Early prediction of patients in need of admission to the intensive care unit (ICU) is essential for maximizing the use of available hospital resources and enhancing the quality of patient care outcomes. This work uses the Covid19MPD Dataset to predict ICU admissions based on various machine learning techniques such as Random Forest, Support Vector Machine, Gradient Boosting, and Multi-Layer Perceptron alongside Explainable Artificial Intelligence (XAI) approaches. Our findings show that the Gradient Boosting model achieved the best accuracy at 97.49% and an F1 score of 71% for ICU admissions. Notably, the study finds that age and pneumonia are important predictors, with patients 45 years and older who come with COVID-19 and pneumonia having a much higher chance of needing ICU care. These findings highlight how important it is to use machine learning models in clinical settings in order to improve ICU admission prediction and facilitate prompt medical intervention.

Flamingos are known for their vibrant pink and reddish hues, which are not merely aesthetic but indicative of their overall health and diet. These colors are derived from carotenoid pigments in their food sources, making coloration a vital marker for monitoring their well-being and environmental conditions. This study introduces a two-stage classification methodology designed to safeguard flamingo populations by leveraging deep learning techniques. Convolutional Neural Networks (CNNs) are used for both shape and color detection, ensuring accurate identification of flamingos and insights into their health status. Simulation results demonstrated the CNNs model’s effectiveness, making it a valuable resource for wildlife conservation efforts aimed at preserving flamingo habitats. The first stage employs digital classification filters to distinguish flamingo images from other species, achieving an accuracy of 97.52%, while the second stage refines these detections through color analysis with an accuracy of 86.27%. This approach promises to mark a significant advancement in wildlife conservation, offering reliable methods for assessing and managing flamingo populations in their natural environments.

This paper focuses on the signal classifier component of a tele-rehabilitation framework that uses wearable surface electromyography (sEMG) devices to identify hand grasps and deliver real-time, adaptive therapy. The prediction accuracy and robustness of a stacking ensemble classifier significantly — with the combination of Support Vector Machine (SVM), Random Forest (RF), and a Logistic Regression meta—learner-improves performance, enabling continuous and personalized monitoring. Through the use of machine learning techniques, this study enables patients to conduct rehabilitation exercises remotely and to track their progress in real-time with therapists.

Osteoporosis, a prevalent skeletal disorder characterized by weakened bone strength and integrity, poses a significant health risk, particularly for older adults and postmenopausal women. Early detection is critical to mitigate fracture risks and improve patient outcomes. This research investigates the potential of pretrained convolutional neural networks for automated osteoporosis detection in knee X-ray images and highlighting the impact of image preprocessing techniques on model performance. We evaluate four pretrained models (ResNet-50, ResNet-101, VGG16, and VGG19) for feature extraction, coupled with a Random Forest classifier optimized using Bayesian Optimization. Our framework explores the effectiveness of different preprocessing methods, including Bilateral Filtering and Contrast Limited Adaptive Histogram Equalization, to enhance feature quality and improve classification accuracy. Evaluation on a dataset of knee X-ray images collected from the University of Sharjah Hospital reveals that VGG architectures, particularly VGG16, demonstrate superior performance in detecting knee osteoporosis. VGG16, preprocessed using Bilateral Filtering and CLAHE, achieved a $\mathbf{7 8 . 3 9 \%}$ accuracy, $\mathbf{8 0 . 0 0 \%}$ precision, and $\mathbf{7 6 . 9 2 \%}$ recall. This study underscores the importance of selecting both model architecture and preprocessing techniques for optimal performance. Further optimization and more data could potentially lead to a more robust and accurate model.

Alzheimer’s disease (AD) is a complex neurodegenerative disorder that presents significant challenges for early and accurate diagnosis. Early diagnostic and treatment strategies can help enhance the circumstances by slowing the progression of the illness and enhancing the patient and family’s quality of life. Machine learning (ML) approaches have shown promise in improving the diagnosis and prognosis of Alzheimer’s based on relevant risk factors. This paper aims to develop and evaluate a machine learning model for classifying Alzheimer’s, mild cognitive impairment (MCI), and normal cognition (NC) using a diverse data set from the ADNI database. The model had high performance with a sensitivity rate of up to 97%, accuracy rate of up to 94%, and specificity rate of up to 96%. Moreover, none of the Alzheimer’s cases were falsely detected as normal cognition but as mild cognitive impairment and none of the normal cognition cases were detected as Alzheimer’s, but as mild cognitive impairment. The algorithm that has the highest number of true positive detections, which is 78 out of 85 Alzheimer’s cases, is the decision tree algorithm. The performance of the system heralds a promising future for Alzheimer’s diagnosis by machine learning with the aim of developing smart health systems.

Anomaly detection of user behavior in smart homes is vital due to its potential impact on improving the quality of life. Many anomaly detection systems proposed in the literature rely on a wide range of user activities performed at different parts of the smart home environment. This paper presents an anomaly detection framework focused on kitchen activities, a specific area that is not sufficiently addressed by similar studies. The two activities are 'eating' and 'making simple food'. This paper exploits the capability of machine learning algorithms, namely: Artificial Neural Network (ANN), Extreme Gradient Boosting (XGBoost), and Support Vector Machine (SVM). The performance of those algorithms was evaluated and ranged between 80% and 100% for best cases, Demonstrating their high anomaly detection potential despite the relatively small size of the scenario and dataset.

This paper proposes a load scheduling approach for a residential home in an islanded PV microgrid scenario based on a Genetic Algorithm (GA). The primary aim is to inform on how a Demand Side Management (DSM) could reduce the capital cost of the residential home energy use and operational cost of the microgrid by minimizing the use of fossil fuel generator. The research study proposes and describes the design for load allocation to achieve utilization of solar PV resources optimally. The proposed scheme is based on the time-of-use (TOU) and improvement of electricity users' comfort. The demonstration of the concept is presented and discussed based on a single smart home scenario using a solar PV microgrid and battery in a rural community of Enugu State, Nigeria.

Loss of control in-flight (LOC-I), following loss of situational awareness and startle has been identified as a leading cause of aviation-based fatalities in recent decades. This has led to significant effort toward improving safety records, particularly in the fields of flight crew training and in-flight support technologies that aid better decision making and management of reactions to a startling occurrence. One way to achieve quality decision making in the cockpit is by providing adequate cueing and response activating mechanisms carefully designed to aid human information processing. These response performances, especially in the context of reactionary management of startle in flight, could be honed through simulator-based training. This paper discusses the setup background development of a startle causality dynamics using fuzzy cognitive mapping. This mapping provides an objective, strategy framework; for determining the required training for a management of the startle reflex in extenuating circumstances made significantly worse by human factors, such as erroneous knee jerk reactions.

An evolutionary-based path planning is designed for an Autonomous Surface Vehicle (ASV) used in environmental monitoring tasks. The main objective is that the ASV covers the maximum area of a mass of water like the Ypacarai Lake while taking water samples for sensing pollution conditions. Such coverage problem is transformed into a path planning optimization problem through the placement of a set of data beacons located at the shore of the lake and considering the relationship between the distance travelled by the ASV and the area of the lake covered. The optimal set of beacons to be visited by the ASV has been modeled through Eulerian circuits. Due to the complexity of the optimization problem, a metaheuristic technique like a Genetic Algorithm (GA) is used to obtain quasi-optimal solutions in both models. The parameters of the GA are tuned and then the obtained Eulerian Circuit is compared with a lawnmower and a random approaches obtaining an improvement of up to the double of the lake.

The ability to forecast natural phenomena relies on understanding causality. By definition this understanding must include a temporal component. In this paper, we consider the ability of an emerging class of neural network, which encode temporal information into the network, to perform the difficult task of Natural Event Forecasting. The Axonal Delay Network (ADN) models axonal delay in order to make predictions about sunspot activity, the Auroral Electrojet (AE) index and daily temperatures during a heatwave. The performance of this network is benchmarked against older types of neural networks; including the Multi-Layer Perceptron (MLP) network and Functional Link Neural Network (FLNN). The results indicate that the inherent temporal characteristics of the Axonal Delay Network make it well suited to the processing and prediction of natural phenomena.

The advancements in Unmanned Aerial Vehicle (UAV) related technologies and wireless communications pave the way for the deployment of wireless mesh networks in the air. These air mesh networks can be suitable for providing communication services in disaster scenarios to ground nodes such as victims and first responders. However, the optimal deployment of UAVs is not an easy as the number of possible scenarios to position the UAVs may reach a computationally challenging level. The combination of global and local search optimization algorithms can be considered as a powerful way for dealing with the massive number of possible solutions. We propose a deployment approach based on a global search algorithm such as the genetic algorithm and a local search algorithm namely the hill climbing algorithm. We show that the combination of both optimization techniques provides promising results for optimal positioning of UAVs in disaster scenarios based on simulation examples.

This study builds on previous research, where a technology acceptance model for electronic health (e-HTAM) was investigated, developed and evaluated. The e-HTAM questionnaire originally used showed some weakness in terms of its overall reliability, which highlighted the need for a second phase of study to address the shortfalls reported by the first model. The second phase of the study was conducted after the questionnaire used in the initial phase was modified. The results suggested that when creating e-Health websites or services, the principle of how e-Health websites and services should be designed and delivered, and under which cultural setting they will be used, should be taken into consideration from the initial design stage.

In this paper a novel application of a particular type of spiking neural network, a Polychronous Spiking Network, for financial time series prediction is introduced with the aim of exploiting the inherent temporal capabilities of the spiking neural model. The performance of the spiking neural network was benchmarked against two “traditional”, rate-encoded, neural networks; a Multi-Layer Perceptron network and a Functional Link Neural Network. Three nonstationary and noisy time series are used to test these simulations: IBM stock data; US/Euro exchange rate data, and the price of Brent crude oil. The experiments demonstrated favourable prediction results for the Spiking Neural Network in terms of Annualised Return, for both 1-Step and 5-Step ahead predictions. These results were also supported by other relevant metrics such as Maximum Drawdown, Signal-To-Noise ratio, and Normalised Mean Square Error. This work demonstrated the applicability of polychronous spiking network to financial data forecasting and that it has the potential to function more effectively than traditional neural networks, in nonstationary environments.

One of the major concerns of research in integrated healthcare information systems is to enable decision support among clinicians across boundaries of organizations and regional workgroups. A necessary precursor, however, is to facilitate the construction of appropriate awareness of local clinical practices, including a clinician's actual cognitive capabilities, peculiar workplace circumstances, and specific patient-centered needs based on real-world clinical contexts across work settings. In this paper, a user-centered study aimed to investigate clinical practices across three different geographical areas-the U.K., the UAE and Nigeria-is presented. The findings indicate that differences in clinical practices among clinicians are associated with differences in local work contexts across work settings, but are moderated by adherence to best practice guidelines and the need for patient-centered care. The study further reveals that an awareness especially of the ontological, stereotypical, and situated practices plays a crucial role in adapting knowledge for cross-boundary decision support. The paper then outlines a set of design guidelines for the development of enterprise information systems for e-health. Based on the guidelines, the paper proposes the conceptual design of CaDHealth, a practice-centered framework for making sense of clinical practices across work settings for effective cross-boundary e-health decision support.

This paper presents the design of Cad Health, a system aimed at enabling knowledge and work practice transfer among clinicians across geographical, regional and workplace boundaries for effective clinical decision support in e-health. The system offers a unifying structure that allows clinicians to make sense of clinical work situations across regional and workplace boundaries in an e-health environment. The approach we have taken in Cad Health is motivated by the fact that 1) patterns of clinical work practice have been found to vary significantly across work settings, and 2) the contextual cues and practice-based knowledge, which are offered by common problem solving contexts in co-located work settings, and which enable clinicians, in such settings, to share information and knowledge to support one another's clinical decision making do not exist in e-health and other distributed work contexts. In particular, we highlight a number of user-informed design considerations, and describe the architecture and prototype of Cad Health.

This study concerns the problem of controlling multiclass (elastic, inelastic and unresponsive) Internet traffic without sacrificing quality of service (QoS) by adopting a unified ‘resource allocation and traffic management’ approach. The aim is to minimise the need for relying on dedicated QoS traffic control mechanisms in order to avoid spiralling complicatedness that, in practice, leads to ‘robust yet fragile’ Internet. In order to address this challenge, the authors first introduce an end-to-end non-convex network utility maximisation-based resource allocation algorithm to guarantee enhanced QoS to elastic and inelastic flows. Then, a pricing-based fair and scalable traffic management scheme, called Purge, is introduced to protect transmission control protocol-friendly traffic from unfairness attacks by unresponsive flows. Finally, the main contribution of this work, the unified algorithm, is developed by adapting Purge to complement link-control of the proposed resource allocation algorithm to enable it to enforce fairness while maintaining a scalable network core. The unified approach thus delivers QoS guarantees for multiclass traffic.

In the context of e-health, professionals and healthcare service providers in various organisational and geographical locations are to work together, using information and communication systems, for the purpose of providing better patient-centred and technology-supported healthcare services at anytime and from anywhere. However, various organisations and geographies have varying contexts of work, which are dependent on their local work culture, available expertise, available technologies, people's perspectives and attitudes and organisational and regional agendas. As a result, there is the need to ensure that a suggestion - information and knowledge - provided by a professional to support decision making in a different, and often distant, organisation and geography takes into cognizance the context of the local work setting in which the suggestion is to be used. To meet this challenge, we propose a framework for context-aware knowledge modelling in e-health, which we refer to as ContextMorph. ContextMorph combines the commonKADS knowledge modelling methodology with the concept of activity landscape and context-aware modelling techniques in order to morph, i.e. enrich and optimise, a knowledge resource to support decision making across various contexts of work. The goal is to integrate explicit information and tacit expert experiences across various work domains into a knowledge resource adequate for supporting the operational context of the work setting in which it is to be used.

Network utility maximization (NUM) provides an important perspective to conduct rate allocation where optimal performance, in terms of maximal aggregate bandwidth utility, is generally achieved such that each source adaptively adjusts its transmission rate. Behind most of the recent literature on NUM, common assumptions are that traffic flows are elastic and that their utility functions are strictly concave. This provides design simplicity but, in practice, limits the applicability of resulting protocols, in that severe QoS problems may be encountered when bandwidth is shared by inelastic flows. This paper investigates the problem of distributively allocating data transmission rates to multiclass services, both elastic and inelastic, and overcomes the restrictive and often unrealistic assumptions. The proposed method is based on the Lagrangian Relaxation for a dual formulation that decomposes the higher dimension NUM into a number of subproblems. We use a novel Surrogate Subgradient based stochastic method to solve the dual problem. Unlike the ordinary subgradient methods, surrogate subgradient can compute optimal prices without the need to solve all the subproblems. For the lower dimension, nonlinear and nonconvex subproblems we use a hybrid particle swarm optimization (PSO) and sequential quadratic programming (SQP) method, where the objective is to achieve fast convergence as well as accuracy. We demonstrate the efficiency of the proposed rate allocation algorithm, in terms maintaining QoS for multiclass services, and validate its scalability and accuracy for large scale flows.

Collaboration between urban planners, government officers and other stake-holders is a key task in the urban planning procedure. This paper presents a VR-centred workspace prototype for supporting collaborative urban planning. This workspace comprises a variety of technologies such as semi-immersive stereo display, table-top display, mobile devices and an optical tracking system. A system framework for VR-centred workspace integrates a user interaction layer along with service and data management capabilities. This open-structured framework can support various visualisation and simulation modules and facilitates collaboration. We have implemented this prototype and some test scenarios have been examined with user groups.

This paper presents an optimisation application to support the construction site planning task by finding efficient paths between two site locations based on a combination of safety, transportation cost, and visibility criteria. These criteria can be combined or individually optimised by mathematical search algorithms, namely Dijkstra and A*, in order to present site planners with the safest path, the shortest distance path, the most visible path, and the paths that reflect a combination of low risks, short distance, and high visibility measures between two site locations. This paper identifies the need for the use of simulation for site layout analysis, in particular, it investigates the potential application of mathematical optimisation techniques for the selection of site paths.

This paper presents the design of a simulation environment for the modelling, visualisation and optimisation of construction site layouts. This construction site workspace application aims to support the site planning task by analysing space and risk on the site and generating automated site layouts which satisfy a combination of cost, efficiency and safety criteria. The construction site simulation environment forms part of an EU funded project called DIVERCITY (Distributed Virtual Workspace for enhancing Communication within the Construction Industry). This is concerned with the development of virtual simulation prototypes to support the client briefing, design review, and construction planning stages of the construction process. The site analysis module comprises a safety analysis component, a space analysis component, and an optimisation component. The safety analysis component is a generic model for the graphical and numerical representation of risk/hazard spaces. The space analysis component represents and classifies the various spaces on the construction site according to their relative importance in terms of accessibility and visibility. The optimisation component takes safety and space analysis information, and performs site layout optimisation according to travelling distance minimisation, risk minimisation, and space use maximisation criteria.