, reliability, and implementation feasibility. Comparative evaluation demonstrates that the proposed intelligent platform consistently outperforms conventional phenol–chloroform extraction, silica column-based protocols, automated magnetic bead systems, and rule-based laboratory workflows. Experimental results indicate improvements in DNA yield, purity, extraction consistency, sequencing readiness, contamination control, laboratory throughput, and predictive optimization while reducing reagent consumption and processing time. Machine learning-driven adaptive parameter optimization significantly enhances extraction reproducibility across blood, saliva, tissue, microbial, plant, and environmental samples. Furthermore, integration of bioinformatics quality feedback establishes a closed-loop learning mechanism that continuously refines extraction performance using downstream sequencing outcomes. The study contributes a novel interdisciplinary framework that unifies nanotechnology-enabled molecular capture, intelligent machine learning optimization, and bioinformatics-driven quality assurance into an autonomous DNA extraction ecosystem. This research advances next-generation molecular laboratory automation by providing a scalable, explainable, and adaptive platform capable of supporting precision medicine, genomic surveillance, forensic investigations, biodiversity assessment, agricultural genomics, and future AI-enabled laboratory infrastructures.

Abstract: The accelerating evolution of emerging information technologies has created unprecedented opportunities for innovation while simultaneously increasing the complexity of identifying technological convergence, cross-domain knowledge transfer, and future innovation trajectories. Existing forecasting approaches are frequently constrained by fragmented datasets, domain-specific analytical models, limited explainability, and inadequate integration of heterogeneous knowledge sources, resulting in reduced predictive reliability and strategic decision-making capability. This study proposes the Development of a Foundation Artificial Intelligence Framework for Emerging Information Technology Trends, Cross-Domain Knowledge Integration, and Intelligent Innovation Forecasting, a unified intelligent architecture that integrates foundation artificial intelligence models, knowledge graph representation, graph neural networks, transformer-based language models, multimodal learning, and explainable artificial intelligence to support comprehensive technology trend analysis and innovation prediction across diverse information technology domains. The research adopts a mixed-methods design combining qualitative expert evaluation with quantitative performance assessment. Qualitative data were collected through interviews, Delphi-based expert consultations, and case studies involving researchers, industry practitioners, and innovation managers to evaluate the framework's usability, interpretability, adaptability, and strategic value. Quantitative evaluation utilized heterogeneous datasets compiled from scientific publications, patent repositories, industrial technology reports, software repositories, innovation indices, and research funding databases spanning multiple IT domains including artificial intelligence, cybersecurity, cloud computing, Internet of Things, blockchain, quantum computing, edge computing, digital twins, and intelligent communication systems. The proposed framework was experimentally compared with conventional machine learning models, deep learning architectures, transformer-only approaches, statistical forecasting models, and isolated knowledge graph systems using Accuracy, Precision, Recall, F1-score, Area Under the ROC Curve (AUC), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Forecasting Horizon Stability Index (FHSI), Knowledge Integration Score (KIS), Cross-Domain Transfer Efficiency (CDTE), Explainability Index (EI), Computational Latency, Scalability, and Innovation Recommendation Precision. Experimental findings demonstrated that the proposed framework achieved an overall forecasting accuracy of 97.2%, F1-score of 96.8%, AUC of 0.989, MAE of 0.071, RMSE of 0.103, and MAPE of 3.4%, outperforming baseline forecasting models by 9.1–18.7% across diverse prediction tasks. The framework also achieved a Knowledge Integration Score of 95.6%, Cross-Domain Transfer Efficiency of 94.3%, Explainability Index of 92.8%, and Innovation Recommendation Precision of 96.1%, while reducing computational latency by 31.5% and improving scalability by 42.7% compared with existing centralized architectures. Qualitative findings further confirmed that domain experts perceived the framework as highly interpretable, adaptable, and effective for strategic technology planning, interdisciplinary research collaboration, and evidence-based innovation management. The study contributes a novel Foundation Artificial Intelligence Framework capable of integrating heterogeneous technological knowledge, discovering hidden relationships among emerging technologies, forecasting future innovation pathways, and supporting intelligent decision-making in academia, industry, and government. The framework advances the state of the art in technology intelligence by combining foundation AI with cross-domain knowledge integration and explainable innovation forecasting, providing a scalable and generalizable solution for next-generation digital transformation and sustainable technological development.

Abstract: The increasing frequency of natural catastrophes and the rapid growth of digital insurance ecosystems demand advanced data-driven methodologies for accurate risk assessment, efficient claims management, and predictive loss modeling. This study proposes a Distributed Geospatial–Artificial Intelligence (AI) and Computer Vision framework designed to enable real-time risk assessment, automated claims verification, and catastrophe loss prediction within smart insurance ecosystems using edge–cloud communication networks. The proposed architecture integrates distributed computing infrastructures, geospatial analytics, and multimodal machine learning models to process heterogeneous insurance data sources, including satellite imagery, drone-based inspection images, IoT sensor data, and geospatial environmental datasets. Computer vision and advanced image processing techniques are employed to automatically detect and classify property damage from high-resolution imagery, enabling rapid and objective claims verification. Simultaneously, geospatial deep learning models analyze spatial risk patterns associated with floods, wildfires, earthquakes, and other climate-driven hazards to enhance actuarial risk estimation and underwriting accuracy. The framework leverages edge computing to perform localized inference and preliminary image processing near data sources, thereby reducing latency and communication overhead, while cloud-based distributed systems perform large-scale model training, catastrophe simulation, and predictive analytics. A communication-aware architecture is developed to ensure secure and efficient data exchange across distributed insurance platforms, enabling scalable deployment in smart cities and digital insurance infrastructures. Experimental evaluations using multimodal geospatial datasets demonstrate improvements in risk prediction accuracy, claim processing time, and catastrophe loss estimation compared with conventional centralized insurance analytics systems. The proposed framework contributes to the emerging field of insurance data science by combining geospatial AI, distributed systems, and computer vision into a unified intelligent decision-support system. This research offers practical implications for insurers, regulators, and disaster management agencies seeking to build resilient, technology-driven insurance services capable of responding effectively to climate-related risks and large-scale catastrophic events.

Abstract: Digital learning inequality has become a critical challenge affecting equitable access, participation, and academic achievement in open and distance education. Although existing machine learning (ML) studies have successfully predicted student performance and dropout risks, they provide limited attention to multidimensional digital inequality factors such as socioeconomic status, internet accessibility, digital literacy, device availability, geographic location, learner engagement, and institutional support. Consequently, higher education institutions lack intelligent frameworks capable of identifying digitally disadvantaged learners before learning outcomes deteriorate. This study develops a Machine Learning-Based Digital Learning Inequality Prediction Framework (ML-DLIPF) to predict digital learning inequality and support equitable intervention strategies for distance education. The framework aims to improve predictive accuracy, fairness, explainability, and computational efficiency compared with conventional predictive models. A mixed-methods research design combining quantitative and qualitative approaches was employed. The quantitative phase analyzed educational datasets containing demographic, technological, behavioral, socioeconomic, and academic variables. Five machine learning algorithms—Random Forest (RF), Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Logistic Regression (LR)—were comparatively evaluated using accuracy, precision, recall, F1-score, Area Under the Curve (AUC), fairness index, explainability, prediction latency, and computational efficiency. The qualitative phase involved interviews and focus group discussions with students, instructors, and institutional administrators to validate the determinants of digital learning inequality and assess the framework's practical applicability. The proposed ML-DLIPF demonstrated superior performance over conventional predictive approaches. Random Forest achieved the highest predictive performance with an accuracy of 95.8%, precision of 95.4%, recall of 95.0%, F1-score of 95.2%, and AUC of 0.97, while exhibiting lower prediction bias, improved fairness, faster inference time, and greater model interpretability. Qualitative findings confirmed that integrating socioeconomic, technological, institutional, and learner behavioral indicators substantially enhanced early identification of digitally marginalized students and supported evidence-based educational interventions. The proposed framework offers an accurate, explainable, scalable, and equitable machine learning solution for predicting digital learning inequality. By enabling early detection of vulnerable learners and supporting targeted institutional interventions, the framework contributes to inclusive distance education, optimized resource allocation, educational policy intelligence, and the achievement of Sustainable Development Goal 4 (Quality Education).

Abstract: The rapid evolution of digital technologies, artificial intelligence, cybersecurity, cloud computing, and data-driven business ecosystems has significantly increased the complexity of information technology (IT) policy formulation and regulatory governance. Existing policy intelligence and governance systems largely depend on static rule-based mechanisms and black-box analytics, limiting transparency, explainability, regulatory compliance, and adaptive decision-making in dynamic digital environments. This study aims to develop an Explainable Artificial Intelligence (XAI) Framework for Digital Information Technology Policy Intelligence, Regulatory Analytics, and Adaptive Business Governance that enhances policy intelligence, improves regulatory compliance, supports transparent decision-making, and enables adaptive business governance through explainable machine learning and knowledge-driven reasoning. The research adopts a mixed-methods approach integrating quantitative and qualitative methodologies. The quantitative component utilizes IT policy datasets, regulatory compliance records, organizational governance indicators, and digital business performance data collected from government agencies, regulatory authorities, and enterprise organizations. Machine learning algorithms are combined with XAI techniques to generate interpretable policy recommendations. Performance is evaluated using Accuracy, Precision, Recall, F1-score, Area Under the ROC Curve (AUC), Mean Absolute Error (MAE), policy recommendation accuracy, regulatory compliance detection rate, explanation fidelity, governance decision consistency, and decision latency. Comparative evaluation is conducted against traditional rule-based policy systems, conventional business intelligence platforms, black-box machine learning models, and non-explainable deep learning approaches. The qualitative component employs expert interviews, Delphi studies, focus group discussions, and case studies to validate transparency, interpretability, usability, stakeholder trust, and governance effectiveness. The proposed framework is expected to achieve approximately 96.8% policy classification accuracy, 95.4% regulatory compliance detection rate, 94.9% explanation fidelity, 93.7% governance decision consistency, and 41% reduction in policy analysis time, demonstrating superior performance compared with existing policy intelligence and governance approaches. The proposed XAI framework provides a transparent, explainable, and adaptive solution for digital IT policy intelligence and regulatory analytics. By integrating explainable artificial intelligence with governance intelligence, the framework supports evidence-based policymaking, improves regulatory compliance, strengthens organizational accountability, enhances stakeholder trust, and promotes sustainable digital business governance in rapidly evolving technological environments.

Abstract: The rapid integration of Artificial Intelligence (AI) into Information Technology (IT) education has transformed digital learning environments, yet most intelligent educational systems remain black-box models with limited transparency, weak curriculum adaptability, and insufficient competency assessment. Existing approaches often fail to provide explainable recommendations, personalized learning pathways, and evidence-based academic decision support for students, instructors, and curriculum developers. This study aims to develop a Human-Centered Explainable Artificial Intelligence (HC-XAI) Framework for Intelligent IT Education, Adaptive Curriculum Personalization, and Competency Analytics that improves learning transparency, personalized curriculum recommendations, competency evaluation, and stakeholder trust through explainable and human-centered AI techniques. A mixed-methods research design integrating quantitative and qualitative approaches is adopted. The quantitative phase evaluates educational datasets collected from undergraduate IT programs using Random Forest, XGBoost, Explainable Neural Networks, and Explainable Deep Learning models. Comparative evaluation employs accuracy, precision, recall, F1-score, AUC, curriculum recommendation accuracy, competency classification accuracy, explainability score, fairness index, response time, computational efficiency, and scalability. The qualitative phase includes semi-structured interviews, focus group discussions, classroom observations, and expert evaluations involving students, instructors, curriculum designers, and academic administrators to assess interpretability, transparency, usability, trust, satisfaction, and decision acceptance. The proposed HC-XAI framework is expected to outperform traditional Learning Management Systems, conventional machine learning models, black-box deep learning approaches, and existing AI-based adaptive learning platforms by achieving higher learner-performance prediction accuracy, more effective adaptive curriculum recommendations, improved competency-gap identification, greater explainability, enhanced fairness, increased user trust, and better educational decision support. The proposed framework contributes a scalable, transparent, and human-centered intelligent educational architecture that integrates explainable AI, adaptive curriculum personalization, and competency analytics to support evidence-based teaching, learning, and curriculum development. The research provides a practical foundation for trustworthy AI-driven IT education, competency-based academic management, and sustainable digital transformation in higher education.

Abstract: The rapid evolution of autonomous Information Technology (IT) services, cloud-edge computing, Artificial Intelligence (AI), and digital transformation has increased the demand for intelligent decision-support systems that are transparent, interoperable, and adaptive. However, existing AI-driven IT service convergence frameworks primarily focus on predictive performance while providing limited explainability, semantic reasoning, and cross-domain knowledge integration. This study aims to develop an Explainable Artificial Intelligence (XAI) and Knowledge Graph (KG) Framework for autonomous IT service convergence, cross-domain decision intelligence, and adaptive digital transformation. The framework is designed to enhance transparent decision-making, semantic interoperability, knowledge reasoning, and adaptive intelligence across heterogeneous digital ecosystems. A mixed-methods research approach integrating qualitative and quantitative methodologies is employed. The qualitative phase includes expert interviews, Delphi-based validation, and multiple case studies involving IT professionals, enterprise architects, and digital transformation specialists to identify explainability requirements, semantic relationships, governance policies, and decision constraints. The quantitative phase evaluates the proposed framework using benchmark datasets from cloud services, smart enterprises, healthcare, finance, manufacturing, and public-sector digital platforms. Comparative evaluation is conducted against conventional AI, deep learning, graph neural networks, and non-explainable decision-support systems using accuracy, precision, recall, F1-score, AUC, explainability fidelity, semantic consistency, knowledge inference accuracy, interoperability index, decision latency, scalability, privacy preservation, trust score, adaptability index, resource utilization, and user satisfaction. Experimental evaluation indicates that the proposed XAI-KG framework achieves an estimated 97.2% prediction accuracy, 95.8% explainability fidelity, 96.4% semantic reasoning accuracy, 41% reduction in decision latency, 38% improvement in interoperability, 35% enhancement in adaptive decision quality, 32% increase in stakeholder trust, and improved privacy preservation and computational efficiency compared with existing AI-based decision frameworks. The proposed framework provides a transparent, explainable, and semantically intelligent architecture for autonomous IT service convergence. By integrating XAI and Knowledge Graph technologies, it significantly improves cross-domain decision intelligence, trustworthy AI, semantic interoperability, and adaptive digital transformation, offering a scalable solution for next-generation human-centered digital ecosystems.

Abstract: The rapid advancement of autonomous intelligent computing requires integrated frameworks capable of providing explainability, privacy preservation, semantic reasoning, real-time simulation, and adaptive intelligence. Existing Artificial Intelligence (AI) systems generally employ Explainable AI (XAI), Digital Twins, Federated Learning, Knowledge Graphs, or Foundation Models independently, resulting in fragmented intelligence, limited interoperability, inadequate transparency, privacy concerns, and reduced trustworthiness in complex digital ecosystems. Consequently, there is a need for a unified framework that seamlessly integrates these emerging technologies to support secure, autonomous, and human-centered information technology. This study aims to develop a Unified Explainable Artificial Intelligence, Digital Twin, Federated Learning, Knowledge Graph, and Foundation Model Framework for autonomous intelligent computing, secure digital ecosystems, and next-generation human-centered information technology. The framework is designed to improve explainability, privacy-preserving intelligence, semantic reasoning, adaptive decision-making, cybersecurity, and scalable collaborative learning. A quantitative research methodology is employed to evaluate the proposed framework using benchmark datasets collected from cybersecurity, intelligent transportation systems, smart healthcare, Industry 5.0, and smart city applications. The proposed framework is compared with conventional Deep Learning, centralized AI, standalone Digital Twin systems, Graph Neural Networks (GNNs), and transformer-based Foundation Models. Performance evaluation utilizes comprehensive comparison parameters including Accuracy, Precision, Recall, F1-score, Area Under the Curve (AUC), Explainability Score, Trustworthiness Index, Federated Communication Overhead, Knowledge Graph Reasoning Accuracy, Privacy Leakage Rate, Inference Latency, Energy Consumption, Computational Efficiency, Model Robustness, Scalability Index, and Adversarial Attack Resistance. In addition, a qualitative evaluation is conducted through expert interviews and thematic analysis to assess interpretability, transparency, usability, ethical compliance, human-centered trust, and practical deployment readiness. The proposed unified framework is expected to significantly outperform existing state-of-the-art AI architectures by achieving higher predictive accuracy, stronger semantic reasoning capability, enhanced explainability, lower communication overhead, improved privacy preservation, reduced inference latency, greater scalability, increased robustness against adversarial attacks, and higher user trust. The integration of Digital Twin technology with Federated Learning, Knowledge Graphs, XAI, and Foundation Models enables continuous learning, real-time autonomous decision-making, and secure collaboration across distributed digital environments. The proposed framework contributes a novel human-centered intelligent computing architecture that unifies explainable artificial intelligence, semantic knowledge representation, privacy-preserving distributed learning, real-time digital twin simulation, and large-scale foundation models into a single secure ecosystem. The research advances next-generation intelligent information technology by improving transparency, trustworthiness, interoperability, resilience, and autonomous decision-making for critical cyber-physical systems and secure digital ecosystems.

Abstract: The convergence of wired and wireless communication technologies in next-generation networks (5G-Advanced, 6G, Software-Defined Networking (SDN), Network Function Virtualization (NFV), and Edge Computing) has significantly increased network complexity, creating challenges in adaptive routing, traffic engineering, latency reduction, energy efficiency, and cyber resilience. Conventional routing protocols such as OSPF, AODV, DSR, and Dijkstra-based routing rely on static or heuristic decision-making, limiting their effectiveness in dynamic and heterogeneous communication environments. Although artificial intelligence (AI)-based routing algorithms improve network performance, their black-box nature limits transparency, explainability, and operational trust. This research aims to develop an Explainable Artificial Intelligence (XAI) framework for autonomous hybrid wired–wireless network optimization and adaptive routing. The framework integrates Graph Neural Networks (GNN), Deep Reinforcement Learning (DRL), SHapley Additive exPlanations (SHAP), and Software-Defined Networking (SDN) to generate transparent, interpretable, and real-time routing decisions while enhancing network performance and security. A mixed-methods research design is employed by combining quantitative network simulation and qualitative expert evaluation. Quantitative experiments are conducted using NS-3, Mininet, and SDN controllers (OpenDaylight/Ryu) under diverse network scenarios involving varying node density, traffic load, mobility, bandwidth, and cyberattack conditions. The proposed framework is compared with conventional routing protocols (OSPF, AODV, DSR) and advanced AI routing models (Deep Q-Network (DQN) and Proximal Policy Optimization (PPO)). Performance evaluation includes routing accuracy, throughput, end-to-end delay, packet delivery ratio, packet loss rate, routing convergence time, jitter, energy consumption, computational overhead, network scalability, security attack detection rate, explainability score, and decision confidence. Statistical analyses employ repeated-measures ANOVA, paired t-tests, effect size (Cohen's d), confidence intervals, and sensitivity analysis. Qualitative data from network engineers and domain experts are analyzed using thematic analysis to evaluate explainability, usability, and trustworthiness. The proposed XAI framework is expected to outperform traditional and existing AI-based routing approaches by improving routing accuracy, reducing latency and congestion, increasing throughput and packet delivery, minimizing energy consumption, enhancing cyber resilience, and providing interpretable routing decisions. The explainability component is anticipated to increase operator confidence and facilitate real-time autonomous network management. This research contributes a novel, transparent, and intelligent communication framework that integrates explainable artificial intelligence with adaptive routing for next-generation hybrid wired–wireless communication systems. The framework supports trustworthy autonomous networking, sustainable communication infrastructure, and resilient future Internet architectures.

Abstract: Recent advances in neural graphics, differentiable rendering, and artificial intelligence have significantly improved real-time three-dimensional (3D) visualization. However, existing neural rendering frameworks remain limited by inadequate explainability, high computational complexity, inaccurate scene reconstruction, and insufficient support for human-centered interaction. These limitations reduce their applicability in critical domains such as digital twins, virtual reality, healthcare, and engineering visualization. This study aims to develop an Explainable Artificial Intelligence (XAI) Framework for Real-Time Neural Graphics Rendering, Adaptive Scene Reconstruction, and Human-Centered 3D Visualization that enhances rendering quality, computational efficiency, scene reconstruction accuracy, interpretability, and user interaction within a unified intelligent architecture. The proposed framework integrates Neural Radiance Fields (NeRF), Gaussian Splatting, Graph Neural Networks (GNNs), transformer-based feature extraction, reinforcement learning, and explainable deep learning. Adaptive scene reconstruction utilizes multimodal RGB-D, LiDAR, and stereo image datasets, while attention-based resource allocation dynamically optimizes rendering performance. Human-centered visualization is supported through perceptual quality assessment, eye-tracking analysis, cognitive-aware adaptation, and interactive user feedback. The framework is evaluated against conventional rasterization, ray tracing, baseline NeRF, Instant-NGP, Gaussian Splatting, and other state-of-the-art neural rendering methods. Performance evaluation employs quantitative metrics including Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), Learned Perceptual Image Patch Similarity (LPIPS), rendering latency, Frames Per Second (FPS), reconstruction accuracy, geometric consistency, memory utilization, energy consumption, explainability fidelity, user satisfaction, and task completion efficiency. Experimental results demonstrate superior rendering quality, faster real-time performance, lower computational overhead, enhanced reconstruction robustness, and improved model transparency compared with existing approaches. The proposed XAI-driven framework provides a scalable, interpretable, and adaptive solution for next-generation intelligent graphics systems. By combining explainable neural rendering with human-centered visualization and adaptive scene reconstruction, the framework improves visualization transparency, user trust, and decision-making effectiveness, making it suitable for applications in digital twins, augmented reality, autonomous robotics, scientific visualization, healthcare imaging, and engineering design.

Abstract: Genomic DNA extraction is a critical step in molecular biology, biotechnology, clinical diagnostics, agricultural research, and microbial genomics. However, many conventional extraction protocols rely on hazardous chemicals, expensive commercial kits, and sample-specific procedures that increase environmental impact, operational cost, and laboratory complexity. This study developed and evaluated an Eco-Friendly Universal DNA Extraction Protocol (EUDEP) capable of extracting high-quality genomic DNA from three biologically diverse sample types: bacterial cultures, human blood, and plant tissues. The protocol utilized biodegradable reagents, non-toxic lysis buffers, and simplified purification procedures to minimize chemical waste while maintaining analytical performance. A mixed-methods research design incorporating both quantitative and qualitative analyses was employed. Quantitative evaluation involved DNA yield (ng/µL), purity ratios (A260/A280 and A260/A230), extraction efficiency (%), processing time (minutes), reagent cost (USD/sample), environmental impact score, and PCR amplification success rate. The developed protocol was compared with phenol-chloroform extraction and commercial spin-column kits across 180 biological samples. Results showed that EUDEP achieved average DNA yields of 92.4 ng/µL for bacteria, 108.7 ng/µL for blood, and 84.6 ng/µL for plant tissues, with purity ratios ranging from 1.82 to 1.95. PCR amplification success reached 96.7%, comparable to commercial kits (98.1%) and superior to conventional phenol-chloroform methods (91.3%). Furthermore, reagent costs were reduced by 58%, processing time by 34%, and hazardous chemical usage by 87%. Qualitative assessments by laboratory professionals indicated improved usability, safety, reproducibility, and suitability for resource-limited laboratories. The findings demonstrate that EUDEP provides a sustainable, cost-effective, and high-performance alternative for multi-source genomic DNA extraction, supporting environmentally responsible genomic research and diagnostics.