Handover management in 5G/6G networks presents significant challenges mainly due to ultradense deployments, high mobility and diversity of scenarios. This work presents a digital twin-based Deep Q-Learning (DQL) strategy to optimize handover processes and resource management in these networks. It is proposed to consider digital twins to support the implementation of real-time network conditions where DQL agents are trained to improve quality of service (QoS) parameters related to throughput. The suggested approach, deployed with the ns-3 mmWave module and ns3-ai framework, shows relevant benefits, including reduced handovers. Moreover, the consideration of digital twins allows adaptive learning and scalability, allowing networks to respond more effectively to dynamic user behavior and environmental modifications. These results underline the potential of combining DQL and digital twins as an efficient solution for smart handover management, and thus, for sustainable and efficient communication in 5G/6G networks.

ABSTRACT Wireless sensor networks (WSNs) is a wireless system including the set of distributed sensor nodes used for physical or environmental observation. A network energy expenditure is considered as a significant concern because of battery restricted sensors of the WSN. Clustering and multi hop routing are considered as effective approaches to enhance the network lifecycle and communication. Achieving the anticipated objective of reducing the energy expenditure, thereby increasing the network lifecycle, is considered as an optimisation issue. In recent times, a nature inspired meta‐heuristic approaches are extensively utilised for solving the different optimisation issues. In this context, this research aims to accomplish the objective by proposing the multiobjective‐perturbed learning and mutation strategy based artificial rabbits optimisation namely M‐PMARO for an optimum cluster head (CH) selection and route discovery. The proposed M‐PMARO incorporates an experience based perturbed learning (EPL) and mutation strategy to identify the capable regions over the search space for enhancing the exploration and avoiding the local optima issue. To formulate the multiobjective, the residual energy, average intracluster distance, average base station (BS) distance, CH balancing factor (CHBF) and node centrality are incorporated for optimum CH discovery while the residual energy and average BS distance are considered for multi hop routing. The M‐PMARO is analysed based on alive nodes, dead nodes, energy expenditure, throughput and data received in BS and network lifecycle. The viability of M‐PMARO is validated by comparing it with existing approaches such as fitness based glowworm swarm with fruitfly algorithm (FGF), energy balanced particle swarm optimisation (EBPSO), improved bat optimisation algorithm (IBOA), graph neural network (GNN) and fuzzy logic and particle swarm optimisation (PSO) based clustering routing protocol namely PFCRE. The alive node count of M‐PMARO is 100 for 1200 rounds, which is higher than the EBPSO.

The rapid growth in the number of devices connected to the Internet has led to a surge in sophisticated, intelligent service-based applications, exploiting new paradigms and technologies such as edge and cloud computing. These applications essentially result from the interaction between various services. However, this interaction requires the collection, integration, and communication of sensitive data, raising significant concerns about data vulnerability. Users are often concerned about the safety and management of their private information, which makes this crucial issue particularly relevant to the design of automatic services. To address this issue, we propose a novel trustful automatic sensitive service composition approach that leverages machine learning, including the Bayesian average, for accurate trust prediction. The proposed Majority Voting and Bayesian Average-based Trustful Service Composition (MVBA_ TSC) heuristic aims to generate a feasible and efficient service composition by dynamically managing a Trust Service Provider Network (TSPN). Our approach focuses on preserving user privacy by exclusively involving trustworthy service providers in the composition process. This heuristic is geared towards selecting the feasible composition by identifying the most reliable path within the trusted network, considering the Quality of Service (QoS) and trust dimensions among service providers. Experimental results confirmed that MVBA_TSC can improve the trustworthiness of sensitive service composition efficiency by 17 % compared to competitive solutions.

The curse of dimensionality significantly hinders healthcare computational models due to the large size of the data. In this benchmarking study, we are using different feature selection and transformation methodologies to compare the computational efficiency and complexity of a Random Forest (RF) and Decision Tree (DT) classifier models for predicting heart disease risk. We investigate the performance of multicollinearity removal using Stratified cross-validation, Factor Analysis (FA), Principal-Component Analysis (PCA), and Linear-Discriminant Analysis (LDA) in terms of their ability to improve prediction accuracy and reduce computational time. The proposed RF model with appropriate feature engineering technique can produce a CHD risk prediction accuracy of 91% with only 7 features which is 8.33% higher than the preliminary model. Notably, we are also able to achieve a more than 33% reduction of time complexity in our best model as compared to the preliminary model. The statistical significance of our models’ results has been validated with a Z-test and Analysis of Variance (ANOVA). Our results provide insights into the trade-offs between prediction accuracy and computational efficiency when using these models for heart disease risk prediction. They have important implications for their practical use in clinical settings.

This COVID-19 pandemic is impacting the world in health and economic terms since 2020 with more than 200 million confirmed infected people and more than 4 million deaths across 190 countries. Treatment used against COVID-19 disease has initially been based on the combination of several medicaments, such as hydroxychloroquine/chloroquine, azithromycin and kaletra, each of which can individually delay the ventricular depolarization and repolarization processes through morphological changes in the patient’s electrocardiogram. These changes can produce serious arrhythmias that lead to the sudden death of the patient.This paper presents an interpretable fuzzy rule-based system for fatal ventricular arrhythmia risk level estimation due to COVID-19 treatment, whose decisions are made on the basis of the evolution of electrocardiogram morphology and certain patient’s clinical information. For the risk level estimation, the proposed fuzzy rule-based system considers three different risk levels (High, Moderate and Low) which are indicated by means of three different colors (Red, Orange and Green). Decisions made by the fuzzy rule-based system present a reliable behavior in comparison with cardiologist’s decision. To be precise, the obtained accuracy, when comparing both decisions, reaches the 96.43%, which, joint to the high measured interpretability of the decision making system, result in a powerful tool in order to avoid death in patients, even in health centers without specialized clinical staff, and to reduce the stress in medical centers by reducing reaction times in critical patient situations.

At the dawn of the fourth industrial revolution, artificial intelligence is impregnating our society by an overwhelming number of applications with rising complexity. This circumstance has triggered a new debate on explainable artificial intelligence in terms of transparency and confidence. Therefore, the relevance of interpretability is remarkable since this concept provides us with transparence and understandability in avoidance of black-box systems. On the other hand, cloud computing is a new paradigm of distributed computation based on the externalization of computing needs offered as services, whose performance has a significant economic and environmental impact and is strongly influenced by the scheduler, which is, likely, the most critical part in charge of allocation computational resources. In this regard, fuzzy rule-based systems are knowledge-based systems that are increasingly arising as an alternative for the development of cloud scheduling systems, mainly due to their intrinsic features such as adaptability to environments, dynamism and capability to cope with uncertainty. Bearing in mind the above-mentioned ideas, this paper presents a study-case in which the interpretability in fuzzy rule-based schedulers for cloud computing, obtained through automatic learning processed, has been analyzed. Results show how an inherent feature, like interpretability, vanishes due to the use of learning processes for both total execution time and power consumption optimization, which put the focus on the relevance of facing interpretability in this kind of systems.

The main objective of this work is to reduce power consumption and energy of a datacenter. There are various power saving techniques. A fuzzy system-based scheduler has been used, comparing its results with other well-known algorithms. The methods used in this paper are based on a combination of the DVFS algorithm and the development of a rule-based expert system to provide power-based planners for task planning domains. The parameters considered in the system are explained in detail and the results obtained are analyzed.

This work proposes the use of bio-inspired knowledge acquisition for Fuzzy Systems founded on Swarm Intelligence-Particle Swarm Optimization (SI-PSO). Swarm-based models consider knowledge entities as particles that move in the space to reach the higher quality. Fuzzy Systems following SI-PSO for knowledge acquisition are categorized in this work as Swarm Fuzzy Systems (SFSs). Specifically, two learning methodologies, KASIA (using rule bases as particles in PSO) and KARP (using rules as particles in PSO) are introduced. SFSs performance is studied in a problem of practical importance nowadays with data sets, the learning of fuzzy meta-schedulers in computational grids. Fuzzy meta-schedulers are Fuzzy Systems doing intelligent allocation of jobs to improve the performance of the grid, such as the reduction of the execution time of workload. The scheduling decisions are taken based on the knowledge of the Fuzzy System and in this way, the relevance of their learning process are critical. In this work, compared results of the performance of the different SFSs and a comparison between SFSs and Genetic Fuzzy Systems are presented. Simulations results show that SFSs can achieve a faster convergence and higher quality with a reduced number of control parameters what makes them a good alternative to Genetic Fuzzy Systems.

While the capability for growing on demand is the design foundation of the new generation of software platforms, physical systems always have limited resources. Therefore, there is a need to optimise the existing infrastructures and evolving them building on interoperability. The optimisation of the infrastructure entails not only addressing the computational and storage capabilities but also the network, its associated intelligence and the exchanged information. Therefore, the interoperability requirement in a large scale cyber-system spans from the lowest layers, interfacing with the physical resources, to the software building blocks, client devices and handled data. In this paper, several initiatives addressing interoperability among cloud IaaS layers, IaaS-PaaS, PaaS-SaaS, Cloud-Network and data are presented.

Knowledge acquisition is a long-standing problem in fuzzy-rule-based systems. In spite of the existence of several approaches, much effort is still required to increase the efficiency of the learning process. This study introduces a new method for the fuzzy-rule evolution that forms an expert system knowledge: the knowledge acquisition with a swarm-intelligence approach (KASIA). Specifically, this strategy is based on the use of particle-swarm optimization (PSO) to obtain the antecedents, consequences, and connectives of the rules. To test the feasibility of the suggested method, the inverted-pendulum problem is studied, and results are compared for two of the most extensively used methodologies in machine learning: the genetic-based Pittsburgh approach and the Q-learning-based strategy, i.e., state-action-reward-state-action (SARSA). Moreover, KASIA is analyzed as a learning strategy in fuzzy-rule-based metascheduler design for grid computing, and performance is compared with other scheduling strategies based on genetic learning and existing scheduling approaches, i.e., EASY-backfilling and ESG+local periodical search. To be more precise, simulation results prove the fact that the proposed strategy outperforms classical learning approaches in terms of final results and computational effort. Furthermore, the main advantage is the capability to control convergence and its simplicity.