International Journal of Reasoning-based Intelligent Systems (46 papers in press)

Regular Issues

• Open Access

  Immersive virtual reality art generation via neural style transfer ( Free Full-text Access )

  by Wei Pu 
  Abstract: Stylised imagery in VR must meet comfort budgets while remaining legible. To address this challenge, a real-time pipeline is presented. First, a feed-forward multi-style engine applies post-lighting stylisation while head-locked interface layers are excluded. Then, temporal and binocular stabilisers curb flicker and rivalry. Finally, a render contract with GPU-resident textures, mixed precision, and deterministic fallback keeps latency under control. Experiments on three scenes sustain 75 FPS, about 15 ms per frame, and motion to photon of 1923 ms on an RTX-class device; a fast baseline reaches 52 FPS at 22 ms and an iterative one 6 FPS at 140 ms. Image fidelity remains high, temporal flicker drops by roughly 31%, and stereo divergence falls to 0.012 with ninety-fifth percentile frame time near 17 ms.
  Keywords: neural style transfer; NST; immersive virtual reality; real-time rendering; temporal consistency.
  DOI: 10.1504/IJRIS.2026.10076376
   
  

• Open Access

  Integrated design of performance-oriented cost accounting in colleges and universities driven by artificial intelligence ( Free Full-text Access )

  by Xiaona Cui 
  Abstract: To resolve disconnected performance assessment and cost accounting systems, and inefficient resource allocation in colleges and universities, this study constructs a performance-oriented integrated cost accounting management system using artificial intelligence technology. The research first combines the functions of teaching, scientific research and management positions, extracts representative and quantifiable performance indicators, and forms a multi-dimensional indicator system. Based on data standardisation and feature engineering processing, the random forest model is introduced to predict the performance score, and the model effect is verified through residual analysis and variable importance ranking. The system links the performance output results with the job cost accounting data, calculates the unit performance output cost, and builds a closed-loop management process of accounting prediction feedback. The results show that the artificial intelligence model performs stably in identifying performance differences and optimising cost structures, and the system has good scalability and management decision support capabilities.
  Keywords: artificial intelligence; AI; performance-oriented cost accounting integrated system; university management.
  DOI: 10.1504/IJRIS.2026.10076377
   
  

• Open Access

  Transformer-GNN hybrid architecture for optimising real-time traffic forecasting on highways ( Free Full-text Access )

  by Hua Cheng, Yupeng Cao, Weiping Li 
  Abstract: Facing the challenge of worsening highway traffic congestion, precise real-time forecasting is crucial for intelligent traffic management. However, traditional models struggle to effectively capture the complex spatio-temporal dependencies and dynamic propagation delays inherent in traffic data. To address this, this paper proposes a hybrid architecture that integrates graph neural networks with transformers. Through a dynamic graph attention mechanism and a delay-aware module, it significantly enhances the modelling capabilities for long-range spatial correlations and temporal propagation effects. Experiments on public datasets such as performance measurement system 04 and performance measurement system 08 demonstrate that the proposed model reduces the mean absolute error by 6.2%9.2% compared to existing state-of-the-art methods within the 1560 minute prediction window, with particularly notable performance improvements during peak congestion periods. The framework presented here has the potential to provide a more reliable technical pathway for traffic state prediction, holding significant practical application value.
  Keywords: traffic flow prediction; graph neural networks; GNNs; transformers; intelligent transportation systems.
  DOI: 10.1504/IJRIS.2026.10076378
   
  

• Open Access

  Legal loophole detection model based on multi-agent reinforcement learning ( Free Full-text Access )

  by Dongli Ma 
  Abstract: Precise detection of legal loopholes is a critical component in upholding judicial fairness. However, existing methods exhibit significant shortcomings in deeply integrating legal knowledge, while their modelling capabilities for imbalanced data remain underdeveloped. To address this, this paper first designs a hierarchical experience replay mechanism. By storing and sampling experiences through temporal difference errors and task priorities, it effectively controls gradient conflicts during cross-task training. Second, a legal loophole detection model based on improved multi-agent reinforcement learning is designed. The knowledge fusion module maps multi-source legal knowledge into a unified representation space, achieving selective knowledge enhancement through a learnable knowledge gating mechanism. Furthermore, an adaptive feature space partitioning is realised through the collaborative mechanism of multiple agents across multi-classification tasks, significantly improving the recognition performance of minority class samples. Experimental results demonstrate that the proposed model achieves a loophole detection accuracy of 92.48%, significantly enhancing detection precision.
  Keywords: legal loophole detection; multi-agent reinforcement learning; hierarchical replay pool; attention mechanism; multi-task training.
  DOI: 10.1504/IJRIS.2026.10076379
   
  

• Open Access

  Optimised design of commercial building interior spaces and landscapes using enhanced ResNet and virtual reality technology ( Free Full-text Access )

  by Ni Yin 
  Abstract: To address the limitations of traditional commercial building design, this study proposes an optimisation method integrating an improved ResNet model with VR technology. The enhanced ResNet50 model, incorporating a CBAM attention mechanism and multi-scale feature fusion, quantifies key metrics like spatial comfort and traffic efficiency, achieving 94.7% feature recognition accuracy. A VR platform then generates interactive scenes based on model outputs to collect real-time user feedback. Experiments on three commercial buildings showed that after optimisation, space utilisation increased by 15.2%, landscape satisfaction rose by 21.6%, and traffic efficiency improved by 18.9%. This data-driven, user-centred approach offers significant practical and academic value for commercial design.
  Keywords: improving ResNet; virtual reality technology; commercial buildings; indoor space; optimisation landscape design.
  DOI: 10.1504/IJRIS.2026.10076628
   
  

• Open Access

  Dynamic cost control methods for computer-integrated engineering based on digital twins ( Free Full-text Access )

  by Chenchen Lai, Sheng Lin, Zhongyuan Chen 
  Abstract: This paper presents a digital twin-based dynamic cost control model for computer-integrated engineering projects. Built upon BIM component modelling, the proposed approach establishes a synchronised virtual-physical interaction framework to support real-time cost monitoring and adaptive regulation. The novelty of this work lies in three aspects: a bidirectional mapping mechanism between BIM components and construction states, a multidimensional cost analysis model integrating time-series and spatial distribution features, and a rule-engine-driven closed-loop feedback mechanism for cost deviation detection and control. Experimental results demonstrate that the proposed model achieves high estimation accuracy across key construction processes, with cost fluctuation margins of 2.8% during concrete pouring and 1.6% during reinforcement operations. The results confirm the effectiveness of real-time feedback in identifying high-risk processes and improving dynamic cost control performance.
  Keywords: digital twin; BIM; rule engine; sensor network; real-time feedback; data fusion; cost evolution modelling.
  DOI: 10.1504/IJRIS.2026.10076891
   
  
• Optimizing Feature Selection in Educational Data Sets Using an Enhanced Teaching-Learning Based Optimization Algorithm  
  by George Amalarethinam, A. Emima 
  Abstract: Educational data mining (EDM) is an emerging study topic that helps schools improve student performance. Selecting only relevant data reduces model input parameters with feature selection. It reduces dimensionality by selecting a subset of features and removing incorrect, superfluous, or noisy ones. It improves learning accuracy, computational cost, and model interpretability. This impacts the accuracy of performance models used to assess student outcomes. Most optimisation methods, including the genetic algorithm, must optimise many governing parameters for greater performance. Optimisation approaches using wrapper feature selection (WFS) improve classifier prediction. The proposed ETLBO algorithm with WFS techniques uses the Euclidean distance formula to assess fitness value and popular control parameters to select the optimal feature subset. The algorithm above is used on the educational dataset. Classification algorithms evaluate the best features from TLBO ETLBO algorithms: 4 algorithms classify performance metrics: GNB, LR, SVM, and K-nearest neighbour. Experimental results suggest that the ELTBO algorithm’s best feature subset improves classification accuracy for GNB, LR, SVM, and KNN compared to TLBO.
  Keywords: classification algorithms; feature selection; FS; optimisation technique; Euclidean distance; enhanced teacher learner based optimisation; ETLBO; teacher learner based optimisation.
  DOI: 10.1504/IJRIS.2024.10068107
   
  
• Advancing Healthcare Intelligent Systems: The Critical Role of Paternity Benefits in Modern Caregiving  
  by Swapna Ashmi, P.R.L. Rajavenkatesan 
  Abstract: The Maternity Benefit Act of 1961 ensures that women are entitled to receive payment for maternity leave and leave in the event of a miscarriage. It is also important to note that Indian law has not appropriately recognised paternity leave. Fathers can take paternity leave after the birth of their child or after miscarriage, adoption, or similar circumstances. The legislation regarding paternity leave in India was officially passed in 2017. The execution of this law needs improvement, as dads’ paternity leave rights are not regulated. Gender-neutral policy guidelines matter in a global economy. Fathers’ contributions to their spouses and children’s well-being make paid parental leave crucial. The study examines how paternity benefits affect children’s development and growth. The study also compared India’s paternity leave policy to many others. Healthcare analysis and kid well-being are also examined. It was given to 317 people from diverse fields. The study evaluates the importance of paternity benefit enforcement in India based on 250-member replies. MS Office was used to draft and organise the research, while Python was used to process and compare data.
  Keywords: childcare; advancing healthcare; intelligent systems; equality and fatherhood; gender-neutral; maternity leave; miscarriage and paternity leave.
  DOI: 10.1504/IJRIS.2024.10068108
   
  
• Leveraging Social Capital and SIoT for Sustainable Entrepreneurship Development  
  by K.M. Ashifa, Mehdi Safaei, HINA Zahoor, Rehab El Gamil, NASIR MUSTAFA 
  Abstract: The current research examines the combined effect of integrating social internet of things technology in entrepreneurial skill development programs for the Irula tribal community, Tamil Nadu, toward socio-economic upliftment. LAS and SCAM were adopted to collect data at the household level of 538 households, besides gathering qualitative information through purposive collection through focused group discussion and an in-depth interview of 60 participants. Quantitative results, as shown by paired t-tests and CR analyses, recorded significant increases in social capital and entrepreneurial skills following intervention. In-depth interviews, FGDs, and workshops brought rich qualitative insights into improved networking, innovation, and decision-making. Increasing communities’ cohesion and resilience resulted in enhanced livelihood
  Keywords: social internet of things; SioT; tribal development; indigenous knowledge; community health; entrepreneurial skills; government interventions; livelihood assessment schedule; LAS.
  DOI: 10.1504/IJRIS.2024.10068109
   
  
• Ancient Epigraphical Monuments' Convolution Neural Network-Based Skeletonized Structural Angularmorphing Character Identification Intelligent Systems  
  by P. Selvakumar 
  Abstract: Tamil is one of the oldest languages, and it is based on several proofs from ancient Kiladhi epigraphic monuments. Tamil texts have various structural styles and projections identified from monuments like palm lead characters, vattezhuthu, and stone inscriptions. By projecting Tamil characters in various angles, the text style may vary due to structural representation, leading the actual character style to differentiate from the old style. Thus, recognition of the specific projection of the old character leads to more features on the dimension level to get the Tamil character and classification. Consider skeletonized structural angular morphing (S2AM) based on a CNN-identified Tamil character from ancient epigraphic monuments for optimum identification. Epigrammatic images will be pre-processed using Gaussian filters, then SMS will glide the character region using CED. Use the skeletonized angular projection to discover text structural components and extract angular information. The selected features will be trained with a DFCNN to
  Keywords: Script Systems Identifying; Tamil character detection; edge detection skeletonization; Character Identification Intelligent Systems; Canny Edge Detection (CED); Deep Features Convolution Neural Networ.
  DOI: 10.1504/IJRIS.2025.10068690
   
  
• Enhancing Critical Thinking Skills through Generative AI Models: Mechanisms and Educational Impacts  
  by Vincent Raj, Eronimus Jeslin Renjith, S.Silvia Priscila, C.Sathish Kumar, S. Suman Rajest 
  Abstract: Enhancing Critical Thinking Skills has even been considered to revolutionize the future of artificial intelligence and has such huge impacts across sectors, especially education. This study looks at the way generative AI models enhance critical thinking in learners based on recent studies; it debates applying them to actual education and their influence. The study adopted a mixed-method approach. One would be to carry out an analysis of learners’ performance using a quantitative method and also obtain a subjective assessment of the development of learners’ critical thinking skills. The results show that generative AI improves scholastic performance; personalised learning tools raise critical thinking scores from 50-80 to 70-100. The conclusion shows that models increase engagement and positive attitudes towards enhanced learning outcomes. AI integration into higher education faces various hurdles, including privacy issues over higher education data and educator training. The study also discusses how educators and governments might use focused
  Keywords: Generative AI; Critical Thinking; Educational Technology; Personalized Learning; Interactive Learning; AI in Education; Pedagogical Strategies; Cognitive Development.
  DOI: 10.1504/IJRIS.2025.10068827
   
  
• Intelligent Techniques for Evaluating Organizational Agility via Contingency Theory in Dynamic Environments  
  by Sivakoti Reddy Manukonda, Seema Bhakuni, Vinayak Anil Bhat, Rameshwaran Byloppilly, Rishi Shukla, Jayesh Solanki 
  Abstract: The contingency approach to management holds that effective management depends on the context of a situation. This paper discusses how the contingency approach works in different managerial settings, focussing on environmental variables like organisational size, task structure, and leadership style that may affect manager effectiveness. This study will combine an in-depth literature review, interview and survey data, and statistical analysis to understand the contingency strategy in practice. Management approaches combined with these environmental variables improved organisational performance, adaptation, and resilience. With organised contingency practices, larger organisations perform better, adaptive management tactics help difficult tasks, and democratic leadership styles work better in varied situations. Regression models and the coefficient of correlation show strong positive correlations supporting these links. Since the contingency technique usually requires ongoing adjustment and integration of various factors into a system, it also highlights its obstacles and limitations from the study. Long-term contingency practices and their effects or
  Keywords: Contingency Approach; Managerial Effectiveness; Organizational Performance; Leadership Style; Environmental Variables; Task Structure; Organizational Size; Adaptability.
  DOI: 10.1504/IJRIS.2025.10068931
   
  
• Reinforcement Learning-Driven Collective Intelligence for Prioritized Spectrum Reservation in Cognitive Radio Network  
  by Meetu Nag, Bhanu Pratap 
  Abstract: In the realm of cognitive radio networks, research aims to enhance spectrum usage by enabling access for more users through different spectrum allocation policies. The dynamic and rapid changes in the communication environment pose challenges in making correct decision for spectrum allocation. To facilitate dynamic spectrum allocation, intelligence is integrated into the cognitive system to analyze environmental parameters, various known parameters have to be analyzed to know about the nature of the radio node. In this paper a novel method is discussed for spectrum allocation by involving a multiple decision system that works on priority-based allocation approach. This system collects environmental data for decision-making, ensuring efficient service in this adaptive communication scenario.
  Keywords: Reinforcement Learning; Collective Intelligence; Spectrum Reservation in Cognitive Radio Network; Spectrum Sensing; Cognitive Radio Network.
  DOI: 10.1504/IJRIS.2025.10069324
   
  
• Predicting Chronic Obstructive Pulmonary Disease (COPD) using Machine Learning with Bio-Inspired Hyperparameter Optimization  
  by Yalin Song 
  Abstract: Chronic obstructive pulmonary disease (COPD) is a prevalent respiratory condition for which early detection is crucial to effective patient management. With LGBM and DTC as the foundational models, this study explores the predictive capability of ML approaches for COPD. Two bio-inspired optimizers, the TSA and ROA, were employed to enhance their performance. These optimizers mimic the collective behavior of biological systems, such as tunicates’ foraging patterns and jellyfish’s pulsating movements, to achieve optimal solutions within the model training process. Relevant features are extracted from patient data, potentially including demographics, medical history, lung function tests, and lifestyle factors. Among the metrics used to evaluate the performance of the optimized models are their accuracy and precision. The DTTS model’s excellent performance shows how well the DTC model predicts COPD. The greatest accuracy and precision scores of 0.907 and 0.911 support its COPD prediction accuracy. These findings demonstrate the DTTS model’s reliability and
  Keywords: Chronic Obstructive Pulmonary; Decision Tree Classification (DTC); Light Gradient Boosting Classification (LGBM); Rhizostoma Optimization Algorithm (ROA); Tunicate Swarm Algorithm (TSA); Machine Learn.
  DOI: 10.1504/IJRIS.2025.10070424
   
  
• Efficient deep mood-based Hindustani raga music recommendation using facial emotion expressions  
  by Yogesh Prabhakar Pingle, Lakshmappa K. Ragha 
  Abstract: Music recommendation is considered as a solution, and the performance is degraded with prediction error. A novel approach for music recommendation based on facial emotions with the objective of extracting better feature information without loss is required. In this paper, an efficient cross-dense network model with multi-pooling is used to detect basic emotions from the face image. The complex cross-dense connections are provided for the extraction of most discriminate feature information. After recognising the emotion from the face, a new attention-based deep collaborative filtering recommendation system is proposed, with a list of Hindustani raga music to improve users moods. The proposed framework is invoked with the Facial Expression Recognition 2013 (FER-2013) dataset, and the recommendation is provided for happy and sad emotions from the ragas. The performance is compared with existing deep learning-based approaches. The proposed approach improves accuracy, precision and recall by 0.9972, 0.9896, and 0.9906.
  Keywords: facial emotion recognition; CrossDenseNet; multi-pooling layer; AttentionNet; collaborative recommendation; Hindustani music.
  DOI: 10.1504/IJRIS.2025.10070903
   
  
• Enhancing academic success: a deep dive into students' performance prediction using decision tree classification models  
  by Tingting Du, Linglanxuan Kong 
  Abstract: Education, a fundamental human right, plays a pivotal role in personal and societal advancement, cultivating critical thinking and problem-solving skills, fostering social integration, and contributing to global progress, with a focus on innovative strategies to elevate education standards and prioritise students' performance. Educational data mining (EDM) is a burgeoning field within DM that investigates patterns in education, covering analysis of student knowledge and behaviour, teacher curriculum planning, and course scheduling, all with the primary goal of enhancing student learning performance and achieving efficiency in education systems. This paper addresses the task of predicting and categorising students' performance in the Portuguese language, emphasising decision tree classification (DTC) models, along with 2 hybrid models optimised using aquila optimiser (AO) and honey badger algorithm (HBA), for a cohort of 649 students. The results underscore the exceptional predictive capabilities of the DTHB model, outperforming the DTAO model in G2 prediction with an impressive F1-score of 0.9428 compared to 0.9381. Additionally, the DTHB model continues to excel in G3 prediction, boasting the best performance at an F1-score of 0.9275.
  Keywords: Student performance; decision tree; aquila optimiser; AO; honey badger algorithm; HBA; teacher curriculum planning; educational data mining; EDM; course scheduling; decision tree classification; DTC.
  DOI: 10.1504/IJRIS.2025.10072219
   
  
• Comprehensive study on digital image encryption using magic square  
  by Vybhavi. Balasundar, K. Mani, Uma Devi, S.Kumar Chandar 
  Abstract: Digital image encryption plays a vital role in safeguarding sensitive images from unauthorised access. Among the emerging methodologies, magic square-based encryption has gained significant attention due to its simplicity, flexibility, and capacity to generate diverse encryption keys. This review provides a detailed analysis of magic square-based techniques for image encryption, emphasising their unique properties and applications. The paper examines several recent algorithms, exploring their design, strengths, and limitations. Furthermore, it highlights the potential of hybrid encryption approaches that integrate magic square techniques with other cryptographic methods to enhance security and efficiency. Finally, the review discusses the current advancements in magic square-based image encryption and identifies key challenges, such as scalability, robustness, and adaptability, clearing the path for additional study and advancement in this area.
  Keywords: digital image encryption; magic square methodology; hybrid encryption; algorithms; techniques.
  DOI: 10.1504/IJRIS.2025.10072799
   
  
• A novel method for solving probabilistic programming problem in interval type-2 fuzzy environment  
  by Babita Chaini, Narmada Ranarahu 
  Abstract: This paper introduces a novel mathematical model for stochastic programming in a type-2 fuzzy environment, addressing the dual uncertainties of fuzziness and randomness through fuzzy normal random variables. The proposed model innovatively converts fuzzy stochastic problems into deterministic ones using a two-step process: the -cut technique to remove fuzziness and the chance-constrained technique to handle randomness. This approach, involving perfectly normal interval type-2 triangular fuzzy numbers, is illustrated with a numerical example. The critical finding is the effective transformation of complex fuzzy stochastic problems into more manageable deterministic forms, enhancing computational efficiency and solution accuracy. The industrial implications are significant, offering a robust decision-making framework for sectors like manufacturing, logistics, and finance, where uncertainty is a critical factor. This methodology improves accuracy and reliability in operational and strategic planning, making it highly relevant for practical applications.
  Keywords: stochastic programming; normal random variables; optimisation techniques; type-2 fuzzy set; T2FS.
  DOI: 10.1504/IJRIS.2025.10073067
   
  
• Multi-scale semantic awareness fusion transformer for sentiment analysis in electricity marketing  
  by Chunlei Liu, Wei Ge, Yanan Cai, Jinghui Chen 
  Abstract: In the context of electricity market marketing, facial recognition-based emotion analysis systems can help enterprises better understand customers emotional feedback, thereby enhancing service experience and improving the precision of marketing strategies. To address these challenges, this paper proposes a multiscale semantic perception and attention fusion model (MSPAF) aimed at improving the accuracy and robustness of customer emotion recognition in the power industry. During the multimodal feature fusion stage, the model applies a multi-level attention pooling strategy to effectively capture emotional correlations between different modalities while reducing feature dimensionality, thereby improving efficiency and generalization. When using generic image encoding features combined with global semantics and local syntax fusion, the models accuracy drops by 1.64% and 2.34%, respectively.
  Keywords: multiscale semantic awareness; transformer; electricity marketing; sentiment analysis; multihead attention mechanism.
  DOI: 10.1504/IJRIS.2025.10073068
   
  
• Cognitive driving: harnessing machine learning to understand driver behaviour  
  by Deepika Arunachalavel, Pandeeswari Nagarajan 
  Abstract: This study presents an innovative approach to enhancing road safety and optimising transportation efficiency by leveraging advanced machine learning techniques. The primary focus is on analysing telematics and sensor data collected from vehicles to model, predict, and classify various aspects of driver behaviour. By utilising a combination of supervised and unsupervised learning methods, the research aims to develop a robust, real-time system capable of detecting patterns associated with safe, aggressive, and distracted driving. Supervised learning techniques are employed to train classification models using a diverse set of features extracted from telematics data, including speed variations, acceleration and braking patterns, steering behaviours, lane discipline, and spatial-temporal characteristics. Emphasis is placed on model interpretability to ensure transparency, reliability, and trust in real-world applications, especially for law enforcement and insurance industries. Additionally, unsupervised learning approaches, such as anomaly detection, are explored to identify deviations from normal driving behaviour without relying on predefined labels. By integrating these techniques, this study contributes to intelligent transportation systems, reducing accidents and improving overall road safety.
  Keywords: road safety; feature extraction; vehicle telematics; analysing telematics; sensor data.
  DOI: 10.1504/IJRIS.2025.10073243
   
  
• Face expression recognition for electricity marketing based on multiscale feature fusion with swin transformer  
  by Yanan Cai, Jinghui Chen, Wei Ge 
  Abstract: In this method, the proposed lightweight SPST module replaces the swin transformer blocks in the fourth stage of the original swin transformer model, significantly reducing the number of parameters and enabling lightweight and efficient inference. Subsequently, an EMA module is embedded after the second stage of the improved model to enhance the perception of subtle facial expression details through multi-scale feature extraction and cross-spatial information aggregation, thereby improving the accuracy and robustness of facial expression recognition in power marketing scenarios. Experimental results show that the proposed method achieves recognition accuracies of 97.56%, 86.46%, 87.29%, and 70.11% on the JAFFE, FERPLUS, RAF-DB, and FANE public facial expression datasets, respectively. Compared with the original swin transformer model, the improved model reduces the number of parameters by 15.8% and increases FPS by 9.6%, demonstrating significantly enhanced real-time performance while maintaining high recognition accuracy.
  Keywords: power marketing; face expression recognition; swin transformer; ST; multiscale feature fusion.
  DOI: 10.1504/IJRIS.2025.10073378
   
  
• Research on neural network-based UAV distribution grid line defect detection methods  
  by Bin Feng, Keke Lu, Shuang Fu, Jun Wei, Yu Zou 
  Abstract: This study presents a neural framework for UAV-based insulator defect detection in power distribution systems, addressing critical challenges in real-time operation, multi-scale defect recognition, and computational efficiency. Extensive experiments on a custom dataset (2,721 images, 6,812 instances) demonstrate state-of-the-art performance with 97.3% mAP@0.5:0.95 and 31.4 FPS on embedded GPUs, outperforming YOLOv5 (89.1% mAP), Faster R-CNN (93.4%), and DETR (89.8%). Ablation studies confirm the complementary nature of proposed components, showing cumulative improvements from 91.1% (baseline) to 97.3% mAP through progressive integration. The framework particularly excels in challenging scenarios with 91.4% AP for sub-10px defects and maintains <5.1% false positive rate under complex backgrounds.
  Keywords: embedded systems; multi-scale attention; power line defects; UAV inspection; wise-IoU loss; YOLOv8.
  DOI: 10.1504/IJRIS.2025.10073667
   
  
• Research on a method for assessing the status of electric power metering assets based on neural network federated learning  
  by Mingxin Jin, Shanshan Li, Guanna Lu, Yanguo Lv, Huinan Wang 
  Abstract: This approach not only avoids the security risks associated with third-party coordination but also enhances the models performance in practical applications such as fault diagnosis and electricity bill recovery risk prediction. Additionally, an incentive mechanism based on multi-dimensional contribution assessment and a block chain-based smart contract implementation scheme is designed to provide a sustainable motivational guarantee for multi-party collaboration. Specifically, by exchanging encrypted intermediate parameters (such as gradients or weight updates) during model training, the method achieves effective integration and joint modelling of multi-party data values.
  Keywords: federated learning; information security; machine learning; neural networks; electricity metering.
  DOI: 10.1504/IJRIS.2025.10074388
   
  
• Research on a deep learning-based coordinated optimisation and control technology for source-load-storage in new-type distribution networks  
  by Xiaomeng Yan, Peng Wang, Tao Liang, Wei Jiang, Yang Liu, Jun Guo, Zhebin Sun 
  Abstract: This paper proposes an intelligent multi-timescale optimisation and control method for active distribution networks based on deep reinforcement learning, taking into account the accuracy of generation-load power forecasting and the operational characteristics of devices. In the day-ahead stage, control plans for energy storage systems and flexible loads are formulated to achieve economic operation of the distribution network and reduce the peak-shaving pressure on the upper-level grid. A corresponding feature extraction method is designed for the multi-node, multi-period state space. In the intraday stage, the optimisation scheduling problem is transformed into a Markov decision process.
  Keywords: active distribution networks; optimised regulation; source-load-storage synergy; deep reinforcement learning; DRL; power prediction.
  DOI: 10.1504/IJRIS.2025.10074597
   
  
• Research on online error estimation method for station gate metering devices based on dynamic bus topology unit energy conservation  
  by Qiang Song, Zhiyi Qu, Jing Yang, Qingqing Fu, Tiejun Cheng 
  Abstract: This enables the establishment of a mapping between metering device errors and deviations in system energy conservation, forming a dynamic error modelling framework that reflects actual operating conditions. Then, a fading memory mechanism is introduced, and the FMRLS algorithm is employed to recursively estimate model parameters, thereby realising online and adaptive estimation of metering device errors. Simulation results demonstrate that, compared with the Levenberg-Marquardt (LM) algorithm and the limited memory recursive least squares (LMRLS) algorithm, the proposed method significantly improves the accuracy and dynamic responsiveness of error estimation while maintaining convergence stability.
  Keywords: error estimation; dynamic line loss; fading memory recursive least squares; online estimation.
  DOI: 10.1504/IJRIS.2025.10074598
   
  
• MIV-3: modified inception V3 architecture for enhancing periodontal diagnostic accuracy with SE attention module  
  by R. Kausalya, J. Anitha Ruth 
  Abstract: Over the recent decades, real-world applications and research which use AI (Artificial Intelligence) have evolved significantly, exclusively in dental and healthcare sectors. Our research discusses the utilisation of AI in X-ray imaging to detect periodontal diseases at an early stage. MIV-3 (modified inception V-3) is a model which enhances feature extraction and accuracy in diagnosis by combining an attention module and a squeeze-and-excitation (SE) module. Separable convolutions are utilised by MIV-3 model for increasing computational efficiency without impacting accuracy. A NPV and sensitivity of 98.37% and 94.68% respectively were depicted in the experimental data. Having a sensitivity of 94.68%, NVP of 98.85%, ROC-AUC of 99.14% and a specificity of 97.65% will help the model in predicting dental caries in a more accurate manner. The results indicate that the detection of periodontal disease happens at a faster pace and more accurately with the proposed AI-driven method. For developing the model, MATLAB program is utilised which offers a strong and dependable tool for diagnosis in clinical applications.
  Keywords: periodontal diagnosis; dental care; deep learning; inception V3; squeeze and excitation.
  DOI: 10.1504/IJRIS.2025.10074670
   
  
• Breast cancer classification refined using ResNet50 parameter tuning with lyre bird optimisation  
  by Sabura Banu Urundai Meeran 
  Abstract: Breast cancer remains a major cause of mortality among women, highlighting the need for accurate and efficient diagnostic methods. Deep learning, particularly CNNs, has improved medical image analysis, yet further optimisation is required for better precision and faster inference. This study optimises ResNet50 using the lyrebird optimisation (LBO) algorithm for hyperparameter tuning. A histopathological image dataset with cancer and non-cancer classes was used for training and evaluation. LBO fine-tuned key parameters such as learning rate, significantly enhancing model performance. The LBO-optimised ResNet50 outperformed standard ResNet50, Inception V3, and VGG16, achieving 98.85% accuracy along with high precision, recall, F1 score, and specificity (98.6%). The model also achieved an AUC-ROC of 99.98%, low log loss (0.0267), and reduced inference time (0.1377 seconds). Confusion matrix results showed fewer misclassifications. While promising for improving diagnostic reliability, additional clinical validation is recommended.
  Keywords: histopathological image analysis; deep learning models; hyperparameter tuning; diagnostic accuracy; medical image classification; confusion matrix analysis; performance optimisation; computer-aided diagnosis.
  DOI: 10.1504/IJRIS.2025.10075375
   
  
• Research on error optimisation algorithm for station gate electric energy metering devices based on triplet Siamese networks  
  by Qiang Song, Zhiyi Qu, Jing Yang, Qingqing Fu, Tiejun Cheng 
  Abstract: The triplet Siamese network not only extracts features from the training samples themselves but also learns the similarities among samples of the same class and differences among samples of different classes, significantly enhancing the clustering effect and discriminative ability of the feature vectors. Simulation results demonstrate that the proposed algorithm achieves high accuracy and superior performance under small-sample conditions, significantly outperforming traditional machine learning methods and other deep learning models. It can effectively support error optimisation for station gate electric energy metering devices and contribute to enhancing the intelligence and security stability of power grid operations.
  Keywords: ternary twin network; Gram’s corner field; plant-station gateway; power metering device error.
  DOI: 10.1504/IJRIS.2025.10075376
   
  
• Research on an online voltage unbalance mitigation method for distribution networks based on deep reinforcement learning  
  by Lin Xu, Chang Liu, Houdong Xu, Yan Gong, Fuxin Li, Yi Zheng 
  Abstract: Key innovations include: achieving model-free online decision-making, eliminating dependence on precise network parameters; possessing dynamic environment adaptability to respond in real-time to load fluctuations and distributed generation (DG) output variations; and simultaneously enhancing voltage balance and governance cost-effectiveness through reward function optimisation. Simulation results demonstrate that this method can effectively suppress voltage unbalance (reduced by over 30% in typical scenarios) within seconds, significantly decrease violation duration, and optimise compensation device switching frequency. It provides crucial technological support for constructing an intelligent and agile new-generation distribution network voltage governance system.
  Keywords: distribution grid; distributed resources; three-phase imbalance; intensive learning.
  DOI: 10.1504/IJRIS.2025.10075377
   
  
• Research on data-driven methods for evaluating and predicting the health status of energy storage cell packs  
  by Ning Li, Pengcheng Wei, Mingyang Wang, Yuan Liang, Dengyou Lei 
  Abstract: To address the limitations of traditional scheduling methods in modelling multi-variable coupling relationships and dynamic response delays, this paper proposes an attention mechanism-based multi-layer neural network (AMNN) optimisation framework. By employing a bidirectional long short-term memory (bi-LSTM) network, a multidimensional time-series prediction model incorporating electricity price fluctuations, battery aging, and meteorological features is constructed to achieve precise perception of the energy storage system's operational status. Validation using real-world operational data demonstrates that compared to the PSO optimisation algorithm, this method reduces scheduling errors by 19.7% during sudden load fluctuations and lowers the lifetime cost per kilowatt-hour by 12.3%.
  Keywords: deep learning; electric energy metering; fault diagnosis; smart grid monitoring.
  DOI: 10.1504/IJRIS.2025.10075378
   
  
• Simulation and dispatch optimisation of electricity spot markets considering renewable energy uncertainty  
  by Xuanyuan Wang, Xu Gao, Zhen Ji, Wei Sun, Bo Yan, Bohao Sun 
  Abstract: This paper aims to develop an integrated electricity spot market simulation and dispatch optimisation model that incorporates the characteristics of renewable energy. First, by establishing probabilistic models for wind and solar power output and employing stochastic programming or robust optimisation methods, the impact of their uncertainty on market clearing is characterised. Second, the objective function comprehensively considers factors such as minimising total system operating costs and maximising renewable energy integration, while the constraints rigorously account for grid security, unit technical limits, and power balance requirements, forming a complex mathematical optimisation problem.
  Keywords: simulation; distribution network; electricity spot markets; hierarchical planning.
  DOI: 10.1504/IJRIS.2026.10076057
   
  
• Research on wide-area protection algorithms for power grids based on fault charge quantity comparison and distributed computing  
  by Yu Sui, Xun Lu, Xiaoyu Deng, Wei Xu 
  Abstract: The method employs fault charge comparison, where local integration is used to extract the positive and negative characteristics of fault currents on both sides, enabling distributed retention and transmission of key features. For switching information, it relies only on the stage II and III starting signals of distance protection, calculates the fault correlation coefficient combined with threshold criteria, and aggregates the basic probability assignments of adjacent lines, thereby reducing the need for centralised transmission across the network. The algorithm applies an improved evidence theory within the distributed framework to fuse multi-source information and reliably identify faulted lines.
  Keywords: fault charge quantity; wide-area protection; fault identification; fault tolerance; distributed computing.
  DOI: 10.1504/IJRIS.2026.10076502
   
  
• Research on low-latency communication network methods for power system automation based on 5G technology  
  by Yabin Chen, Wei Xu, Xiaoyu Deng, Yu Sui 
  Abstract: This paper focuses on the stringent requirements for real-time performance, reliability, and security in communication transmission for power system automation, and researches low-latency communication network methods based on 5G technology. Traditional communication methods face challenges in meeting the low-latency and high-reliability demands of new services such as distributed intelligent control, wide-area protection, and precise load monitoring. It emphasises the in-depth integration of 5G technology with power automation services and the design of end-to-end communication solutions to support the future smart grid, thereby enhancing its rapid response and handling capabilities for renewable energy integration and complex faults.
  Keywords: 5G technology; power system automation; low latency; communication networks.
  DOI: 10.1504/IJRIS.2026.10076503
   
  
• Research on modelling and anomaly analysis methods for metering errors at factory stations  
  by Zhiyi Qu, Qiang Song, Pengcheng Li, Qinghui Chen, Tiejun Cheng 
  Abstract: A norm-feedback algorithm is introduced to optimise the distribution of weights during training, thereby improving the convergence and predictive performance of the model. Based on the extracted error features, statistical methods are employed to derive confidence intervals for metering errors, enabling early identification and assessment of abnormal metering behaviours. Experimental results demonstrate that the method can accurately model the characteristics of gateway metering errors and identify multiple extreme conditions that may lead to significant deviations, providing theoretical support and technical means for enhancing the reliability and anomaly monitoring capability of substation gateway metering.
  Keywords: electricity metre; confidence interval of measured values; convolutional neural network: norm feedback algorithm.
  DOI: 10.1504/IJRIS.2026.10076504
   
  
• Research on load dispatch and grid restoration in disaster-resilient emergency response for distribution networks based on nonlinear programming  
  by Hao Dai, Guowei Liu, Lisheng Xin, Longlong Shang, Qingmiao Guo, Hao Deng 
  Abstract: This paper addresses large-scale blackout scenarios in distribution networks following extreme disasters and investigates post-disaster emergency restoration strategies based on nonlinear programming. By constructing a multi-objective optimisation model that aims to maximise the total restored load and minimise switching operations, it comprehensively considers security constraints such as line capacity, voltage deviation, and radial operation, forming a mixed-integer nonlinear programming problem. The model effectively coordinates flexible resources like distributed generation and soft open points to achieve the coordinated optimisation of load transfer and network reconfiguration. Simulation results demonstrate that the proposed strategy significantly enhances the efficiency and resilience of post-disaster restoration.
  Keywords: deep learning; disaster-resilient; grid restoration; prediction model.
  DOI: 10.1504/IJRIS.2026.10076505
   
  
• Research on source-load cooperative planning of distribution network based on carbon emission of distributed power sources  
  by Hanyun Wang, Jianjie Jiang, Yang Liu, Tao Wang, Jiaqian Chen, Wei Zheng, Hai Liu 
  Abstract: The paper proposes a source-load cooperative planning method for distribution networks based on load carbon emission characteristics. Based on the carbon potential and net load curve of the distribution network, a quantitative model of load carbon emission characteristics considering carbon emissions from electricity consumption, distance low carbon, trend low carbon and low carbon electricity consumption rate is established to obtain the low carbon planning priority of each distribution network within the regional grid. The carbon emission reduction benefits of the proposed method are analysed and verified by simulation on the improved IEEE30 node system.
  Keywords: carbon intensity; load carbon emission characteristics; low-carbon demand response; source-load synergy; low carbon planning.
  DOI: 10.1504/IJRIS.2026.10076684
   
  
• Research on sparse prediction training technology for brain-like models based on pulse neural networks  
  by Guoliang Zhang, Peng Zhang, Fei Zhou, Zexu Du, Jiangqi Chen, Zhisong Zhang, Qingyu Kong 
  Abstract: This paper proposes a sparsity-prediction-based SNN training method, which introduces a predictive sparsity mechanism into the network structure to effectively reduce redundant computations and unnecessary synaptic updates. Specifically, during the training phase, an improved global recursive partitioning optimisation strategy is employed to enhance inter-cluster communication efficiency. Experimental results on five representative SNN benchmark models demonstrate that the proposed method significantly reduces communication latency and energy consumption while maintaining model accuracy, thereby improving training efficiency. Compared with existing approaches, it also exhibits clear advantages in terms of sparsity utilisation and energy efficiency.
  Keywords: pulse neural networks; brain-inspired processors; sparse prediction; training techniques.
  DOI: 10.1504/IJRIS.2026.10076685
   
  
• Research on organisational forms and operational mechanisms of multimodal industry-education integration platforms  
  by Na Xie 
  Abstract: This study examines multimodal industry-education integration platforms, aiming to systematically analyse their diverse characteristics in organisational structure, constituent entities, and connection methods, along with their applicable scenarios. Through a combination of theoretical and practical analysis, this study aims to construct a more efficient and sustainable organisational and operational framework model for multimodal industry-education integration platforms. This framework seeks to effectively address the structural barriers and functional challenges encountered in the process of deepening industry-education integration, thereby supporting the enhancement of technical and skilled talent cultivation, strengthening industrial innovation momentum, and serving the high-quality development of regional economies.
  Keywords: multimodal; industry-education integration; organisational structure; operational mechanism.
  DOI: 10.1504/IJRIS.2026.10076686
   
  
• Research on task management in English teaching based on multimodal fusion neural networks  
  by Tiantian Tang, Xin Guo 
  Abstract: This paper addresses the issues of insufficient personalisation and efficiency bottlenecks in traditional English teaching task management by proposing a novel management model based on multimodal fusion neural networks. By deeply integrating multimodal data including student learning behaviour videos, voice interactions, text assignments, and classroom expressions the model employs neural networks for feature extraction and collaborative analysis. This enables precise perception of learning states, dynamic adaptation of task difficulty, and intelligent recommendation of teaching resources. Experimental results demonstrate that this framework effectively enhances task planning efficiency and personalisation levels, providing a new technical pathway and practical reference for intelligent English teaching management.
  Keywords: multimodal; fusion neural networks; English teaching; task management.
  DOI: 10.1504/IJRIS.2026.10076689
   
  
• A study on multitask deep learning-based prediction of student dropout risk and analysis of influencing factors  
  by Guohua Sun, Hongxia Jia 
  Abstract: To address the challenge of predicting student attrition risk in higher education institutions, this study proposes a multi-task deep learning-based early warning model for student dropout. This approach enables precise prediction of attrition risk and in-depth analysis of key influencing factors. By jointly learning multi-dimensional data including academic performance, behavioural characteristics, and personal attributes through a shared feature representation layer, the method simultaneously accomplishes attrition classification and factor analysis tasks. Experimental results demonstrate that this model achieves significant improvements in both prediction accuracy and stability compared to traditional single-task models. It effectively identifies key factors influencing student attrition, such as academic performance, attendance rates, and engagement levels, providing data-driven decision support for universities to implement targeted interventions and academic support.
  Keywords: multitask deep learning; student attrition; risk prediction; influencing factors.
  DOI: 10.1504/IJRIS.2026.10076691
   
  
• Fermatean neutrosophic sets and their role in advanced decision-making systems  
  by Prasanta Kumar Raut, K. Saritha, M. Gayathri Lakshmi, R. Rajalakshmi 
  Abstract: In recent years, the need for effective representation and management of uncertain, imprecise, and inconsistent information has grown rapidly, especially in complex decision-making environments. Fermatean neutrosophic sets (FNS), a novel extension of neutrosophic sets, have emerged as a powerful mathematical tool capable of capturing higher degrees of uncertainty by relaxing conventional constraints. This paper presents a comprehensive overview of Fermatean neutrosophic sets, highlighting their foundational structure, key properties, and advantages over classical and intuitionistic fuzzy paradigms. Furthermore, we explore the pivotal role of FNS in advanced decision-making systems, including multi-criteria decision making (MCDM), risk assessment, and data classification problems. Illustrative examples and potential application domains are discussed to showcase the effectiveness of Fermatean neutrosophic models in real-world decision scenarios.
  Keywords: Fermatean neutrosophic set; FNS; uncertainty modelling; indeterminacy; neutrosophic logic; fuzzy systems.
  DOI: 10.1504/IJRIS.2026.10076734
   
  
• A hybrid model of fuzzy logic to enhance data mining accuracy incorporating intra-concentration and inter-separability loss into neighbourhood component analysis  
  by Hemangini Mohanty, Santilata Champati 
  Abstract: Data mining is crucial to discovering meaningful insights and patterns from massive datasets. However, the accuracy and efficiency of data mining algorithms are often challenged by the curse of dimensionality and the complexity of real-world data. In this article, we propose a novel approach to enhance the accuracy of data mining by enriching the concept of intra-concentration and inter-separability (I2CS) loss into neighbourhood component analysis (NCA). NCA is a dimensionality reduction technique that focuses on preserving local neighbourhood information, thus improving classification accuracy. Fuzzy logic, on the other hand, provides a flexible framework to handle uncertainty and vagueness in data, enabling more nuanced decision-making. By integrating fuzzy C-means clustering with I2CS-NCA, we aim to leverage the complementary strengths of both approaches to enhance the accuracy and robustness of data mining algorithms. Also, the experimental results show that the proposed model gives the highest accuracy.
  Keywords: I2CS loss; neighbourhood component analysis; NCA; fuzzy C-means clustering; random forest.
  DOI: 10.1504/IJRIS.2024.10067117
   
  
• Information fusion method on hexagonal fuzzy number-based multi-criteria decision-making problems  
  by V. Lakshmana Gomathi Nayagam, R. Bharanidharan 
  Abstract: Receiving information from experts is a crucial stage in fuzzy multi-criteria decision-making (MCDM) problems. Various types of fuzzy numbers are used in fuzzy MCDM problems. In particular, hexagonal fuzzy number is widely used in fuzzy MCDM problems because of its convenience on piecewise linearity. The major drawback of fuzzy MCDM problems is non-availability of information for some alternatives with respect to some criteria while collecting information from the experts. To overcome this, researchers found some methodologies which are known as information fusion/infusion methods. In this paper, we have proposed two infusion methods based on score function and similarity measure. We have analysed the proposed infusion algorithms by giving illustrative numerical examples. Further, due to the needfulness, a new similarity measure on hexagonal fuzzy numbers have been introduced and used in the infusion method.
  Keywords: hexagonal fuzzy numbers; information fusion; missing data MCDM; similarity measure on HXFN.
  DOI: 10.1504/IJRIS.2024.10068105
   
  
• Ensemble of transfer learning with convolutional neural networks for writer recognition in historical documents  
  by Radmila Janković Babić, Alessia Amelio, Ivo Rumenov Draganov, Marijana Ćosović 
  Abstract: In the cultural heritage domain, writer recognition has become a challenging classification task still explored for historical documents, due to the presence of different types of noise in the documents, i.e., ink bleed-through, ink corrosion, stains on paper or parchment, difficulty in the character discrimination, elements different from the text, such as images, etc. that limit the effectiveness of existing techniques. To further advance in terms of robustness of classification and experimental setting, we propose a new deep learning model which ensembles pre-trained convolutional neural networks for writer recognition. Specifically, the ensemble is composed of three pre-trained Inception-ResNet-v2 models with different hyperparameter values. Results obtained on the benchmark ICDAR 2019 dataset of handwritten historical documents prove that the proposed approach is very promising in recognising the handwritten characters of different writers, also when compared with other deep learning models.
  Keywords: convolutional neural networks; CNNs; writer recognition; cultural heritage; historical documents; ensemble learning; artificial neural networks; document analysis; deep learning; transfer learning.
  DOI: 10.1504/IJRIS.2024.10067482
   
  
• Rice plant nutrient deficiency classification using deep learning techniques  
  by D. Sindhujah, R. Shoba Rani 
  Abstract: Every day, half of the world's population eats rice. The World Bank predicts that by 2025, the demand for rice consumption will have increased by 51%. Mineral deficiency is one of the variables that impact rice yield. Plants need a variety of minerals and nutrients to flourish, especially while they are in the process of blooming or developing fruit. Critical plant growth disorders, which impact agricultural productivity, are caused by nutrient deficiencies. As soon as farmers see signs of nutrient inadequacy in their plants, they may use effective nutrient management measures to remedy the situation. New possibilities in non-destructive field-based analysis for nutritional deficiencies have emerged with computer vision and deep learning algorithms. In this research, we presented a ResNet50 model that has been fine-tuned to identify nutritional deficits in rice images. Our suggested model is combined with the ADAM optimiser and the softmax classifier to get the best possible outcome. Using our model, we will determine whether the rice plant is deficient in nitrogen, phosphorus, and potassium. Our findings show that our model outperforms the competition with an accuracy of 94.34%.
  Keywords: image augmentation; ResNet50; ADAM optimiser; softmax classifier; critical plant growth disorders; deep learning algorithms; nutrient inadequacy; agricultural productivity.
  DOI: 10.1504/IJRIS.2024.10068106
   
  
• Embracing creativity and encouraging teacher satisfaction at intelligent systems  
  by Neenet Baby Manjaly, S.A. Vignesh Karthik, H. Lekha, V. Ameena Babu, Gayathri Joshi 
  Abstract: Many firms can now achieve high employee performance thanks to self-motivated working cultures. Employee behaviour and job satisfaction at the organisational climate level have been extensively studied. Academics need job dedication and satisfaction to boost productivity, student advancement, retention, and cognitive and personal growth. Academic independence, creativity, professional commitment, and job joy are examined in this study. This research will examine the relationship between these factors. The study tested work-life balance theories for Chennai's private professional teachers. Data was collected using a self-administered questionnaire, and 353 were analysed. The model's validity and reliability were assessed using multivariate statistics. Data was analysed using structural equation modelling for normalcy, reliability, and discriminant validity. Results demonstrated that employment independence boosts creativity, dedication, and satisfaction. All components boost job happiness. Freedom at work and job commitment facilitated creativity, supporting the mediation hypothesis. The results also showed that job dedication mediates flexibility at work and job happiness. Workplace independence increases worker satisfaction, creativity, and commitment, according to the study. This research greatly improves our understanding of academic workplace dynamics.
  Keywords: embracing creativity; encouraging teacher; freedom at work; FAW; job commitment; intelligent systems; job satisfaction; employment independence; teaching profession.
  DOI: 10.1504/IJRIS.2025.10068691