Many suggested projects carried out using a cryptography key are meant to guarantee security and privacy of Health Information Systems (HIS). Still, HIS's weakest flaw depending on these cryptographic techniques is key management and resource consumption in computers. This work presents an experimental comparison of asymmetric and symmetric cryptography keys; specifically, RSA (asymmetric) and AES, (symmetric), to ascertain effectiveness and efficiency with respect to the level of consumption of computer resources viz-a-viz Processing Time, computing memory, and CPU consumption. Using the anonymised dataset at https://wiki.openmrs.org/display/RES/Demo+Data , which has been ad-judged to fit for study, was undertaken. The results reveal that throughout the experiment, HIS data increases the RSA and AES both in processing time although RSA indicates higher value compare to AES in all the HIS input. In the same vein, memory usage for the RSA and AES rises as HIS data rises across the trial. For processor consumption, while RSA consistently increased, AES increased only twice, the first one was when HIS input increased from 500 to 1000, AES increased from 0 to 0.01 respectively. Also, HIS input increased from 4500 to 5000, AES increased from 0.01 to 0.03 respectively.

The Internet of Things (IoT) technologies, such as wearable sensors and smart devices, may continuously monitor vital signs, track symptoms, and collect data in healthcare. Complex systems that identify and monitor COVID-19 early can be developed using IoT. Finding a scalable IoT strategy for reliable pandemic dataset collection of COVID-19 suspicious datasets from diverse healthcare centers is still a baby. The proposed framework has three layers: 1) IoT-DL-based pandemic data gathering layer, which is a sensor-based X-ray machine (simulated); 2) network layer with Internet gateway for signal transmission and cloud for COVID-19 data temporary storage; and 3) application layer: healthcare professionals and policymakers may review the results to make quick decisions that stop the spread. About 2 gigabytes of historical COVID-19 chest X-ray (CXR) images were used. All iterations demonstrate IPv6 latency is lower than IPv4. This shows IPv6 transfers data more efficiently in the investigated circumstance. Little variation in IPv6 latency across rounds. 100 iterations of IPv4 and IPv6 scalability evaluations show how each handles escalating network loads. IPv6's scalability range is better and more versatile, 10-14, with a maximum approaching 15. IPv4 is less scalable (6-10). Repeated IPv4 and IPv6 data collecting periods differ greatly, reflecting network performance or environmental issues. The IPv4 is shorter peak-trough range indicates more stability. IPv6 is more prone to network that has higher peak and trough transmission time. This study demonstrates a scalable pandemic (X-ray data) transportation via simulation with a proof of concept.

Traditional security methods need to be improved as a result of security difficulties over time. Biometrics was introduced as a result of this. The sclera has been an area of extensive study recently as far as biometrics is concerned. This is because it is accurate; nevertheless, the application of this biometric feature has been limited by its segmentation. It is still necessary to improve segmentation accuracy even though several techniques have been published in the literature. This study recommends using a sclera detection tracing (SDT) approach in conjunction with the circular Hough transform. Additionally, a system based on the discrete wavelet transform (DWT) fusion of local binary-based features of the iris and sclera was proposed by the study. The fusion was passed to a Siamese network for classification. A comparison between the outcomes of the unimodal and bimodal systems was conducted. The result showed that the best performance of 98.5 was obtained for the fusion of the two biometrics. Likewise, the sclera result based on the sclera detection algorithm performed better than the segmentation that was done with the convolutional neural network (CNN).

The security of medical information has become a significant challenge with the move from traditional filing systems to electronic records. This study proposes the use of blockchain technology to address these concerns. The system registers patients and medical staff with unique IDs and stores patient diagnoses as immutable records on the blockchain. A central interplanetary file system stores the collected data, which can be accessed by authorized users like nurses, pharmacists, and patients via special access details. Users must log in before accessing medical records through the Electronic Record Management system. This approach can be scaled to multiple hospitals. After testing, the system showed some latency issues with 100 nodes, but performance improved with more nodes (300-500), demonstrating better scalability as the system handles more data and hospitals. Overall, the proposed blockchain-based system offers a secure, scalable solution for managing and accessing medical records.

Precise crop yield estimation is essential for food safety, resource utilization, and green agriculture. Classical machine learning approaches pose privacy and communication issues as they involve centralized use of data. Federated Learning (FL) overcomes this by supporting decentralized model training, but FL models based on CNN are unable to handle long-distance spatial relationships in multispectral images. To address this, we introduce FedViT-YIELD, a privacy-preserving FL framework using Vision Transformers (ViTs) for better spatial and contextual representation. ViTs operate on entire image patches, extracting both local and global features. We present FedPMA, a personalized aggregation method to address non-IID data and maintain client-specific features. For privacy, we use Differential Privacy and Homomorphic Encryption. FedViT-YIELD surpasses conventional CNN-based FL models with a 12% improvement in accuracy and 35% communication overhead reduction, with an R2 score of 0.89 on real-world datasets. Our approach is scalable and secure, laying the groundwork for smart, privacy-preserving precision agriculture.

Organ transplantation is a life-saving procedure, but matching donors with recipients remains a highly complex and critical task. Accurate donor-recipient matching reduces the risk of organ rejection and enhances the recipient's survival chances. In this project, we propose a Recurrent Neural Network (RNN)-based approach to improve the efficiency of donor-recipient matching by considering both historical medical data and real-time physiological metrics. The system leverages RNN's ability to analyze time-series data, including recipient's medical history, blood type compatibility, organ type, and donor viability. By using deep learning, the system provides optimized matches that improve transplant outcomes, significantly reducing organ rejection and post-transplant complications. Organ transplantation is a life-saving procedure that requires precise matching between donors and recipients to ensure compatibility and reduce the risk of organ rejection. This project presents a deep learning-based system for the identification of matching donors and recipients using Recurrent Neural Networks (RNN). The proposed system integrates pre-processing techniques to handle missing and noisy data, making it robust for real-world applications in healthcare institutions. Experimental results demonstrate that our model achieves 98% accuracy, 97.3% precision, and 96.5% recall on the donor-matching dataset, outperforming existing models like Logistic Regression and SVM. Superior performance over existing methods, providing a novel, technically robust, and practically viable solution which improve donor-recipient matching for organ transplantation.

Gender identification from videos is a challenging task with significant real-world applications, such as video content analysis and social behavior research. This study proposes a novel approach, the White Shark Optimizer-Support Vector Machine (WSO-SVM), tailored specifically for gender identification from video data. The WSO-SVM integrates the White Shark Optimizer, a bio-inspired optimization algorithm mimicking the hunting behavior of white sharks, with the Support Vector Machine is a highly effective machine learning techniques used for categorization. By combining these two methods, we aim to exploit the advantages of both algorithms and enhance gender identification accuracy. To evaluate the performance of the WSO-SVM in gender identification, the work conducted extensive experiments using a diverse dataset of video clips containing individuals of various genders and backgrounds. The work compared the results with conventional SVM-based gender identification and state-of-the-art techniques. The outcome of this study demonstrates that the WSO-SVM achieves superior accuracy in gender identification compared to traditional SVM-based approaches. The WSO-SVM’s ability to efficiently explore the solution space and select optimal SVM parameters contributes to its improved performance. Moreover, the WSO-SVM demonstrates resilience in dealing with fluctuations in lighting conditions, stances, and facial expressions, rendering it highly suitable for gender recognition tasks in real-world video scenarios. The outcomes derived from the SVM approach demonstrate that WSO-SVM produced an average FPR of 7.14

Comparing the various ML models is essential to understanding their advantages and disadvantages so that they can be used effectively for tasks in the futureDespite the contributions of ML models to prediction tasks, the empirical evaluation of these models for prediction tasks is reliant on synthetic datasets, which is still an issue in the field; so, this work is informed. In this paper, we carried out an empirical study comparison between three variations of ML models, namely, Logistic Regression (LR), Decision Tree (DT)and Random Forest (RF) in predicting COVID-19. All models used in this study were evaluated with different metrics such F1-score, precision, recall, accuracy and confusion matrix. The study approach presents a complete prediction procedure of the COVID-19 status using patients readings dataset. It has four steps: collecting data, getting it ready, training the model, and testing the ML model. The results show all the ML models in this study performed very. Each model reported an accurate score of more than 0.9. Overall, RF outperformed the three ML models tested in this study. RF said that recall, precision F1 score, and accuracy were 096, 1, 0.98, and 0.98, respectively. RF outperforms other models in learning the pattern and association in datasets with sequential or temporal properties due to their inherent architecture for extracting correlation in the dataset.

Internet of Things (IoT) is playing a vital role in healthcare by automating the real time monitoring of patients seamlessly with the help of a variety of sensors. In the current scenario of smart healthcare, IoT with the integration of machine learning and deep learning algorithms is instrumental in the management of chronic diseases like chronic obstructive pulmonary diseases (COPD). IoT when deployed for COPD, supports in the monitoring of lung functioning, uninterrupted patient monitoring with personalised treatment planning and also facilitates in the implementation of telemedicine technology. This paper presents a comprehensive insight on the diagnosis and treatment strategies of COPD from multiple modalities viz. smart sensors for respiratory monitoring, auscultations and imaging-based diagnosis. An exhaustive experimentation of various machine learning and deep learning algorithms for the diagnosis of respiratory illness from the auscultation recordings corresponding to four classes: normal, crackle, Rhinchi and wheeze has been presented as a case study. Lung sounds were analysed in different dimensions by considering the signal as one-dimensional (1D) vector and multiple time frequency representations derived from the 1D signals. Statistical, spectral and cepstral features derived from the signals were classified using decision tree, support vector machine, K- nearest neighbor and random forest classifiers. In a different spectrum, convolutional neural networks were employed to perform classification of lung sounds from 1D signals as such and spectrogram, melspectrogram and scalogram representations derived from 1D signals. From the outcomes of experimentation, it is inferred that visual representations of signal frequency spectrum in the form of melspectrogram representations play a significant role in categorizing the patterns belonging to four classes of respiratory diseases. The proposed model COPDScope upon experimenting with different optimizers, it is found that Melspectrogram representation records a better performance over the other two representations.

Utilizing microservices, the Internet of Things (IoT), and advanced machine learning algorithms, this study addresses the escalating global demand for sustainable and effective environmental monitoring systems. The focus is on augmenting data gathering, processing, and visualization, with the study delving into the development and deployment of a microservice-based application dedicated to environmental monitoring. Initial data analysis is conducted using simpler models like Logistic Regression. Advanced predictive models, such as XGBoost and AdaBoost, are developed and applied to forecast environmental parameters. Comparative analysis using Random Forest ensures the robustness and accuracy of the chosen models. These models collectively enhance the environmental monitoring system's capability to provide real-time insights and support decision-making processes. The paper showcases the efficacy of machine learning techniques such as Random Forest, Logistic Regression, AdaBoost, and XGBoost in anticipating and evaluating environmental data. Drawing on a dataset sourced from IoT sensors, the research demonstrates the potential for precise and real-time environmental monitoring. The methodology section provides a comprehensive outline of the project strategy and scientific techniques, accompanied by a thorough review of relevant literature to identify potentials and gaps in the field. The proposed microservice architecture is not only robust and scalable but also adaptable to diverse applications, contributing significantly to the discourse on environmental monitoring. The study's results suggest promising avenues for further research in this emerging area and lay a solid foundation for enhancing data-driven decision-making in environmental management.

Explainable Decision Support Systems (XDSS) have emerged as a critical tool for integrating artificial intelligence (AI) into decision-making processes, combining predictive accuracy with interpretability to foster user trust and accountability. Despite their increasing adoption across various domains, XDSS face significant computational challenges, including data complexity, scalability, real-time processing demands, and ensuring fairness and robustness. These challenges are further compounded by the unique requirements and constraints of different application areas, which directly influence system performance and utility, making the strategic selection of application areas a crucial step in optimizing XDSS performance. Therefore, this paper employs an adaptation of Multi-Criteria Decision Analysis (MCDA) to systematically evaluate and rank potential application areas based on domain-specific factors such as data characteristics, explanation requirements, and computational constraints. Through a detailed analysis of challenges and application contexts, this paper underscores the importance of domain selection in maximizing the practical utility and computational efficiency of XDSS. The findings emphasize that selecting the right application area is foundational to ensuring XDSS efficiency and highlight how the MCDA framework can be extended to support further configuration decisions within selected domains. This paper contributes to the strategic planning and development of future XDSS frameworks, offering guidance for developers and business leaders aiming to implement these systems more effectively.

The integration of Africanization in computing education is a growing necessity to ensure relevance, inclusivity, and contextualized learning. However, little research has explored how Africanization can be effectively incorporated into technical subjects such as Distributed Systems. This paper presents a pedagogical reflection on an attempt to infuse Africanization into a third-year Distributed Systems course, using the development of an African language dictionary as a case study. This study critically reflects on the experience of incorporating Africanization into Distributed Systems teaching, using Gibbs' Reflective Cycle to assess teaching methods, challenges, and learning outcomes. The goal is to explore how contextualized teaching approaches can be applied in computing education while addressing technical and pedagogical complexities. Data sources include teaching observations, lesson planning insights, and post-course reflections on classroom dynamics. The African language dictionary project served as a practical implementation of distributed computing concepts, requiring students to engage with distributed data storage, synchronization, and fault tolerance, while also applying computational principles to an Africanized context. Observations indicate increased engagement when localized context were provided, yet challenges such as balancing technical depth with cultural contextualization, scalability of language-based projects, and curricular constraints emerged. This paper underscores the potential of Africanization as a pedagogical tool in computing education, bridging local knowledge and modern technology. It contributes to the broader discourse on decolonizing STEM curricula in particular and educational curricula in general, advocating for culturally relevant computing education. The paper concludes with recommendations for refining Africanized computing pedagogy and exploring scalable models for integration into technical curricula.

The research presents a deep learning approach for detecting complications like preterm birth and endometriosis using machine learning techniques using two datasets: a dataset of ultrasound images which consists of various parts of the fetus and mother, and another dataset of histopathological image samples which are microscopic photos of the endometrium tissue present in the uterus lining. These images capture maternal-fetal development for monitoring parts of fetus and maternal cervix at different weeks of the prenatal period and provide valuable insights. The model uses convolutional neural networks (CNNs) architecture based on Densely Connected Convolutional Networks and Efficient Nets to analyze these images, classifying ultrasound images into features such as Fetal Abdomen, Fetal Brain, Fetal Femur, Fetal Thorax and Maternal Cervix. By training on the Maternal fetal ultrasound dataset of 12400 images, the model achieves a 94.40

In today’s data-driven business environment, organizations encounter challenges in effectively utilizing predictive analytics to drive decision-making while ensuring ethical data practices. This paper proposes an AI-powered predictive analytics framework to improve operational efficiency, enhance decision accuracy, and maintain data privacy and security. The system integrates machine learning techniques with ethical governance policies, enabling real-time insights from large datasets. AI helps businesses analyse past data and predict future trends by utilizing predictive and prescriptive analytics. This allows for more proactive and well-informed initiatives. Machine learning is essential for automating data-driven decisions and providing businesses with real-time insights that enable them to react rapidly to changing market circumstances. A comparative evaluation demonstrates that the system achieves 92.5

This study proposes a system for biometric access control utilising the improved Cultural Chicken Swarm Optimization (CCSO) technique. This approach mitigates the limitations of conventional Chicken Swarm Optimization (CSO), especially in dealing with larger dimensions due to diversity loss during solution space exploration. Our experimentation involved 600 sample images encompassing facial, iris, and fingerprint data, collected from 200 students at Ladoke Akintola University of Technology (LAUTECH), Ogbomoso. The results demonstrate the remarkable effectiveness of CCSO, yielding accuracy rates of 90.42%, 91.67%, and 91.25% within 54.77, 27.35, and 113.92 s for facial, fingerprint, and iris biometrics, respectively. These outcomes significantly outperform those achieved by the conventional CSO technique, which produced accuracy rates of 82.92%, 86.25%, and 84.58% at 92.57, 63.96, and 163.94 s for the same biometric modalities. The study's findings reveal that CCSO, through its integration of Cultural Algorithm (CA) Operators into CSO, not only enhances algorithm performance, exhibiting computational efficiency and superior accuracy, but also carries broader implications beyond biometric systems. This innovation offers practical benefits in terms of security enhancement, operational efficiency, and adaptability across diverse user populations, shaping more effective and resource-efficient access control systems with real-world applicability.

Over the years, farmers have continuously faced the issue of planting various crops at locations that are not suitable for the growth of certain crops and that in turn affects the crop yield per season. This is because they still use the traditional method of making these inaccurate predictions. Several literatures have proposed different machine learning techniques for crop yield prediction. However, not many literatures have compared the performance of different machine learning technique in the prediction of crop yield. Hence, in this study examines the performance of three machine learning techniques in predicting crop yield. For this study, soil and climate factors were considered. The system started with a general data cleaning. The cleaning was aimed at preparing the data for classification while the correlation was aimed at determining the relevance of factor to the yield of the farm. The three methods considered include Multi-layer Perceptron, Artificial Neural Network and Long Short-Term Memory. Amongst the three techniques considered, LSTM had the best accuracy of 97%, ANN had 86% and MLP had 87% accuracy respectively.

DDoS attacks have emerged as the most serious Internet risks, causing significant damage to businesses and governments. This study utilized the BoT-IoT dataset obtained from the University of New South Wales (UNSW) website on machine learning (ML) algorithms to prevent HTTP-based DDoS attacks in the Internet of Things environment. Character encoding was performed to convert the strings such as DDoS attack to 1 and Non-malicious to 0. Feature scaling and normalization were carried out to normalize numerical data with large values using a Python module called standard scaler to avoid model overfitting or underfitting. According to the performance evaluation metrics results, decision tree is the best model with 99.96% accuracy, 0.9979 R 2 , 0.0003 MAE, 0.0003 MSE, 0.0195 RMSE, 1.0 precision, 0.9994 recall, 0.9997 F1_score after dimensionality reduction. After comparing various ML models, this study concludes that the decision tree is the golden model for HTTP-based DDoS attack prevention in IoT devices. The study recommends the utilization of other performance evaluation metrics such as Matthew Correlation Coefficient (MCC), Kappa Statistic and Huber loss for further studies to avoid bias and subjectivity in the model selection where there are surrogate models. Also, deep learning paradigms can be implemented for the prevention of HTTP-based DDoS attacks in cloud environments.