The study investigates the social capital of smallholder farmers in the Central Gondar Zone, Ethiopia, with a focus on their access to agricultural information. A mixed-methods research approach was employed, and data were collected through 368 household surveys and 36 key informant interviews. The findings suggest that smallholder farmers have low participation in groups, associations, and networks, with bonding social capital being more useful for accessing information. The study recommends designing programmes that enable farmers to form their own groups or associations for sharing agricultural information. The study also suggests the need for holistic information sharing and networks through participatory methods.
The combined use of a Modified Gorilla Troops Optimization (MGTO) algorithm and a U-Net framework for medical imaging was evaluated in this research. The MGTO has been modified to find hyperparameters necessary to stabilize training (learning rate, batch size, and filter size) as well as to improve convergence time. In that sense, the MGTO's ability to increase the degree of balance between exploration and exploitation during hyperparameter search can improve efficiency while navigating through the complex, non-convex optimization surfaces of deep learning models. Results based on the BraTS20 Dataset support that the MGTO optimized learning configuration provided significantly better performance than some standard optimization technique methods (99.37% accuracy; IoU = 0.86; loss = 0.012 etc.). Additionally, the MGTO-generated improvements in convergence characteristics reduced the total training duration approximately 25% compared to more traditional optimization techniques. Overall, the results indicate that systematically optimized hyperparameters contribute directly to improved learning efficiency, segmentation accuracy and computational performance, further demonstrating how effective the proposed MGTO framework is for use in medical imaging applications.
The growing demand for intelligent decision support in Strategic Human Resource Management (SHRM) drives the need for advanced predictive models for employee attrition and performance. This paper presents a hybrid deep learning framework combining TabNet and Long Short-Term Memory (LSTM) networks within a cloud-based architecture. TabNet's attention-based feature selection and LSTM's ability to model temporal relationships enable the model to effectively process both structured tabular data and time-series behavioural patterns. The model is trained and validated using the IBM HR Analytics Employee Attrition & Performance dataset, consisting of 1,470 employee records with 35 demographic and organizational features. Performance evaluation is conducted using metrics such as accuracy, precision, recall, F1-score, ROC-AUC, and regression metrics (R2, RMSE, MAE). Results show that the proposed TabNet-LSTM framework significantly outperforms traditional models, including Artificial Neural Networks, Reinforcement Learning, and Gradient Boosting, in terms of predictive accuracy and reliability. These findings highlight the potential of hybrid cloud-enabled deep learning systems to support proactive employee retention and performance optimization strategies. The original contribution of the study lies in the integration of interpretable tabular learning and sequential modelling within a scalable cloud-based HR analytics framework for enhanced strategic workforce decision-making.
The purpose of this study is to address the critical challenge of insider threats and privacy risks in Human Resource Management (HRM) employee data systems by developing a secure and scalable framework for authorized data access and threat detection. The proposed methodology integrates Artificial Intelligence (AI), Machine Learning (ML), and blockchain technologies, where redundant employee records are removed, sensitive information is extracted using the Fuzzy Basis Cubic Spline Rule (FBCSR), and privacy is preserved through the R & ouml;ssler Attractor K-Anonymity (RAKA) approach. In the unified workflow, RAKA first anonymizes sensitive employee attributes, HSMPC then encrypts and securely stores the protected data on the blockchain for authorized sharing, and finally FBCSR enables real-time insider threat detection from user access activities, providing an end-to-end security solution for HRM applications. Employee data are further partitioned using Fuzzy C-Means (FCM) clustering and securely encrypted via Hybrid Secure Multi-Party Computation (HSMPC) before being stored on a blockchain supported by the PoSABS consensus mechanism and context-aware smart contract access control. Experiments conducted using the Employee dataset and the EventSim dataset demonstrate that the proposed framework achieves an insider threat detection accuracy of 96.19% while effectively preventing unauthorized access with improved scalability and encryption efficiency. The study provides an important implication for HRM organizations and policy makers by recommending blockchain-enabled privacy-preserving access governance to strengthen employee data protection and insider threat monitoring in real-world enterprise environments. The originality of this research lies in the novel integration of FBCSR-based threat detection with HSMPC encryption and scalable PoSABS blockchain consensus, offering a robust contribution toward secure employee data management in modern HRM systems.
Finance, taxation, auditing, and accounting often operate in institutional silos, despite their interdependence in ensuring governance and economic efficiency, particularly in emerging economies such as India. Although digitization initiatives and regulatory reforms have progressed, comprehensive frameworks that technologically and regulatorily integrate these domains remain limited. This fragmentation creates operational inefficiencies, compliance risks, and restricts the full strategic benefits of digital transformation. To address this gap, this study adopts a qualitative approach, using interviews and focus groups, to examine how blockchain, artificial intelligence, and digital tax systems can be systematically integrated into India's regulatory environment. Findings indicate that while digital tax regimes are expected to enhance efficiency, adoption remains constrained by technological barriers, fragmented regulations, and insufficient standardization. Based on these insights, the study proposes a technology-driven integration model incorporating centralized reporting and blockchain-based real-time auditing to improve transparency, data reliability, and compliance. The research contributes by offering a practical and policy-oriented framework tailored to India's financial ecosystem, highlighting the need for coordinated regulatory reform and institutional collaboration to support sustainable economic growth.
This paper aims to develop an efficient fog computing-based brain fog consortium framework for early detection and continuous monitoring of cognitive impairment using IoT-enabled physiological sensing. The proposed methodology integrates EEG, ECG, blood pressure, temperature, and behavioural sensors with fog-layer analytics to distinguish persistent brain fog patterns from temporary cognitive fluctuations. Data are collected from a consortium of over 200 individuals and processed using signal preprocessing, clustering-based behavioural analysis, and lightweight CNN-based cognitive-state classification deployed at the fog layer, while long-term validation is supported by cloud resources. Experimental results demonstrate improved detection accuracy, reduced latency, and reliable identification of risk-prone individuals through real-time fog-level processing. The study highlights the practical implications of enabling remote, low-latency cognitive health monitoring and decision support for clinicians and caregivers, reducing dependence on centralized cloud systems. The original contribution of this research lies in proposing a unified brain fog consortium model that combines physiological sensing, fog computing, and intelligent analytics to support scalable, real-time cognitive assessment and intervention.
Wearable sensors are transforming physical education by enabling real-time monitoring and feedback to improve athletes' performance. These sensors, combined with advanced technologies, bridge the gap between traditional methods and data-driven training models. However, current practices are subjective, leading to discrepancies in measuring physical activity, performance, and progress, with slow feedback for corrective actions. This research proposes the IoT-empowered Proactive-Pulsed Energizing (IoT-PE) model, integrating wearable sensors with cloud computing and mobile apps. The IoT-PE system collects real-time data on heart rate, movement, and calories burned, providing instant feedback through mobile and web applications. It ensures consistent, accurate assessments and enhances engagement through interactive feedback tools. The model personalizes training, allowing educators to tailor sessions based on individual needs. It also provides teachers with actionable insights and data visualizations for informed decision-making. By improving the accuracy of physical activity evaluations, fostering positive athlete interactions, and supporting continuous improvement through real-time feedback, the IoT-PE model enhances the quality of physical education. Furthermore, it encourages long-term fitness practices among athletes, contributing to a data-driven strategy for promoting lifelong health and fitness.
In India, mental health facilities experience high demand, necessitating accurate estimates of patient capacity to support efficient planning, resource allocation, and the development of effective policy. While traditional approaches of forecasting have typically used basic statistical methods, which may not account for complex temporal patterns, seasonality, or external socio-economic variables to predict. To address the shortcomings of these types of forecasts, this paper presents a machine learning framework to forecast patient capacity trends in mental health facility settings in India using past patient data, time series, and external variables. The methodology follows these general steps: data collection from a district-wise mental health patient dataset (2021-2022); subsequent data processing, including data cleaning (excluding missing values) and normalization of numerical features. Lastly, Random Forest Regression was used to model historical patterns and forecast patient capacity, through the process of ensemble averaging, it helped increase robustness and represent non-linear relationships. The proposed machine learning framework demonstrates strong predictive performance by effectively capturing temporal patterns and nonlinear relationships in mental health patient data. Experimental results show low forecasting errors, with a Root Mean Square Error(RMSE)of[RMSE value]and a Mean Absolute Error(MAE)of[MAE value], indicating high prediction reliability. These results confirm the practical strength of the model for demand forecasting, resource planning, and policy support in mental health services across India.
The integration of AI with hexacopter-based UAVs enables high-resolution crop imaging and automated mapping. This study proposes a Hybrid AE-ViT framework, optimized using CoatiOA, for UAV-based crop health classification to distinguish healthy and weedy rice. DJI Mavic 3 Multispectral RGB and multispectral images undergo denoising, augmentation, standardization, and patch preprocessing. High-resolution RGB imagery captures textural and structural crop features, while multispectral bands (green, red, red-edge, and near-infrared) encode spectral and vegetation health information, enhancing model robustness and accuracy under varying field conditions. The Autoencoder performs spectral-spatial representation localization, while the Vision Transformer (ViT) identifies global contextual dependencies, with CoatiOA tuning fusion and network parameters. Testing confirms excellent generalization, achieving 98.4% accuracy, 1.000 precision, 96.7% recall, 0.983 F1-score, 0.9936 AUC, and 0.9956 Average Precision. The confusion matrix shows 572 healthy plants and 563 weedy rice plants correctly classified, with minimal misclassifications and a 0.000 False Positive Rate. Additionally, GPS-based geo-mapping generates spatial health distribution maps, supporting precise field-level decision-making. These results demonstrate the AE-ViT model's reliability, strength, and scalability for UAV-powered crop monitoring in precision agriculture.
In high-throughput financial reconciliation contexts, traditional SQL engines often face significant challenges performing intricate workload-related transactions due to complex multi-dimensional processing. This study proposes a new SQL optimization strategy which is orchestrated by AI systems with an embedded self-learning algorithm that intelligently restructures the execution paths for specific queries in real time, optimizes the workload partitioning, and increases overall reconciliation throughput. This architecture combines machine learning components with a rule-based profiler to discover inefficiencies in the system and re-query based on the given context and surrounding patterns. Extensive experiments on synthetic and real-world financial datasets showed that the system achieved over a 65% reduction in query response latency. Similar improvements in the CPU, memory consumption, the execution resources used, and efficiency absorbing the active transactional load with precision and accuracy while preserving operational integrity of the process were noted. Other benchmarks using PostgreSQL, Oracle, or SAP HANA all confirmed that the adaptability and flexibility of the system were preserved. These findings show that AI-based orchestration drives automated SQL execution systems for modern architecture of financial infrastructure providing dynamic optimization methods for sophisticated revaluation processes.
This paper investigates the benefits and challenges of integrating artificial intelligence (AI) and data analytics in South African digital banking, addressing the research question: What are the benefits and challenges of using AI and data analytics in digital banking in South Africa? Guided by an interpretive qualitative design, the study used semi-structured online interviews with 11 professionals drawn from banks, fintech firms and regulatory institutions, and applied thematic and content analysis to the resulting data. The findings show that AI and data analytics enhance customer experience, improve operational efficiency and support informed decision-making, while key challenges include data privacy and security concerns, high implementation costs, resistance to technological change and legacy-system constraints. A major policy implication is the need for robust regulatory and governance frameworks, combined with targeted skills development and infrastructure investment, to enable ethical and sustainable AI deployment in digital banking. The study offers an original contribution by providing context-specific empirical evidence from an emerging market, extending a literature base that has largely focused on developed economies and informing both policymakers and practitioners on how to optimize AI and data analytics in South African digital banking.
Threats are metamorphosing simultaneously at an alarming speed, demanding instantaneous responsiveness in defence-related surveillance applications. Conventionally, these systems involve monitoring by using rule-based and manual detection, which implies delays in detection and incorrectness at the highest rate of false positives. To mitigate such inconveniences, the present framework is intelligent enough to recognize suspicious human activities such as trespassing, fighting, or weapon handling from live video streams. The issue of rule-based systems causing delayed and incorrect detection in defence surveillance is discussed. A deep learning-based solution is introduced that utilizes the R (2+1) D 3D CNN for temporal recognition and YOLOv8 for real-time object detection, which occurs with exceedingly high precision. The DCSASS and UCF-Crime databases serve as the source of the training datasets which guarantees precision for activity detection in real-time. The 3D model outperforms 2D CNN: 98% accuracy, lower loss. Hence, the incorporation retains movement subtleties that are critical for classifying these actions while keeping an edge on the computation time. Performance validation is conducted on benchmark datasets such as DCSASS and UCF-Crime, which speak volumes about high precision, recall, and prompt response times. A comparison shows how considerably advanced this system is than the existing system. Furthermore, it is scalable for urgent deployment in sensitive operational zones.
Blockchain technology has the potential to eradicate fraud, improve institutional legitimacy, and increase efficiency, security, and transparency. Despite its potential, the adoption of blockchain technology among staff in higher education institutions (HEIs) remains limited, particularly in emerging economies like Oman, due to limited technological infrastructure, insufficient preparedness of institutions, and a lack of skilled employees. Research undertaken in developed countries has identified the advantages of blockchain for credential verification and administrative purposes, but these findings cannot be generalized in Oman due to its distinct context. To fill this research gap, the objective of this study is to identify, explore, and prioritize the critical factors that influence staff adoption of blockchain technology in HEIs in Oman. This study employed the Analytic Hierarchy Process (AHP) to prioritize the factors, and data were collected from 16 staff members employed at universities in Oman. The findings identified Leadership and Governance as the most important factors, highlighting the need for strong top management support, policies, and strategic direction. The second-most important factor, financial management, emphasizes the importance of financial investment, cost-benefit analysis, and financial audit. Technical Infrastructure ranked third, highlighting scalability and security, data security and privacy, and system interoperability. Collaboration and Partnerships ranked fourth, emphasizing collaboration and partnerships with academia, industry, and government, while Human Resources ranked lowest, indicating the need for skilled workforce training and capacity building to effectively adopt blockchain technology. This study offers insights for policymakers, HEIs, administrators, staff, and blockchain technology developers. The originality of this study lies in bridging the gap between academic theory and institutional decision-making by providing the policy-relevant prioritization model for blockchain technology adoption among staff in HEIs, grounded in the Oman context and validated through expert consensus.
The complex relationship between student participation and resource usage is frequently missed by traditional predictors of academic performance, such as past academic records and demographic data. These predictors have limited insight into real-time learning behaviours and often act as lagging indicators compared to behavioural and contextual activities. In contrast, library usage behaviours like study frequency, resource utilization, and reading habits provide dynamic, process-oriented evaluations of student engagement. This study examines the possibility of library usage as a non-traditional predictor by analyzing data on student library usage frequency and the extent to which the student performs the reading, studying, information-seeking, researching, and resource utilization activities to predict student performance. The study uses library usage and likely academic performance data of 1062 students of the five higher learning Institutions collected via online Google Forms. A hybrid method combining k-means clustering and kNN classification algorithms was employed; k-means was explicitly used for clustering processes to perform the initial formulation of 'k'. The clustering identified three distinct clusters: students with predominantly disagreeing responses (students whose performances were not caused by the library usage), those with agreeing responses (students whose performances were caused by the library usage), and a neutral-to-slightly-agreeing (students whose performances showed no relationship with library usage) group. Having obtained a confidence interval of 98.61% and a p-value of 2.2e-16, the results indicate that library usage behaviours can significantly be used to predict student performance. These findings suggest that higher education institutions must reinforce policies encouraging and monitoring active library involvement. The original contribution of this study is the introduction of a machine learning model that uses library usage behaviours as dynamic predictors of student performance, providing a novel perspective over conventional academic and demographic performance indicators.
Several studies have examined how algorithmic systems shape visibility, particularly in how non-Western regions are represented through inherited biases and colonial narratives. Moving beyond a focus on algorithms alone, especially in the African context, this study examines travel-related content on TikTok to understand how platform curation and user engagement work to shape representation. Using content analysis, the study shows that TikTok frequently foregrounds wildlife and nature-focused portrayals of Africa, reinforcing familiar narratives. It also shows that these portrayals are sustained through feedback loops, where users engage with and create content that aligns with what the platform already promotes, further informing algorithmic recommendations. These findings point to the importance of algorithmic transparency, digital literacy, and more careful attention to representation in shaping how Africa is encountered on digital platforms.
