International Journal of Ad Hoc and Ubiquitous Computing (28 papers in press)

Regular Issues

• Open Access

  Dynamic load balancing model for data centres based on bidirectional network edge detection algorithm ( Free Full-text Access )

  by Diangang Hu, Xi Song, Bingling Gu, Kai Song, Xiaodong Wei 
  Abstract: To resolve link congestion and QoS deterioration caused by elephant flow conflicts in data centre networks under conventional multipath routing, this study proposes an innovative dynamic load balancing model. It combines software-defined networking (SDN) for centralised control with a Transformer-BiLSTM-DNN deep learning framework: the transformer encoder grasps global spatial correlations among traffic flows, BiLSTM retrieves intricate temporal variations, and DNN achieves deep feature integration for precise load state categorisation, enabling refined proactive traffic regulation. Evaluated on Google Borg Cluster Trace v2, the model excels in key metrics like load balancing degree, job completion time and resource utilisation, outperforming traditional baselines significantly. Ablation tests verify the essentiality of each component, providing a sturdy intelligent approach to enhance data centre network efficiency and dependability.
  Keywords: data centre; load balancing; bidirectional network; transformer; BiLSTM; SDN scheduling.
  DOI: 10.1504/IJAHUC.2026.10076107
   
  

• Open Access

  Intelligent automotive trade market forecasting and decision-making system: multi-source data fusion under meta-learning ( Free Full-text Access )

  by Kan Lu, Mingting Huang, Min Song 
  Abstract: With the continuous advancement of intelligent vehicle technology, the smart automotive trade market has become increasingly complex and dynamic. This study proposes an intelligent forecasting framework that integrates multi-source data through a meta-learning-based approach. Specifically, a meta-attention gated recurrent unit (MA-GRU) model is developed to enhance the accuracy and robustness of market predictions. The model first extracts key automotive performance indicators and market-related features using a GRU network, and then applies an attention mechanism to capture the most informative temporal dependencies. To address the challenge of data scarcity in the emerging smart vehicle market, meta-learning is introduced to improve the model's adaptability and generalisation across diverse datasets. Experimental evaluations demonstrate that the proposed MA-GRU framework achieves superior predictive performance even under limited data conditions, providing a solid technical foundation for market trend analysis and strategic decision-making in the intelligent automotive trade.
  Keywords: intelligent marketing prediction; intelligent cars; gated recurrent unit; GRU; meta-learning.
  DOI: 10.1504/IJAHUC.2026.10076104
   
  

• Open Access

  Design and optimisation of multi-party secure computing protocols supporting dynamic participant switching ( Free Full-text Access )

  by Zhipeng Li, Haiping Zhao, Shaobo Liu, Hao Yang, Shuo Shen, Rui Dong, Yu Zhang 
  Abstract: Most existing SMPC protocols assume static participants, unsuitable for dynamic scenarios where participant numbers and identities change. This paper proposes an SMPC protocol enabling dynamic participant switching, introducing a dynamically joinable model with variable committee sizes per round and supporting arbitrary member counts. It uses secret sharing to transfer computations between participants without revealing intermediates and consolidates all validations into a dedicated phase to reduce communication complexity. Experiments with seven-member committees show its communication costs (for 10,000-10,000,000 data entries) and computational overheads (for 10,000,000 entries) outperform existing protocols. A platform based on Spring Boot and MyBatis, integrating libscapi for core crypto operations, is developed with optimised architecture and visualisation tools. The protocol offers efficient, secure solutions for data sharing and collaboration, with significant theoretical and practical value.
  Keywords: multiparty secure computing; secret sharing; participant role switching.
  DOI: 10.1504/IJAHUC.2026.10076106
   
  

• Open Access

  Integrating multi-modal emotion recognition with strategy generation: a transformer approach to sustainable marketing ( Free Full-text Access )

  by Anqi Wu, Jing Wang, Jie Zhang, Wei Qiu 
  Abstract: This paper proposes the multi-modal emotion perception and strategy optimisation transformer (MEPSO-Transformer), a dual-task architecture designed to jointly perform emotion classification and green marketing strategy recommendation based on heterogeneous user data. The model incorporates modality-specific encoders, hierarchical gated cross-attention fusion, and two parallel decoders, and is trained under a multi-task learning scheme. To evaluate its effectiveness, experiments are conducted on three datasets: CMU-MOSEI, MELD, and a constructed domain-specific dataset, GreenPromo-Emotion, which contains 6,200 multi-modal samples annotated with seven emotional categories and five predefined marketing strategies. Results show that MEPSO-Transformer achieves an average accuracy of 84.3% and a macro-F1 score of 81.2% on CMU-MOSEI, outperforming the best baseline by +2.3% and +2.6% respectively. On MELD, the model attains a 75.0% macro-F1 and a Hit@1 score of 65.7%. These findings demonstrate the model's applicability to sustainable marketing tasks requiring real-time affective understanding and strategy selection.
  Keywords: multi-modal emotion recognition; green marketing; strategy optimisation; transformer; sustainable AI; affective computing; multi-task learning.
  DOI: 10.1504/IJAHUC.2026.10076105
   
  

• Open Access

  Intelligent monitoring of patient vital signs based on adaptive attention fusion spatiotemporal graph neural network ( Free Full-text Access )

  by Shunda Cheng, Jie Zhu, Shengjiang Guan, Jie Cheng, Tong Dou 
  Abstract: This study proposes a vital signs monitoring framework that addresses the limitations of traditional threshold-based alarms and existing deep-learning models in capturing multimodal physiological interactions and spatiotemporal dynamics. The method integrates an adaptive attention fusion mechanism that dynamically adjusts the importance of heterogeneous physiological parameters, a spatiotemporal graph neural network that jointly models inter-parameter correlations and temporal evolution using multi-scale windows, and a reinforcement learning module that enables active, strategy-driven early warning and clinical decision support. Evaluated on the MIMIC-III and eICU datasets, the proposed system achieves 96.3% anomaly detection accuracy, 38.5-minute early warning capability, and a 0.912 F1-score, outperforming existing methods. Ablation studies confirm the contributions of adaptive fusion, spatiotemporal graph modelling and policy optimisation.
  Keywords: adaptive attention mechanism; spatiotemporal graph neural network; ST-GNN; vital sign monitoring; deep reinforcement learning; multimodal fusion.
  DOI: 10.1504/IJAHUC.2026.10076081
   
  

• Open Access

  Adaptive deep reinforcement learning-based error analysis model for strength training ( Free Full-text Access )

  by Guobao Zhang, Haotian Li, Xurui Liu 
  Abstract: To tackle limitations of conventional strength training movement evaluation, such as inadequate movement stage segmentation and insufficient fine-grained error perception, this study proposes ARLE-Net. This framework integrates improved YOLOv8 with CBAM attention mechanism to enhance local movement feature detection, HR-Net for high-resolution human keypoint extraction, and adaptive reinforcement learning for dynamic action segmentation and standardised score prediction. Experiments on Fitness-AQA and ARFit datasets show it outperforms state-of-the-art methods in action segmentation accuracy and score prediction mean square error, with strong stability. Ablation studies confirm component effectiveness, providing support for intelligent fitness guidance and fitting the journal's enhanced ubiquitous computing focus.
  Keywords: strength training; pose estimation; reinforcement learning; action quality assessment.
  DOI: 10.1504/IJAHUC.2026.10076102
   
  

• Open Access

  Personalised service and sentiment analysis for intelligent customer relationship management systems based on natural language processing ( Free Full-text Access )

  by Beibei Li 
  Abstract: To enhance acoustic affect recognition, this study proposes the ECAPA-BGTDNN framework, which combines SE-Res2Block and Bi-GRU. It first extracts speech features via MFCC, enhances feature expressiveness through SE-Res2Block, captures temporal attributes using Bi-GRU, and achieves high-precision emotion classification via pooling and classification algorithms. Moreover, the recognised emotional states are further transformed into adaptive service strategies, enabling intelligent and real-time adjustment of customer interaction. Experimental results demonstrate that ECAPA-BGTDNN achieves 0.706 precision, 0.729 recall, and 0.717 F1-score, outperforming ECAPA-TDNN by 2.0%, 3.2%, and 2.3%, respectively. In practical deployment on a self-constructed customer voice set, the model further attains 0.837 average precision, enabling stable affect tracking during full conversations exceeding 50 seconds.
  Keywords: NLP; customer management; personalised services; affective computing.
  DOI: 10.1504/IJAHUC.2026.10079000
   
  

• Open Access

  Modelling and simulation for biomechanical responses of acupuncture points based on multi-physics field coupling ( Free Full-text Access )

  by Xin Xiang, Dongming Li, Songlin Liu, Hongfeng Wang 
  Abstract: Accurate acupoint localisation is essential for standardised acupuncture therapy and intelligent systems, yet remains challenging due to soft-tissue variability, ambiguous anatomical boundaries, and cross-domain distribution shifts. This study proposes a unified perception framework that integrates multiple complementary mechanisms within a single localisation pipeline. The proposed ACUR-Net incorporates high-resolution feature representation, geometry-aware relational modelling, uncertainty-aware regression, and domain adaptation to address anatomical variability and domain heterogeneity in a coordinated manner. The underlying assumption is that reliable acupoint localisation benefits from the joint modelling of anatomical topology, cross-domain feature alignment, and prediction uncertainty, rather than from isolated architectural modifications. Experiments conducted on the AcuSim-FAcupoint dataset show that the proposed framework achieves improved feature transfer stability and spatial consistency. Comparative evaluations indicate that ACUR-Net outperforms PFLD, PIPNet, HRNet, and ViTPose in terms of NME, PCK@0.05, and OKS-mAP. The results suggest that multi-factor integration is effective for enhancing localisation robustness under realistic conditions.
  Keywords: Acupuncture point recognition; Multiphysics modeling; Cross-domain adaptation; Deep learning.
  DOI: 10.1504/IJAHUC.2026.10076574
   
  

• Open Access

  An optimisation framework for audit decision-making based on deep convolutional neural networks and reinforcement learning ( Free Full-text Access )

  by Bing He, Shan Jiang, Osama Sohaib 
  Abstract: In contemporary auditing and risk management, transaction complexity and high-dimensional audit data challenge accurate risk assessment and efficient resource allocation. Traditional methods relying on manual expertise or static heuristics fail in large-scale, multimodal, dynamic audit scenarios. To solve these, this paper proposes RLMCN-Net, a framework combining multimodal convolutional neural networks and deep reinforcement learning for audit risk assessment and action optimisation. Its risk identification module estimates audit targets' underlying risk from heterogeneous data; the reinforcement learning module determines audit actions considering costs, resource constraints, and expected returns. Risk labels, audit actions, and decision outcomes are modelled separately. Experiments on two benchmark datasets show RLMCN-Net outperforms traditional baselines in risk identification accuracy, audit return, and resource efficiency; ablation studies verify its robustness and generalisation. These results indicate RLMCN-Net effectively supports risk-oriented target screening and dynamic resource allocation, advancing intelligent auditing systems.
  Keywords: audit strategy optimisation; deep reinforcement learning; multimodal feature extraction; risk identification.
  DOI: 10.1504/IJAHUC.2026.10078811
   
  
• Energy Harvesting-Based Routing for Prolonging Sensor Node Lifespan in Underwater Networks Using Stochastic Network Calculus  
  by Vignesh S. R, Rajeev Sukumaran 
  Abstract: Underwater wireless sensor networks (UWSNs) are crucial for marine monitoring, disaster prediction, and underwater surveillance, but their performance is limited by energy constraints and harsh, unpredictable environments. Conventional deterministic models fail to capture underwater randomness, leading to inefficient energy management and unreliable communication. This work proposes an adaptive framework integrating stochastic network calculus (SNC) with piezoelectric energy harvesting (PEH) to enhance energy efficiency and network reliability. The model incorporates pH-based environmental variations within the SNC framework to dynamically regulate energy utilisation. Performance is evaluated against depth-based routing (DBR) and vector-based forwarding (VBF) using metrics such as: packet delivery ratio (PDR), packet loss ratio (PLR), end-to-end delay (E2E), throughput, path loss, and energy consumption. Results show significant improvements in network lifespan, energy efficiency, latency reduction, and data reliability. The proposed probabilistic approach provides a scalable and sustainable solution for real-time, next-generation UWSNs operating in dynamic aquatic environments.
  Keywords: Underwater Wireless Sensor Networks (UWSNs); Stochastic Network Calculus (SNC); Piezoelectric Energy Harvesting; pH-Based Modeling; Network Sustainability.
  DOI: 10.1504/IJAHUC.2025.10076088
   
  
• Wireless Channel Allocation in Smart Building using Cross-Interference Model  
  by Gabriel Galdino, Francisco Cardoso, Rafael Gomes 
  Abstract: IoT paradigm allowed a proliferation of connected wireless devices, that utilize various technologies such as Bluetooth, Zigbee, and Wi-Fi, all operating within the common 2.4GHz frequency band. However, sharing the same frequency poses challenges due to significant interference among these devices, making their coexistence in the same environment quite daunting mainly in smart buildings. Calculating interference between devices on different floors necessitates factoring in signal loss as it passes through ceilings or floors, along with distance-related signal attenuation between the devices. Within this context, this article proposes an interference model that is applied in an algorithm to allocate the channels of the devices in a smart building environment the model proposed aims to minimize the total interference of the environment between devices using different access technologies, improving the performance of the network and the applications that rely on it leading to a suitable expectation of the network capabilities.
  Keywords: Wireless Channel; Smart Building; Cross-Interference; Internet of Things.
  DOI: 10.1504/IJAHUC.2025.10076213
   
  
• Spectrum sensing using Reconfigurable Intelligent Surfaces with Acoustic Energy Harvesting  
  by Faisal Alanazi  
  Abstract: This paper investigates a novel spectrum sensing framework that integrates Reconfigurable Intelligent Surfaces (RIS) and acoustic energy harvesting in a cognitive radio environment. Specifically, a primary user (PU) harvests ambient acoustic energy to power its data transmission to a primary destination. The transmitted signal is reflected by a strategically placed RIS and subsequently received at the secondary user (SU), which performs spectrum sensing using an energy detection technique. The proposed system model captures the interplay between acoustic energy harvesting dynamics, RIS-assisted signal reflection, and the impact on the detection performance of the energy detector at the SU. Analytical expressions for detection probability is derived, taking into account acoustic energy harvesting. Simulation results validate the analytical findings and demonstrate the potential of the proposed architecture to enable efficient and green spectrum sensing in energy-constrained environments.
  Keywords: Spectrum sensing; RIS; acoustic energy harvesting.
  DOI: 10.1504/IJAHUC.2025.10077632
   
  
• DDPG-Based Joint Energy and Offloading Optimisation in UAV-Aided Mobile Edge Computing  
  by Changchun Qin, Yongzhi Ran, Fei Wang, Junwei Luo 
  Abstract: In this paper, we consider an unmanned aerial vehicles (UAV)-aided mobile edge computing (MEC) system, where a fixed-wing UAV provides computation resources for terminal devices (TDs). The UAV can effectively establish line-of-sight communication links with TDs. However, the limitations of energy capacity and transmission coverage of UAV and TDs are still challenges in UAV-aided MEC. The energy capacity affects the service lifetime of the UAV-aided MEC and the transmission coverage has an impact on the quality of service. To address these issues, we study a joint energy and offloading optimisation problem, where we aim to minimise the energy consumption of UAV and maximise the total offloaded data volume of TDs by optimising TDs transmission power and UAVs flight angle and speed. We propose a deep deterministic policy gradient (DDPG)-based algorithm to solve this problem. Simulation results show that our proposed algorithm has good convergence and is better than other algorithms.
  Keywords: Mobile edge computing (MEC); unmanned aerial vehicles (UAV); energy consumption; offloaded data volume; flight trajectory; deep deterministic policy gradient (DDPG).
  DOI: 10.1504/IJAHUC.2025.10077741
   
  
• Age-of-Computing Constrained Fairness-aware Energy Efficient Resource Allocation in MEC-assisted HSRNs  
  by Yeshen Li, Ke Xiong, Zhifei Zhang, Yingying Wu, Pingyi Fan 
  Abstract: This paper investigates the mobile edge computing (MEC)-assisted high-speed railway network (HSRN), where train users tasks can be offloaded to the MEC server deployed at the ground base station (GBS). The age of computing (AoC) is used to measure the timeliness of users task computation and the fairness-aware energy efficiency (FEE) is maximised by jointly optimising the task offloading ratio, channel selection factor, power allocation vector and the computing resource assignment vector at the MEC server subject to the constraints of AoC of the users tasks. To tackle such a non-convex mixed integer nonlinear programming (MINLP) problem, we propose a heterogeneous multi-agent twin delayed deep deterministic policy gradient-based resource allocation (HMATD3-RA) algorithm with a designed FEE-AoC reward function that linearly combines the violation of AoC and the FEE. Simulation results show that HMATD3-RA achieves the highest FEE-AoC reward compared with baselineswhile revealing the trade-off between FEE and AoC.
  Keywords: Mobile Edge Computing; Energy efficiency; Age of Computing; High-Speed Railway Networks; Multi-Agent Reinforcement Learning.
  DOI: 10.1504/IJAHUC.2025.10077876
   
  
• Optimizing Intrusion Detection Systems Using Ensemble Voting Classifiers: A Multi-Dataset Performance Analysis  
  by Sayantan Singha Roy, Amrita Bhadra, Munesh Chandra Trivedi, Awnish Kumar 
  Abstract: As cyberattacks continue to increase in size and complexity, so does the need to protect sensitive data. Ensemble voting classifier has risen as one of the most extensive methods to improve the performance of intrusion detection systems which play a pivotal role in securing networks. The basis of this study is to use six common classifiers [logistic regression (LR), random forest (RF), decision tree (DT), AdaBoost (ADB), Gaussian Naive Bayes (GNB), and K-nearest neighbours (KNN)] on three benchmark datasets (KDD-CUP, CIC-DDoS2019, UNSW-NB15) and compare the classifiers given their final metrics (accuracy, recall, precision and F1-score). This study has explored 62 ensemble configurations to assess their suitability as base classifiers. Results show that ADB and RF ensembles perform best on KDD-CUP, ADB-DT-GNB-RF combinations excel on CIC-DDoS2019, and ADB-RF achieve top results on UNSW-NB15. The study provides practical guidance for developing data-driven IDS solutions and enhances ensemble-based cybersecurity research.
  Keywords: Ensemble Machine Learning; Intrusion Detection Systems ; Ensemble Voting Classifiers; Benchmark Datasets; Comparative Analysis.
  DOI: 10.1504/IJAHUC.2026.10077881
   
  
• A Hybrid Ant Colony Genetic Optimisation Framework for Automated CNN Architecture and Hyperparameter Tuning in Image Classification  
  by Hong-Ren Chen, Mu-Yen Chen, Jia-Lang Xu, Yi-Syuan Wang, Po-Yen Hsu 
  Abstract: Hyperparameter and architectural optimisation remain critical challenges in deep learning, directly affecting model accuracy, generalisation, and efficiency. This study proposes a hybrid ant colony genetic optimisation (ACGO) algorithm for image classification, integrating the pheromone-guided search of ant colony optimisation with the genetic diversity mechanisms of crossover and mutation. A novel encoding scheme for convolutional neural network (CNN) hyperparameters enables automated tuning across datasets of varying complexity. The proposed ACGO is evaluated against five metaheuristic algorithms genetic algorithm, ant colony optimisation, coral reef optimisation, particle swarm optimisation, and Grey Wolf optimiser on CIFAR-10 and CIFAR-100. ACGO achieved 81.1% accuracy on CIFAR-10 and 52.7% on CIFAR-100 (an absolute improvement of approximately 4% and 3% against competing methods), demonstrating strong adaptability and computational efficiency. These results highlight ACGOs potential as a scalable, robust optimisation framework for deep learning-based image classification.
  Keywords: metaheuristic algorithms; hyperparameter optimization; image classification; model encoding; convolutional neural networks.
  DOI: 10.1504/IJAHUC.2026.10078301
   
  
• Resource Bargaining and Server Selection Framework in Multi-Access Edge Computing Platform  
  by Sungwook Kim 
  Abstract: In this study, we consider both the allocation of different types of resources in heterogeneous MEC servers and server selection. To solve these two control problems, we formulate a joint control framework based on the interactive mechanism. For the multi-resource allocation problem, we design the new compromise bargaining solution (CBS) to distribute different type resources. By combining well-known bargaining solutions, the CBS can provide a comprised consensus solution in a fair-efficient manner. To address the server selection problem, the novel multi-criteria selection (MCS) method is developed. Based on the combination of ideal solution and weighted average, the MCS method selects an appropriate MEC server for each device. In the heterogeneous MEC dynamics, the key insight of our proposed approach is to provide a collaborative control paradigm through the interaction of CBS and MCS processes
  Keywords: Multi-access edge computing; Multi-resource allocation problem; Compromise bargaining solutions; Multi-criteria selection; Cooperative game theory.
  DOI: 10.1504/IJAHUC.2025.10078326
   
  
• Path Planning of Multi Drone System for Disaster Analysis and Management  
  by Abhijeet Pandey, Bhavi Khator, J. Sathish Kumar 
  Abstract: Disasters often cause unpredictable damage that complicates emergency response efforts. While drones (UAVs) offer promise for accessing hard-to-reach areas, current systems lack robust autonomous navigation and 3D situational awareness needed for efficient disaster management. This paper presents a novel approach using multiple drones to collaboratively generate detailed 3D models and employs an advanced deep learning-based obstacle detection and 3D path-planning system. Extending traditional 2D algorithms into 3D, our method integrates new cost functions to optimise drone swarm navigation considering environmental and operational constraints. Using pre-disaster 3D maps and real-time data, the system enables faster, safer, and more efficient search and rescue operations. Validated in AirSim simulations, this framework significantly enhances UAV autonomy and coordination, addressing key limitations in current disaster response technologies and improving both pre- and post-disaster management strategies.
  Keywords: Path Planning; Disaster Management; Unmanned Aerial Vehicles.
  DOI: 10.1504/IJAHUC.2026.10078440
   
  
• A Federated Blockchain Framework for Secure and Scalable Remote Patient Monitoring Using IoT Devices  
  by Sai Dheeraj Peketi, Pranav Sanand Puligandla, S.S.R. Sri Harsha Kedarisetty, Praveen Raj Kanickairaj, J. Sathish Kumar 
  Abstract: The increasing demand for secure, efficient, and scalable healthcare solutions has led to the integration of advanced technologies such as blockchain, IoT, and AI in Remote Patient Monitoring (RPM). Traditional RPM systems suffer from security vulnerabilities, lack of interoperability, and limited scalability. In this paper, we propose a novel Federated Blockchain-based RPM system leveraging Hyperledger Fabric, designed to provide decentralized, secure, and efficient patient data management. Our approach ensures data integrity, privacy, and controlled accessibility through smart contract automation. Comparative performance analysis highlights substantial improvements in transaction throughput, latency reduction, and data reliability. Furthermore, we discuss practical deployment challenges, regulatory compliance, and interoperability with EHRs and the Internet of Medical Things (IoMT). This work aims to revolutionize remote healthcare by delivering an intelligent, secure, and patient-centric approach, fostering innovation in real-time health monitoring and predictive medical interventions.
  Keywords: Remote Patient Monitoring (RPM); Federated Blockchain; Hyperledger Fabric; Decentralized Data Management; Electronic Health Records (EHRs).
  DOI: 10.1504/IJAHUC.2026.10078635
   
  
• Edge-Cloud Integrated Fraud Detection for E-Commerce Transactions Using Hybrid Siamese-LSTM Model  
  by Winner Pulakhandam, Vallu Visrutatma Rao, Pramod Begur Nagaraj, Archana Chaluvadi, Priyan Malarvizhi Kumar 
  Abstract: E-commerce systems are increasingly vulnerable to fraud, requiring efficient detection methods capable of handling large-scale transaction data. However, existing approaches struggle to adapt to evolving fraud patterns and often incur high computational costs. This study proposes a novel edge-cloud integrated fraud detection framework using a hybrid Siamese neural network with long short-term memory (Siamese-LSTM). The e-commerce transaction dataset is pre-processed using K-nearest neighbours for data imputation and Tukeys Fences for outlier removal. The processed data is analysed in an edge-cloud environment, where behavioural patterns and their frequencies are extracted at the edge. Additional indicators, such as unusual transactions and login location mismatches, are forwarded to the cloud for further analysis. The Siamese-LSTM model detects deviations from normal spending behaviour to identify fraudulent transactions. Experimental results demonstrate high performance, achieving 99.47% accuracy, 99.55% precision, and 99.38% recall, while ensuring efficiency and scalability
  Keywords: Fraud Detection; E-Commerce; Edge-Cloud Integration; Deep Learning; Siamese Neural Network; Long Short-Term Memory.
  DOI: 10.1504/IJAHUC.2026.10078656
   
  
• Efficient Edge Resource Allocation Using Deep Neural Network-XGBoost for Distributed Computing in Edge Networks  
  by Venkataramesh Induru, Karthik Kushala, Vijai Anand Ramar, Priyadarshini Radhakrishnan, Premalatha R 
  Abstract: The proposed Deep Neural Network (DNN) and XGBoost combined with Pufferfish Optimization Algorithm (POA) improve the performance of edge computing substantially with better outcomes on the most important metrics. In terms of latency (22 ms), resource usage (95%), and energy efficiency (75 J), the model performs exceptionally well, finishing jobs in 2.8 seconds with 98% accuracy. With 96% performance and 98% adaptability, it is perfect for power-constrained, dynamic IoT and edge computing scenarios that need scalability, accuracy, and speed.This total performance reflects the model's promise in making edge computing a more efficient, effective, and sustainable platform.
  Keywords: Edge Computing; Resource Allocation; IoT; Puffer fish optimization; Deep Neural network; XG Boost.
  DOI: 10.1504/IJAHUC.2026.10078661
   
  
• Building Resilient AI Architecture with MLOps, Chaos Engineering, and SRE for Scalable and Reliable ML Deployments  
  by Pramod Begur Nagaraj, Archana Chaluvadi, Winner Pulakhandam, Vallu Visrutatma Rao, Padmavathy R 
  Abstract: Machine Learning Operations (MLOps) has proved an important area for developing models, deploying them, and monitoring them in machine learning (ML) production environments. ML systems face challenges in scalability and reliability. This work introduces MLOps to optimize models, with Chaos Engineering and Site Reliability Engineering (SRE) ensuring resilience. Data from e-commerce, healthcare, and finance is processed using Matrix Factorization Imputation for missing values and Quantile Clipping for outliers. Features are derived via Quantile Binning and Feast. Lasso Regression aids feature selection, Quantile Regression Forests capture non-linearities, and Evidently AI monitors drift, while SRE ensures reliable deployment.The experimental findings indicate that the ML model works very well across many industries, such as E-Commerce, Healthcare and Finance, obtaining high levels of predictive accuracy and low levels of error metrics. This assists in developing machine learning systems that become more scalable and continuously improve efficiency throughout a wide range of applications.
  Keywords: Machine Learning Operations; Chaos Engineering; Site Reliability Engineering; Machine Learning; Observability; Automation; E-Commerce; Healthcare; and Finance.
  DOI: 10.1504/IJAHUC.2026.10078670
   
  
• Optimising Job Scheduling in Apache Spark: A Simulation Framework for Algorithm Comparison and Performance Prediction  
  by Vishnu Prasad Verma, Santosh Kumar, Srinivas Naik Nenavath 
  Abstract: Selecting the best scheduling algorithm is essential for optimising Apache Sparks performance and resource usage. This study conducts a comparative analysis of different scheduling algorithms, including first in, first out (FIFO), fair, earliest deadline first (EDF), round robin, shortest job next, least laxity first, priority scheduler, and multilevel feedback queue using a Python-based simulation framework. The analysis focuses on critical performance metrics such as turnaround time, waiting time, deadline adherence, and violation rates. Our findings highlight notable performance differences among scheduling algorithms when tested under various workload conditions. Among them, the multilevel feedback queue (MLFQ) consistently emerged as the top performer, achieving the lowest turnaround and waiting times, a remarkable earliest arrival ratio of 97.3%, and no deadline violations. The experimental findings guide researchers, cloud infrastructure designers, and Spark system architects in successfully choosing and refining scheduling techniques to manage various dynamic workload needs
  Keywords: Distributed Computing; Spark job scheduling; Big data; Scheduling algorithms; resource management; optimization.
  DOI: 10.1504/IJAHUC.2025.10078686
   
  
• Generative Artificial Intelligence in Ubiquitous IoT Systems: Advances, Applications and Challenges  
  by Yang Li, Leifeng Wei, Jing Nie, Joel Rodrigues, Muhammad Attique Khan, Margo Sulistio, Hua-Tsung Lin 
  Abstract: Ubiquitous IoT systems generate massive multi-modal data, yet transforming this deluge into actionable intelligence for real-time decision-making remains a fundamental challenge. The convergence of Generative AI and Large Language Models offers a transformative paradigm. This review proposes a generic cognitive orchestration architecture wherein the LLM functions as a central reasoning engine, composing specialised GAI models for simulation and content generation while mediating natural human-system interaction. We demonstrate this domain-agnostic framework across smart manufacturing, transportation, healthcare, built environments, and agriculture. Critical challenges persist, including data scarcity and heterogeneity, model hallucination and computational cost, and systemic ethical concerns including accountability, privacy, and the digital divide. Priority research directions include retrieval-augmented generation, domain-adapted foundation models, edge-native efficiency, and value-aligned governance. Realising dependable, equitable GAI-LLM systems is essential for the coming generation of truly intelligent ubiquitous computing environments.
  Keywords: Generative AI; Large Language Models; Ubiquitous computing; IoT intelligence; Edge AI; Cognitive IoT.
  DOI: 10.1504/IJAHUC.2026.10078834
   
  
• Query-Response generation system for reliable responses to agriculture-related queries  
  by Y.Y. Narayana Reddy, Thulasiram Narayanan, Adusumalli Balaji 
  Abstract: A smart deep learning based agricultural query response generation system that can provide trustworthy and semantically accurate answers in this research. The proposed system introduces two key innovations. Initially, a term inverse added Word2Vector (TIA-W2V) embedding model is developed to improve semantic representation by combining term importance with contextual embeddings. Second, a Cosine similarity enclosed triplet long shortterm memory (CT-LSTM) network is proposed to learn complex semantic relationships and predict responses. This enhances the model's ability to focus on pertinent parts of the query and response, which improves context comprehension and response generation. By effectively managing long sequences, understanding context, and capturing semantic connections, proposed framework can assist in producing more precise and relevant responses to questions about agriculture. The simulation results show that the proposed method performs better in terms of accuracy (97.9%), precision (96.58%), recall (96.12%), F1-score (96.98%), Query response time (15sec), BLEU (7.98%) and METEOR (8.59%), respectively.
  Keywords: Deep learning; Query-Response generation system; pre-processing; Term Inverse Added Word2vector; Cosine similarity enclosed triplet long short-term memory; text mining models.
  DOI: 10.1504/IJAHUC.2026.10078838
   
  
• An Efficient Data-Collection Routing Algorithm to Improve Energy Efficiency and Network Lifetime in Wireless Sensor Networks  
  by Hanif Zafor, Tasher Ali Sheikh, Amitava Nag 
  Abstract: In WSN, data transmission has a serious issue known as a hotspot problem and can be eliminated by deploying a mobile sink (MS) device. In Ms also there are some serious problems like battery capacity of MS, routing problem of MS, and time window for data collection. To improve network lifetime and minimize energy consumption, a hybrid variable neighborhood search with local search (VNS-LS) routing algorithm is proposed for data collection using a mobile data collector (MDC). The simulation results were performed based on the parameter matrices: network throughput (NT), energy consumption (EC), data collection time (DCT), and packet loss ratio (PLR). The results are then compared with some existing methods, the two-phase energy minimized load balancing scheme (TPEMLB), the heuristic tour planning algorithm (HTPA), and the gravitational search algorithm (GSA). The proposed method shows better results with higher NT, minimum EC, DCT, and PLR than the existing methods.
  Keywords: Wireless Sensor Network (WSN); Variable Neighbourhood Search (VNS); Local Search (LS); Data Collection; Energy Consumption (EC); Network Lifetime (NL).
  DOI: 10.1504/IJAHUC.2025.10078839
   
  
• KERSC: A Knowledge-Enhanced Reasoning Semantic Communication Framework for Text Transmission  
  by Lexuan Wang, Bo Shen 
  Abstract: Recently, deep learning-based semantic communication has attracted considerable attention. However, most existing approaches assume shared prior knowledge between transmitter and receiver, overlooking the role of external knowledge in semantic reasoning. To address this limitation, this paper proposes a Knowledge-Enhanced Reasoning Semantic Communication (KERSC) framework for text transmission, focusing on improving semantic reconstruction at the receiver. By modeling the semantic and structural features of external knowledge, we construct a semantically smooth embedding space to support reasoning. An effective interaction mechanism is introduced, comprising a contrastive learning module for precise knowledge alignment and a two-layer relation-aware graph attention network for structured reasoning. Finally, a knowledge-enhanced decoder with an adaptive gating mechanism fuses the transmitted semantics with knowledge representations to boost performance. Simulation results demonstrate that the proposed model effectively enhances semantic reconstruction accuracy under various channel conditions.
  Keywords: Semantic Communication; Knowledge-enhanced Reasoning; Knowledge Graph; Knowledge Representation Learning; Deep Learning.
  DOI: 10.1504/IJAHUC.2026.10078982
   
  
• PPTFA-SC: Privacy Preserving Three Factor Authentication scheme for Smart City IoT Network  
  by Vinod Mahor, R. Padmavathy, Santanu Chatterjee 
  Abstract: Application of IoT in the field of smart city is resulting in revolutionized quality of life, efficient public services and digitally managed infrastructure The concept of smart city is realized by deploying IoT enabled sensor and actuator nodes in the field of interest These nodes then operates autonomously to collect data from its surrounding environment and logs to the gateway node for analysis and decision making The gateway node can connect to internet which enables remote data access to the users Data can be accessed from gateway node or from specific node Ensuring data security and genuinity of accessed information is crucial for development of resilient infrastructure and efficient delivery of public services. This necessitates the strong entity authentication mechanism to prevent adversarial attacks In this paper, we propose novel and secure authentication scheme to address security challenges in smart applications and perform its security analysis and simulation.
  Keywords: Authentication; Security; WSN; IoT; Smart City; AVISPA; Privacy.
  DOI: 10.1504/IJAHUC.2026.10079026