International Journal of Reasoning-based Intelligent Systems (61 papers in press)

Regular Issues

• Open Access

  A two-stage BPE-NPMI framework for core phrase discovery in unsegmented classical Chinese corpora: a systematic, reproducible evaluation ( Free Full-text Access )

  by Lei Peng, Zheng Teng 
  Abstract: Classical Chinese poetry survives as an unbroken stream of characters without word boundaries, lexicons or annotated corpora, so finding its poetic phrases automatically is a low-resource text-processing problem, and solving it matters for imagery studies, dictionary compilation and text retrieval. This paper proposes BPE-NPMI, a two-stage framework: byte pair encoding surfaces two-character candidates by frequency-driven merging, and normalised pointwise mutual information re-ranks them by cohesion. Tested on Complete Tang Poetry (42,452 poems) and Complete Song Ci (21,025 lyrics) with pooling-based expert gold standards and seven comparison methods, the framework reaches precisions of 0.98 and 0.965 in its top-200 candidates, significantly above all baselines. Further analysis shows that the performance gain stems mainly from the cohesion measure rather than the candidate generator. The framework requires no language-specific resources and scales near-linearly, making it a practical lexicon-discovery tool for unsegmented historical and low-resource texts.
  Keywords: phrase discovery; unsegmented text; byte pair encoding; BPE; normalised pointwise mutual information; PMI; classical Chinese; Tang poetry; Song Ci; evaluation methodology.
  DOI: 10.1504/IJRIS.2026.10081652
   
  

• Open Access

  Multiscale corporate cash flow volatility forecasting using VMD-SSA-transformer: adaptive decomposition, gated fusion and cross-industry validation ( Free Full-text Access )

  by Ruixia Tian 
  Abstract: Corporate cash flow varies across months because receivable collection, concentrated payments, settlement schedules, financing activities and operating conditions affect the timing and magnitude of inflows and outflows. This study develops a VMD-SSA-transformer framework for monthly net cash flow forecasting through adaptive decomposition, gated multiscale fusion and attention-based temporal modelling. The evaluation uses an independent 84-month series and 5,376 monthly observations from 64 enterprises in five industries under enterprise-grouped rolling-origin and leave-one-industry-out protocols. The model achieves MAE 6.53, RMSE 8.41, MAPE 4.38% and R2 0.948. Across ten runs, mean RMSE is 8.43?0.11 with a 95% confidence interval of 8.358.51; Holm-adjusted comparisons with neural baselines yield p < 0.001. Industry-holdout RMSE ranges from 7.96 to 8.88. RMSE reaches 10.87 under 5 dB noise and 10.52 at the five-month horizon. These results establish forecasting accuracy and robustness within the evaluated data, while end-to-end computational cost and management value require separate consideration.
  Keywords: VMD; SSA; transformer; corporate cash flow; multiscale forecasting; gated fusion; cross-industry generalisation.
  DOI: 10.1504/IJRIS.2026.10081777
   
  

• Open Access

  A self-supervised unified representation architecture for multi-source heterogeneous data ( Free Full-text Access )

  by Lian Tong, Liyan Zhou 
  Abstract: The explosive growth of multi-source heterogeneous data in various industrial and scientific fields has posed fundamental challenges to the traditional representation learning paradigm. This paper proposes a self-supervised unified representation architecture specifically designed for the environment of multi-source heterogeneous data UniExtract. This architecture introduces modality-specific tokenisation and combines the spatio-temporal normalisation method to map dissimilar data streams into a coherent latent space. Its core component, the adaptive feature extraction module, combines local context encoding with a global cross-attention fusion mechanism, dynamically capturing multi-scale structural correlations without relying on downstream labels. The entire network is trained end-to-end through a joint self-supervised optimisation objective that includes contrast alignment loss and mask reconstruction loss. Under a controlled retraining protocol (group A), UniExtract improves over the strongest retrained baseline on the five reported primary metrics; comparisons against large-scale pretrained reference models (group B) are reported as reference points only.
  Keywords: multi-source heterogeneous data; feature extraction; self-supervised learning; cross-modal alignment; cross-attention fusion; deep representation learning.
  DOI: 10.1504/IJRIS.2026.10081855
   
  

• Open Access

  Time series analysis of learning behaviour and construction of academic early warning mechanism for digital transformation of education ( Free Full-text Access )

  by Zhongguo Wang, Xiaohong Yang 
  Abstract: Traditional academic early warning systems rely on static features or aggregated statistics, overlooking the temporal dynamics of learning behaviour and resulting in delayed warnings with poor interpretability. This paper proposes a temporal analysis framework that integrates the hidden Markov model (HMM) with a BiLSTM-Attention network to model behavioural sequences. On the OULAD dataset, we construct multi-dimensional time-series features encompassing behavioural frequency, temporal regularity, and cognitive engagement. The proposed model achieves 91.3% accuracy in early warning tasks, outperforming logistic regression and random forest by 15.7% and 9.2%, respectively. Visualisation of the state transition matrix and attention weights illustrates the typical temporal pathways of high-risk behaviours as captured by the model. The mechanism identifies at-risk students 2-4 weeks before course completion by rigorously learning temporal dependencies exclusively within the historical data available up to the prediction cut-off.
  Keywords: digital transformation of education; learning behaviour analysis; time series analysis; academic warning; deep learning.
  DOI: 10.1504/IJRIS.2026.10081886
   
  

• Open Access

  Teacher-governed generative AI-assisted training for piano improvisational accompaniment: effects and a multi-source explanatory framework ( Free Full-text Access )

  by Huixin Chang 
  Abstract: Generative artificial intelligence can produce music substitutes, but its contribution to performance learning depends on how these substitutes are used. This quasi-experiment examined a ten-week piano improvisation accompaniment training program led by teachers at a Chinese university (N = 120; 60 in each condition). Three blinded experts scored the independent pre-test, mid-test, and post-test performances on six dimensions. Questionnaires, platform logs, and interviews provided process evidence. The experimental group received a score of 13.95, while the control group received 5.57. Cluster robust estimation showed that the experimental group achieved 3.583 and 8.380 points respectively at the mid-test and post-test stages. All six aspects improved. The research results support a limited explanatory framework centred on task variation within constraints, comparison, teacher correction, and repetition of unassisted performance.
  Keywords: generative AI; piano improvisational accompaniment; teacher governance; generative practice; performance assessment; music education.
  DOI: 10.1504/IJRIS.2026.10081970
   
  

• Open Access

  An econometric analysis of the association between traditional Chinese medicine policy support, industrial agglomeration and innovation performance ( Free Full-text Access )

  by Zheng Xie, Yufang He, Lu Zhang, Yi Gao, Haowei Xu, Jingshuo Liu 
  Abstract: Although policy support for traditional Chinese medicine (TCM) has intensified, evidence on its relationship with industrial agglomeration and innovation performance remains limited. Using panel data for 30 provincial-level regions from 2011 to 2023, this study constructs a policy support index, an industrial agglomeration index, and a two-stage Super-NSBM innovation-performance measure. Two-way fixed-effects, statistical transmission, spatial Durbin, and panel threshold models are employed. Higher policy support is significantly associated with stronger industrial agglomeration and innovation performance. Industrial agglomeration accounts for 33.1% of the total estimated association, with related-diversification agglomeration showing the strongest indirect association. Spatial dependence is positive, and the policy-support association varies with agglomeration and knowledge absorption capacity. The results are interpreted as robust conditional associations rather than definitive causal effects.
  Keywords: traditional Chinese medicine policy; industrial agglomeration; innovation performance; spatial spillovers; threshold effect.
  DOI: 10.1504/IJRIS.2026.10081971
   
  

• Open Access

  Real-time optimal scheduling framework for construction machinery multi machine collaborative work combining federated learning and edge computing ( Free Full-text Access )

  by Wubin Xie 
  Abstract: This paper proposes a device-edge-cloud collaborative real-time scheduling framework (FLEC-Sched) that integrates federated learning and edge computing to address data silos, limited responsiveness, and privacy risks in collaborative scheduling of construction machinery clusters. Raw operational data remain local through distributed state perception and personalised modelling; asynchronous federated reinforcement learning enables distributed strategy optimisation across heterogeneous machines; and a lightweight edge-end decision-execution feedback loop ensures real-time responsiveness. In a five-machine earthmoving simulation, FLEC-Sched attains an average task completion time of 121.6 ? 5.4 s and total energy consumption of 5.34 ? 0.28 105 J, within 6.4% and 3.1% of the centralised upper bound, while reducing communication overhead by 92.3%. Extended experiments confirm scalability to 20 machines, millisecond-level local responses, and statistically significant advantages over the comparison baselines. Ablation studies validate the personalised modelling, asynchronous aggregation, and edge-end feedback components, establishing FLEC-Sched as a simulation-based proof of concept.
  Keywords: federated learning; edge computing; engineering machinery; collaborative scheduling; real-time optimisation; privacy protection.
  DOI: 10.1504/IJRIS.2026.10081983
   
  

• Open Access

  A cluster target audit and validation framework for learning analytics applied to an archived multimodal campus behaviour pipeline ( Free Full-text Access )

  by Zhiming Shi 
  Abstract: This study introduces the cluster target audit and validation (CTAV) framework for learning-analytics pipelines that combine unsupervised profiling with downstream supervised classification. CTAV integrates established leakage control, nested validation, clustering-stability analysis, and reproducibility practice through three original decision devices: a target-dependency trace, a surrogate-necessity test anchored to direct assignment by the frozen clustering rule, and four evidence grades that separate provenance, descriptive clustering, surrogate fidelity, and prospective outcome validity. The framework is applied to an archived multimodal campus-behaviour analysis involving 5,234 students, daily aggregated records, a reported 128-feature representation, four K-means profiles, and two historical MLP configurations. The archive does not preserve the executed MLP input map, fold-specific preprocessing objects, clustering diagnostics, or fold-level predictions. CTAV converts this case-specific audit into a reusable protocol that specifies claim definition, dependency analysis, outer-fold isolation, training-only clustering, stability testing, direct-rule comparison, nested tuning, untouched evaluation, statistical comparison, and artefact archiving.
  Keywords: learning analytics; cluster-derived targets; k-means; surrogate classification; target leakage; nested cross-validation; clustering stability; methodological audit; reproducibility.
  DOI: 10.1504/IJRIS.2026.10082050
   
  

• Open Access

  5G technology-based exploration of next-generation smart campus - taking ZTBU as an example ( Free Full-text Access )

  by Ping Wang, Zipei Zhang, Yuna Yao 
  Abstract: Intelligent campus construction suffers from bandwidth limits, low transmission rates and fragmented deployment. While 5G empowers higher education digital transformation, existing research remains conceptual without systematic IoT design, quantitative assessment or empirical validation. This study develops a five-layer 5G-IoT framework integrated with edge computing, analysing campus scenarios covering teaching, examination, evaluation, security and management. Deployed at ZTBU, the architecture achieves 18.7 ms average latency, 711.5 Mbps peak throughput and 99.98% uptime. NS-3 simulations confirm stable support for one million concurrent terminals with packet loss below 0.13%. It cuts service latency by 72.3% versus 4G and shows good linear scalability. The standardised deployment paradigm and unified data cockpit break data silos and improve collaborative governance for similar private colleges.
  Keywords: 5G; smart campus; IoT; multi-access edge computing; MEC; integrated 5G-IoT; network slicing; quantitative evaluation.
  DOI: 10.1504/IJRIS.2026.10082051
   
  

• Open Access

  Internet of things sensing-driven deep learning diagnosis for nuclear power plant equipment faults ( Free Full-text Access )

  by Qiaonian Xu, Pengyuan Wang, Yanhong Yang 
  Abstract: Heavy rotating machinery in critical industrial facilities operates under extreme stress. When microscopic early damages remain undetected, they rapidly escalate into catastrophic safety failures. However, current monitoring systems lack the capability to distinguish meaningful warning signals from overwhelming background interference. To address this urgent challenge, a cognitive-driven dynamic spatiotemporal graph routing network is proposed. This architecture replicates human cognitive allocation to automatically filter excessive sensor noise and accurately isolate early spatial and temporal machine degradation patterns. Comprehensive experiments using the XJTU-SY accelerated bearing degradation dataset demonstrate that the proposed method achieves a diagnostic accuracy of 97.2%, surpassing all compared baselines, including the state-of-the-art graph transformer GraphGPS (95.8%). Furthermore, the network significantly reduces the false alarm rate to 0.8% and detects a statistically significant representation shift preceding the conventional alarm threshold, validating its potential as a laboratory proof-of-concept for safety-critical applications.
  Keywords: fault diagnosis; graph neural networks; GNNs; intelligent fault diagnosis.
  DOI: 10.1504/IJRIS.2026.10082052
   
  

• Open Access

  Artistic image style transfer based on an attention generative adversarial network ( Free Full-text Access )

  by Xiaohan Liu, Linghe Fan 
  Abstract: Artistic style transfer converts natural photographs into artworks in the style of reference paintings, with applications in digital art creation and cultural-heritage digitisation. However, applying complex artistic styles to detailed photographs often distorts geometric shapes and blurs object boundaries, producing severe visual artefacts. To overcome this destructive blending problem, this paper proposes a novel generative adversarial network guided by a spatial attention mechanism. Instead of forcing an artistic style across the entire image, the proposed method calculates probability masks to dynamically separate structural content from background elements during feature fusion. Evaluated on diverse datasets, the proposed approach substantially reduces structural collapse. Specifically, the proposed model achieves a Frechet inception distance score of 18.45, improving over the strongest retrained diffusion-prior baseline by 3.86 points, and simultaneously reduces perceptual distortion by 24.9% relative to the style transformer baseline, as measured by learned perceptual image patch similarity, providing a practical high-fidelity synthesis framework.
  Keywords: image style transfer; generative adversarial network; spatial attention; feature decoupling.
  DOI: 10.1504/IJRIS.2026.10082053
   
  

• Open Access

  Deep learning reconstruction network for high dynamic range scene imaging ( Free Full-text Access )

  by Jingjing Wen, Xinlei Wang 
  Abstract: This paper proposes a dual-branch deep reconstruction network for high dynamic range scene photography. Current computational photography methods struggle with ghosting artefacts and colour distortion in highly dynamic scenes with extreme lighting variations. By filling the gap in physically constrained illumination mapping, this paper explicitly models the nonlinear camera response function within the feature extraction stage. The network integrates a physical constraint loss and a multi-scale spatial alignment module. Extensive evaluations demonstrate that this paper improves the tone-mapping-space PSNR by 1.38 dB and the linear-space PSNR by 1.63 dB, reduces the LPIPS perceptual error by 28.28%, and improves the HDR-VDP-2 quality score by 4.41% compared to state-of-the-art methods, achieving the closest agreement with the reference among the evaluated methods on the tested dynamic scenes and exposure configurations.
  Keywords: high dynamic range; deep learning; computational photography; illumination estimation; ghosting reduction.
  DOI: 10.1504/IJRIS.2026.10082054
   
  
• 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
   
  
• 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
   
  
• 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
   
  
• Vibration analysis of Fe-based soft magnetic composite core reactor based on improved particle swarm algorithm  
  by Yangyang Ma, Wenle Song, Jie Gao, Yang Liu, Yilei Shang, Weimei Zhao, Fuyao Yang 
  Abstract: Using central composite design combined with finite element simulations, the study investigates the influence of different air-gap structural parameters on vibration responses and establishes an orthogonal polynomial-based response prediction model for accurately estimating core vibration displacement. Taking the minimisation of core vibration as the optimisation objective while maintaining the inductance value nearly constant, the optimal air-gap length of the reactor is obtained. The results show that under the optimised structural parameters, the maximum core vibration displacement is reduced by 10%, while the inductance variation is only 0.051%. This optimisation method provides significant reference value for reducing vibration and noise.
  Keywords: iron core reactor; electromagnetic-structural force field coupling; air gap structure; core vibration; orthogonal polynomial model.
  DOI: 10.1504/IJRIS.2026.10077021
   
  
• An autonomous UAV trajectory optimisation and continuous stitching method for refined inspection of transmission lines  
  by Lin Ao, Teng Ma 
  Abstract: Combined with a spherical-threshold-based spatial density filtering method, redundant trajectory points near shooting locations are removed. Furthermore, a minimum turning radius constraint and arc smoothing are introduced to achieve trajectory smoothness, and flight safety validation is completed through safety distance constraints. Experimental results demonstrate that the proposed method can reduce the number of trajectory points for a single tower inspection by more than 90% while ensuring shooting consistency and flight safety. This significantly enhances the efficiency of UAV autonomous inspections and the reusability of trajectories, providing an engineering-feasible solution for refined and continuous autonomous inspections of transmission lines using UAVs.
  Keywords: autonomous inspection; trajectory optimisation; Douglas-Peucker algorithm; unmanned aerial vehicle; UAV.
  DOI: 10.1504/IJRIS.2026.10077022
   
  
• Exploring the value and pathways of integrating algorithmic ethics education into ideological and political courses in the new era  
  by Ge Chen, Xiaodong Yang 
  Abstract: This paper explores the practical value and implementation pathways of integrating algorithm ethics education into ideological and political courses in the new era. Algorithmic technologies are profoundly transforming social life, and the ethical challenges they pose urgently require educational guidance to address. The study argues that integrating such education into ideological and political courses helps cultivate students correct understanding of technological ethics, sense of social responsibility, and value judgment capabilities, thereby achieving the unity of technological rationality and humanistic spirit. Specific pathways include: developing interdisciplinary teaching cases that integrate topics such as algorithmic transparency, fairness, and privacy with core socialist values; innovating teaching methods through scenario-based discussions and ethical deliberation; strengthening faculty training to enhance teachers technological ethics literacy; and establishing collaborative education mechanisms involving universities, enterprises, and societal stakeholders.
  Keywords: algorithmic ethics; educational integration; new era; ideological and political education courses.
  DOI: 10.1504/IJRIS.2026.10077113
   
  
• Research on teaching and intelligent management based on multimodal fusion deep learning behaviour analysis  
  by Wenjing Sun, Weisong Wang, Teng Ma 
  Abstract: Addressing bottlenecks in traditional classroom teaching evaluations such as high subjectivity and delayed feedback this study explores intelligent teaching-management feedback mechanisms centred on multimodal fusion deep learning for behavioural analysis. By constructing an end-to-end intelligent analysis framework that integrates multi-source data including classroom visuals, audio, and text, and employing attention-based deep fusion strategies, it achieves fine-grained recognition and contextual understanding of teacher-student instructional behaviours. This research not only provides an innovative technical approach for classroom behaviour analysis but also drives a paradigm shift from experiential teaching management toward precision-driven, personalised educational governance through data-driven intelligent feedback mechanisms.
  Keywords: deep learning; instructional networks; teaching; classroom ecologisation; management system.
  DOI: 10.1504/IJRIS.2026.10077117
   
  
• Research on deep learning-driven adaptive course resource recommendation and instructional planning  
  by Jiaxue Liu, Xiaoxian Su 
  Abstract: This study addresses the common challenges faced by online learning platforms, such as resource overload, rigid learning pathways, and lack of personalisation, by constructing an integrated framework for adaptive course resource recommendation and teaching planning based on deep learning. This framework combines a dual-channel learner dynamic perception model based on transformer and knowledge graph embedding, a resource representation approach that integrates knowledge structure and semantic information, and a long-term teaching planner based on deep reinforcement learning. Experiments on the public datasets ASSISTments2012 and EdNet indicate that our model improves recommendation accuracy by 12.5% compared to the best baseline methods, achieves a path rationality score of 4.5/5.0 as evaluated by experts, and enhances learning gains by 15%. The results suggest that the proposed framework effectively enables personalised cognitive navigation and teaching pathway planning, providing a feasible technical path for the development of the next generation of adaptive learning systems.
  Keywords: deep learning-driven; adaptive; course resource recommendation; teaching planning.
  DOI: 10.1504/IJRIS.2026.10077118
   
  
• Research on the ecological management system of English teaching classrooms based on deep learning network technology  
  by Jingshu Wu, Yawei Hu, Haodong Guo 
  Abstract: With the rapid advancement of artificial intelligence technologies such as deep learning, traditional English teaching models are undergoing profound transformation. This study aims to establish an ecological management system for English classrooms. By constructing this ecological management model, the objectives are to achieve precise allocation of teaching resources, dynamic optimisation of teaching processes, intelligent recommendation of personalised learning paths, and diversified comprehensive teaching evaluation. This approach promotes the synergistic evolution and balanced development of all elements within the classroom ecosystem. Establish a theoretical framework and practical pathways for creating a new ecosystem of intelligent and harmonious English teaching.
  Keywords: deep learning; instructional networks; English teaching; classroom ecologisation; management system.
  DOI: 10.1504/IJRIS.2026.10077119
   
  
• Research on the microstructure and magnetic properties of dual-phase composite magnetic materials  
  by Jie Gao, Fuyao Yang, Yang Liu, Cong Wang, Pinpin Zhu, Zhibin Nie 
  Abstract: This study systematically investigates the intrinsic relationship between microstructural characteristics and macroscopic magnetic properties in dual-phase composite magnetic materials. By adjusting preparation process parameters, composite microstructures with varying phase compositions, grain sizes, and interface morphologies were obtained. Combining microscopic analysis with magnetic measurements, the distribution, coupling state, and interface effects between soft and hard magnetic phases were analyzed in detail. Results indicate that exchange coupling between the two phases significantly influences the materials coercivity, remanence ratio, and maximum energy product. Optimising the microstructure effectively enhances magnetic properties, providing crucial theoretical and experimental foundations for designing and fabricating high-performance composite permanent magnet materials.
  Keywords: dual-phase composite; magnetic materials; microstructure; magnetisation; interphase interface.
  DOI: 10.1504/IJRIS.2026.10077121
   
  
• Research on route planning methods based on an improved particle swarm optimisation algorithm  
  by Yanfeng Xu, Yang Wang, Xiaobo Li, Xiang Xu 
  Abstract: Addressing the need for intelligent optimisation of equipment layout and route schemes in complex geographical environments, this paper integrates geographic information systems (GIS) with intelligent optimisation algorithms to propose a route planning method based on an improved particle swarm optimisation (PSO) algorithm. The equipment arrangement and route planning problem is modelled as an optimisation model with multiple constraints. To solve this model efficiently, improvements are made to the standard PSO algorithm by introducing an adaptive inertia weight and a hybrid learning strategy, which effectively balance the algorithms global exploration and local exploitation capabilities, thus preventing premature convergence.
  Keywords: particle swarm optimisation; PSO; route planning; equipment scheduling; spatial analysis; intelligent algorithms.
  DOI: 10.1504/IJRIS.2026.10077122
   
  
• A novel bipolar neutrosophic soft topological model for agricultural decision analysis  
  by Prasanta Kumar Raut, R. Rajalakshmi, K. Saritha, M. Gayathri Lakshmi 
  Abstract: Agricultural decision-making is frequently influenced by uncertainty stemming from environmental dynamics, soil heterogeneity, and varying agronomic conditions, which often lead to incomplete, vague, and even contradictory information. Conventional uncertainty modelling approaches, including fuzzy and intuitionistic frameworks, are not sufficiently equipped to address such complexity in a unified manner. In this paper, a novel analytical framework based on bipolar neutrosophic soft topological structures is proposed to effectively model these challenges. The introduced framework integrates the parameter-driven nature of soft sets with the capability of bipolar neutrosophic logic to simultaneously handle favourable and unfavourable information, together with the organisational advantages of topological structures. This combined model offers a refined representation of uncertainty and indeterminacy in decision environments. The applicability of the proposed methodology is demonstrated through multi-criteria decision-making problems in agriculture, including crop selection and resource management under uncertain scenarios. The results obtained from a detailed case study indicate that the proposed approach produces more robust, transparent, and reliable decisions compared to existing fuzzy-based techniques, highlighting its potential as a valuable tool for agricultural decision support systems.
  Keywords: bipolar neutrosophic set; BNS; soft topology; agricultural decision-making; uncertainty modelling; multi-criteria analysis; crop management.
  DOI: 10.1504/IJRIS.2026.10077123
   
  
• Research on optimising the absorption capacity of feeders in medium and low-voltage distribution networks based on deep learning  
  by Haitao Li, Xinming Xu, Xiaobin Chen 
  Abstract: With the large-scale integration of distributed renewable energy, medium- and low voltage distribution networks face severe challenges regarding feeder hosting capacity. Traditional assessment methods rely on precise physical models and static scenarios, struggling to handle high-dimensional uncertainties and real-time optimisation needs. This paper proposes a deep learning-based method for optimising feeder hosting capacity, constructing a collaborative prediction-assessment-decision framework. Through a hierarchical-coordinated optimisation architecture (centralised upper-layer optimisation and distributed fast-response lower layer), simulation case studies verify that this method significantly enhances system hosting capacity (Power/psi index), smoothes power fluctuations at the point of common coupling, eliminates voltage limit violations, and improves three-phase unbalance and voltage fluctuation ratios. It provides an effective technical pathway for the safe and high-quality operation of distribution networks under high-penetration distributed energy integration.
  Keywords: absorption capacity; low-voltage; distribution networks; deep learning; DL.
  DOI: 10.1504/IJRIS.2026.10077602
   
  
• Research on pre-trained transformer models incorporating domain knowledge in government big data analysis  
  by Yunlong Sun 
  Abstract: This paper explores the application of domain-knowledge-integrated pre-trained Transformer models in government big data analysis. Addressing the specialised nature and complex structure of government data, the study adapts general pre-trained models to specific domains and enhances their knowledge by incorporating domain dictionaries, entity knowledge, and business rules. It proposes training objectives and an integrated architecture tailored to the characteristics of government texts, thereby improving the models comprehension and reasoning capabilities in tasks such as policy analysis, public sentiment assessment, and service recommendation. Experiments demonstrate that this approach effectively improves the accuracy and interpretability of government text classification, information extraction, and decision support, providing robust technical support for smart government development.
  Keywords: domain knowledge integration; pre-training; transformer; government big data.
  DOI: 10.1504/IJRIS.2026.10077603
   
  
• Research on resource aggregation scheduling and optimal control strategies for virtual power plants based on spatio-temporal data mining analysis  
  by Hongtao Li, Dongyu He, Zhengzhe Li, Haoze Zhang, Liguo Zheng, Xingwang Jia, Xia Zhang 
  Abstract: This paper addresses the resource allocation demands for flexibility in new power systems, focusing on the aggregation scheduling and optimisation control of virtual power plants (VPPs). It proposes a strategy framework based on spatio-temporal data mining and computational analysis. By mining the spatio-temporal correlation characteristics and operational patterns of distributed energy resources and loads, the study constructs a resource aggregation model that accounts for multiple uncertainties. Subsequently, a collaborative optimisation method integrating forecasting, scheduling, and feedback control is designed to achieve economically efficient scheduling and dynamic real-time control of virtual power plants across multiple time scales from day-ahead to intraday. Simulation results demonstrate that the proposed strategy effectively enhances the aggregation and control capabilities of virtual power plants over distributed resources, significantly improving system operational economics and renewable energy integration levels. This provides theoretical foundations and technical references for the engineering application of virtual power plants.
  Keywords: spatio-temporal data mining; computational analysis; virtual power plant; VPP; resource aggregation and dispatch.
  DOI: 10.1504/IJRIS.2026.10077604
   
  
• Dynamic evolution analysis of relationships between traditional culture and Chinese literary figures based on knowledge graphs and graph neural networks  
  by Li Cai, Qi Wang 
  Abstract: This paper addresses the limitation in traditional culture and Chinese literary studies where character relationship analysis often remains static. It proposes an evolutionary analysis framework integrating dynamic knowledge graphs with graph neural networks (GNNs). By constructing a multimodal knowledge graph with temporal slices to encode characters, relationships, and their semantics, and combining GNNs with long short-term memory (LSTM) networks to capture network structure and temporal dependencies, this framework enables micro-level tracking and macro-level modelling of the entire process of character relationship formation, reinforcement, transformation, and rupture in classic works such as Dream of the Red Chamber and Romance of the Three Kingdoms. The research reveals that the evolution of literary character relationships exhibits a dynamic pattern intertwining event-driven abrupt changes with gradual shifts in emotional and ethical dynamics. This study provides a new semantic-driven computational pathway for digital humanities, advancing paradigm innovation in traditional cultural analysis.
  Keywords: knowledge graph; graph neural network; GNN; traditional culture; Chinese literary character relationships.
  DOI: 10.1504/IJRIS.2026.10077605
   
  
• Research on intelligent accounting decision-making and financial education pathways based on deep reinforcement learning  
  by Jia Feng, Xiaorui Xue 
  Abstract: This paper addresses the limitations of traditional accounting decision-making, which relies heavily on experience and rules, and the lack of personalised pathways in financial education. It explores the application of deep reinforcement learning in the fields of intelligent accounting and financial education. The study constructs an intelligent accounting decision-making model based on deep reinforcement learning, optimising financial decisions through dynamic environmental interaction and reward mechanisms. Concurrently, an adaptive financial education path recommendation system is designed to dynamically adjust teaching content and difficulty levels in real-time based on learner behavioural data. Results demonstrate that this approach effectively enhances the precision and automation of accounting decisions while providing personalised, interactive learning solutions for financial education. It holds both theoretical and practical value for advancing the intelligent transformation of accounting and innovating financial talent cultivation models.
  Keywords: deep reinforcement learning; DRL; intelligent accounting decision-making; financial education; path analysis.
  DOI: 10.1504/IJRIS.2026.10077606
   
  
• Research on uncertainty quantification algorithms based on Bayesian neural networks and a posteriori regularisation  
  by Li Zhang 
  Abstract: In security-sensitive applications of deep learning, accurately quantifying model prediction uncertainty is crucial. Bayesian neural networks, through their probabilistic framework, provide a theoretical foundation for simultaneously quantifying epistemic and aleatic uncertainty. However, traditional approximate inference methods (such as variational inference) often suffer from biased posterior distribution estimates due to approximation errors or inaccurate prior settings, leading to underestimated uncertainty or poor calibration. To address this, this paper proposes a novel Bayesian neural network uncertainty quantification method incorporating posterior regularisation. By introducing regularisation terms based on information theory or task-specific knowledge into the variational objective function, this approach constrains the shape of the posterior distribution, thereby guiding the model to learn more accurate and better calibrated uncertainty estimates.
  Keywords: Bayesian; neural network; a posteriori regularisation; uncertainty quantification.
  DOI: 10.1504/IJRIS.2026.10077733
   
  
• Research on deep learning-based evaluation mechanisms for ideological and political education  
  by Yaxin Li, Xiaojuan Zhang 
  Abstract: As the process of educational informatisation continues to deepen, traditional ideological and political education evaluation mechanisms face challenges in terms of dynamism, precision, and scientific rigor. This study focuses on applying deep learning technology to innovatively construct an evaluation system for ideological and political education. It aims to achieve multidimensional, intelligent analysis of the learning process through methods such as neural networks and natural language processing. The study explores a hybrid evaluation model integrating student behavioural data, textual sentiment analysis, and cognitive feedback to quantitatively assess educational outcomes, dynamically track ideological development, and provide data support for personalised teaching interventions. Results indicate that deep learning assisted evaluation mechanisms effectively enhance objectivity, real-time responsiveness, and predictive capabilities, offering a viable technical pathway and practical reference for advancing the precision and scientific rigor of ideological and political education.
  Keywords: deep learning; supportive ideological and political education; educational evaluation mechanisms; political education.
  DOI: 10.1504/IJRIS.2026.10078009
   
  
• Subset predicate encryption supporting a wholesaler  
  by Kamalesh Acharya 
  Abstract: Predicate encryption (PE) is a public key cryptographic primitive which gives fine-grained access structure in the cryptographic framework. Subset predicate encryption (SPE) is a variant of PE in which the decryption of the ciphertext corresponding to set S will be successful by using a secret key corresponding to set T if T S. In this work, we introduce subset predicate encryption supporting a wholesaler (SPeW), a new variant of SPE that incorporates an additional entity called as the wholesaler who purchases content in bulk and distributes it to a group of subscribers while preserving fine-grained cryptographic access control. Unlike previous SPE, SPeW provides a three-layered security guarantee: group privacy, a bound on group size, and security against illegal users. Our construction achieves adaptive security in the standard model while maintaining constant-size ciphertexts and secret keys, a property not achieved simultaneously in prior work. We implement SPeW using the PBC library and benchmark its performance across subscriber group sizes ranging from 1,200 to 1,800. The results show that setup and key generation, verification, and encryption times remain almost constant with group size, while decryption and group token generation scale linearly with group size.
  Keywords: subset predicate encryption; SPE; constant-size ciphertext; wholesaler; group privacy; bound on group size; security against illegal users.
  DOI: 10.1504/IJRIS.2026.10078084
   
  
• Construction and application research of intelligent analysis model for course management data based on deep learning  
  by Qian Ma, Yunqiao Peng 
  Abstract: This study aims to construct a deep learning-based intelligent analysis model for course management data and explore its practical application value. The research first performs pre-processing and feature engineering on multi-source heterogeneous data within course management, establishing a high-quality data foundation for model training. Subsequently, tailored to the characteristics of educational data, a hybrid deep learning model integrating CNN and LSTM Networks is designed and constructed. This model effectively captures local features and long-term temporal dependencies within the data, enabling precise analysis for critical tasks such as academic performance early warning, learning behaviour pattern recognition, and course teaching quality evaluation.
  Keywords: deep learning; course management; data intelligence analysis; teaching management.
  DOI: 10.1504/IJRIS.2026.10078085
   
  
• A method for detecting ablation in high-voltage cable buffer layers based on convolutional neural networks with waveform feature fusion  
  by Kang Guo, Zhengping Wang, Zhibo Tian, Qian Li, Siying Wang, Zhenwei Yang, Dongmei Sun 
  Abstract: To address the challenge of effectively identifying erosion defects in the aluminium sheath-buffer layer of high-voltage cables, this paper proposes an intelligent detection method based on waveform feature fusion convolutional neural networks (CNNs). By extracting multidimensional time-frequency domain features from oscillation wave test (OWTS) response signals, a fusion waveform feature vector is constructed. A one-dimensional convolutional neural network is designed to perform adaptive deep feature learning and state classification. Experimental results demonstrate that this method can accurately distinguish between normal, minor burn-through, and severe burn-through states, achieving a detection accuracy of 98.2%. This performance significantly outperforms traditional signal analysis methods, providing a reliable technical approach for the early diagnosis of buffer layer burn-through faults in high voltage cables.
  Keywords: waveform characteristics; convolutional neural networks; CNNs; power cables; buffer layer; ablation detection.
  DOI: 10.1504/IJRIS.2026.10078340
   
  
• Artificial intelligence empowering ideological and political education: research on scenario applications, risk regulation, and ethical boundaries  
  by Wen Jiang, Yan Lin 
  Abstract: This paper focuses on innovative practices in AI-empowered ideological and political education, systematically exploring its application pathways in scenarios such as intelligent teaching, personalised tutoring, and public opinion analysis. The study indicates that AI technology significantly enhances educational effectiveness through precise content delivery and innovative interaction models, yet simultaneously faces multiple risks including algorithmic bias, data privacy concerns, and lack of emotional interaction. In response, the paper proposes establishing a risk regulation system encompassing technological oversight, ethical review, and legal safeguards. It further delineates the ethical boundaries of human-machine collaboration, emphasising the imperative to uphold educational autonomy, value-driven leadership, and humanistic warmth. This framework provides theoretical reference and practical guidance for the healthy development of ideological and political education in the AI era.
  Keywords: artificial intelligence empowerment; ideological and political education; scenario applications; risk regulation; ethical boundaries.
  DOI: 10.1504/IJRIS.2026.10078342
   
  
• Multi-attribute decision-making and dynamic threshold-based sequencing of start-up power sources and stability control strategy for isolated microgrid start-up  
  by Jinhao Shen, Hua Zhang, Xueneng Su, Yinwen Gao, Kun Zheng, Cheng Long, Hangqian Hou 
  Abstract: This paper addresses the black start challenge in isolated microgrids by proposing a strategy that integrates multi-attribute decision-making with dynamic threshold adjustment for power source sequencing and stable control during the start-up process. First, a comprehensive evaluation index system for start-up power sources is established by considering multiple attributes such as power capacity, response speed, and regulation capability. Multi-attribute decision-making methods are then applied to rank candidate power sources. Building upon this foundation, a dynamic threshold control mechanism is designed to adaptively adjust critical operational parameter thresholds based on the microgrids real-time status and load restoration progress, achieving dynamic optimisation and active stability control during the startup process. Simulation results demonstrate that the proposed strategy can scientifically select the optimal startup power source sequence, effectively enhancing the stability and restoration efficiency of the microgrid startup process.
  Keywords: multi-attribute decision-making; dynamic threshold; isolated microgrid; start-up power source sequencing.
  DOI: 10.1504/IJRIS.2026.10078343
   
  
• Distributed model predictive control method for resource clusters with communication delays and resource  
  by Jinhao Shen, Hua Zhang, Xueneng Su, Yinwen Gao, Kun Zheng, Cheng Long, Hangqian Hou 
  Abstract: To address the common challenges of communication delays and resource heterogeneity in distributed resource clusters, this paper proposes a novel distributed model predictive control method. By designing a collaborative predictive control framework with asynchronous update and delay compensation mechanisms, the proposed method effectively mitigates the impact of non-ideal communication between heterogeneous nodes. Simultaneously, an adaptive resource optimisation strategy is introduced, enabling each heterogeneous node to dynamically adjust the complexity of its local predictive model and control law based on its computational and storage capabilities. Theoretical analysis and simulation results demonstrate that this approach significantly enhances the control robustness and operational efficiency of heterogeneous clusters under delayed conditions, while ensuring system stability and optimised performance.
  Keywords: communication delay; resource heterogeneity; distributed resources; clustered distributed; predictive control.
  DOI: 10.1504/IJRIS.2026.10078344
   
  
• AI+ journalism project driven teaching: integrated media content production practice incorporating neural  
  by Qian Ma, Chune Shen 
  Abstract: This paper explores a project-driven teaching model for AI+ journalism, focusing on the practical application of neural network algorithms in converged media content production. Guided by real-world projects, this model integrates intelligent technologies throughout the entire news workflow from gathering and writing to editing and distribution. By establishing a teaching framework that merges algorithmic models with journalistic practice, it cultivates students abilities to utilise AI tools for data mining, automated writing, and personalised content creation. Practice demonstrates that this interdisciplinary teaching approach not only enhances the efficiency and innovation of news production but also provides a practical pathway for cultivating journalism talent in the era of converged media. It holds significant promise for advancing the intelligent transformation of the journalism industry.
  Keywords: AI+ news; drive-based teaching; convolutional neural networks; converged media content.
  DOI: 10.1504/IJRIS.2026.10078345
   
  
• Analysis based on recommendation algorithm logic and Marxist alienation  
  by Weisong Wang, Wenjing Sun 
  Abstract: This paper examines the inherent logic and societal impact of recommendation algorithms in the digital age from the perspective of Marxist alienation theory. The study reveals that through datafication, personalisation, and behavioural steering, recommendation algorithms gradually strip users of autonomy in information acquisition and value judgment. Users become trapped within platform-driven information silos and consumption cycles, intensifying multiple forms of alienation manifesting in labour, products, human essence, and interpersonal relationships. Through theoretical analysis and case studies, the paper reveals how algorithms, beneath their facade of efficiency enhancement, conceal a logic of subjectivity dissolution and capital control. It ultimately advocates for constructing an algorithmic governance pathway centred on the comprehensive development of humanity, thereby steering technological applications back to serving their social essence: the human being.
  Keywords: recommendation algorithms; logic; Marxism; alienation theory.
  DOI: 10.1504/IJRIS.2026.10078522
   
  
• Exploration of teaching reform in innovation and entrepreneurship courses at higher education institutions based on knowledge graphs  
  by Mingzhe Li, Jinghuan Zhu 
  Abstract: This paper explores pathways and methodologies for applying knowledge graph technology to teaching reform in innovation and entrepreneurship courses at higher education institutions. It addresses practical challenges in such courses including insufficient industry-education integration, fragmented knowledge systems, and monolithic evaluation mechanisms while aligning with the trend toward educational digital transformation. The study proposes a knowledge graph-based teaching reform framework encompassing course knowledge system construction, personalised learning path recommendations, and dynamic teaching effectiveness assessment. This aims to systematise teaching content and refine learning processes. Finally, the paper summarises the practical value of knowledge graph-driven teaching model innovation, providing theoretical references and practical directions for enhancing the quality of innovation and entrepreneurship education.
  Keywords: higher education institutions; innovation and entrepreneurship courses; knowledge graphs; teaching reform; industry-education integration.
  DOI: 10.1504/IJRIS.2026.10078523
   
  
• Research on real time 3D reconstruction method for distribution network channels using pseudo-stereo vision and SLAM-based UAVs  
  by Jiaxing Fu, Xiangdong Zu, Changyong Huang, Chuang Yu, Hai Zhao 
  Abstract: This paper investigates autonomous UAV flight and precise control technologies for vegetation encroachment inspection in distribution network corridors. Aiming at the complex corridor environment characterised by narrow spaces and multiple obstacles, this research focuses on achieving fully autonomous and high-precision inspection operations. The core contributions include: firstly, developing a multi-sensor data fusion perception method based on onboard vision and LiDAR to reconstruct high-precision 3D spatial positions and distributions of targets such as conductors, utility poles, and trees in real-time, thereby constructing local environment maps for autonomous navigation. Secondly, designing an online planning algorithm based on corridor geometric constraints and inspection objectives to dynamically compute optimal shooting positions, safe flight trajectories, and gimbal attitudes that comprehensively cover key vegetation encroachment areas while ensuring image quality and completeness.
  Keywords: drone inspection; visual SLAM; pseudo-stereoscopic 3D reconstruction; pseudo-stereoscopic vision system.
  DOI: 10.1504/IJRIS.2026.10078524
   
  
• Research on generating and recommending personalised professional development pathways for teachers based on practice communities and instructional  
  by Can Wang 
  Abstract: This study focuses on constructing personalised professional development pathways for teachers by integrating communities of practice (CoP) with instructional behaviour data. Through systematic analysis of interactive discussions and experience sharing within CoPs, alongside routine instructional behaviour data (such as classroom recordings, student feedback, and assessment results), a multidimensional competency profile for teachers is established. Building upon this foundation, data mining and intelligent recommendation technologies are employed to generate and recommend personalised professional development pathways tailored to each teacher's developmental stage and practical needs. The research aims to overcome the limitations of traditional one-size-fits-all training models by providing teachers with precise, dynamic, and sustainable professional growth support. This approach seeks to enhance the effectiveness of teacher professional development while exploring theoretical and practical pathways for data-driven innovation in teacher development models.
  Keywords: practice communities; teaching behaviour data; teacher personalisation; professional development pathways.
  DOI: 10.1504/IJRIS.2026.10078619
   
  
• Research on path identification and effect evaluation of deep integration between digital technology and the real economy  
  by Huanxia Xue, Jing Ji 
  Abstract: This paper focuses on key issues concerning the deep integration of digital technologies with the real economy, systematically exploring the pathways and comprehensive effects of such convergence. The study first identifies primary integration pathways, including technological empowerment, platform ecosystem construction, data-driven innovation, and organisational process restructuring. Building upon this foundation, a multidimensional evaluation framework is constructed to empirically analyse the significant positive effects of the integration process on industrial efficiency, innovation models, employment structures, and the quality of economic growth. Concurrently, potential challenges such as transformation risks and the digital divide are identified. Finally, the study proposes targeted policy and strategic recommendations aimed at providing theoretical foundations and practical references for promoting the organic integration of digital technology and the real economy, thereby fostering high-quality economic development.
  Keywords: digital technology; real economy; deep integration; path identification; effect evaluation.
  DOI: 10.1504/IJRIS.2026.10078946
   
  
• Fileless malware attack detection based on multimodal large language models and verifiable reward reinforcement learning  
  by Biao Liang, Yongming Chen, Wanling Zhao, Yongxing Lai, Mingjie Xu, Ge Jin 
  Abstract: With the rapid advancement of smart grid construction and the integration of distributed energy resources, large-scale deployment of edge devices in power information systems is increasingly threatened by the surge of fileless attacks, creating an urgent need for lightweight detection techniques to ensure edge security. To address these challenges, we propose a lightweight, multimodal large language model framework for fileless attack detection, augmented with reinforcement learning guided by verifiable rewards. Memory dumps from active execution are converted into multimodal textual and visual representations for model training. Efficiency is improved via visual encoder compression, image patching, and attention pruning, while robustness is enhanced through group-relative policy optimisation. Extensive experiments on real-world datasets demonstrate that our approach achieves superior detection accuracy while maintaining practical efficiency, enabling real-time deployment on edge devices.
  Keywords: fileless malware attack detection; multimodal large language models; MMLMs; pruning; group-relative policy optimisation; reinforcement learning.
  DOI: 10.1504/IJRIS.2026.10079165
   
  
• Research on an optimisation model for power grid companies electricity purchase and sale portfolio targeting multi-variety power transactions  
  by Le Liu, Mingyu Che, Wenbo Yang, Wen Song, Ling Li, Xitong Ma, Jiafan Yang 
  Abstract: In response to the increasing complexity of power purchase and sale decisions faced by grid companies amid the growing diversity of electricity market trading products, this paper constructs an optimisation model for power purchase and sale portfolio management tailored to multi-product electricity trading. The model comprehensively considers various transaction types, including medium- and long-term contract trading, spot market trading, and ancillary service trading. It aims to minimise the companys power purchase costs while maximising power sales revenue, incorporating constraints such as power balance, market pricing, and risk limitations to establish a stochastic optimisation model. By quantifying market volatility risk through the value at risk (VaR) metric and employing linear programming with scenario analysis for model solution, empirical research demonstrates that the proposed model effectively optimises power grid enterprises electricity purchase and sale portfolio strategies in multi market environments. It significantly enhances overall operational efficiency while controlling risk.
  Keywords: multi-product electricity trading; power grid companies; electricity purchase and sale; portfolio optimisation.
  DOI: 10.1504/IJRIS.2026.10079987
   
  
• Research on hyper-personalised learning path generation algorithms tailored to cognitive characteristics in educational ecosystems  
  by Qin Kong, Luhong Tang, Yihua Li 
  Abstract: This paper investigates hyper-personalised learning path generation algorithms tailored to the cognitive characteristics of educational ecosystems. Addressing issues such as the homogeneity and lack of dynamic adaptability in traditional learning paths, it proposes a multidimensional modelling approach that integrates learner cognitive states, knowledge graphs, and behavioural data. By analysing individual cognitive differences and group interaction patterns within educational ecosystems, it constructs dynamic learner profiles and designs a hybrid algorithm based on reinforcement learning and sequence recommendation. This enables real-time personalised generation and optimisation of learning paths. Experiments demonstrate that this algorithm effectively enhances learning efficiency and knowledge mastery, providing theoretical support and technical pathways for precision teaching within smart education environments.
  Keywords: educational ecosystem; cognitive characteristics; hyper-personalised learning; path generation.
  DOI: 10.1504/IJRIS.2026.10080274
   
  
• Innovation and practice evaluation of journalism and communication course teaching models based on cultural identity construction  
  by Li Cai, Wei Shang 
  Abstract: This study centres on cultivating cultural identity as its core objective. Addressing the challenges of weak cultural substance and disconnect between theory and practice in journalism and communication courses, it establishes an integrated teaching model comprising four interconnected dimensions: theoretical foundation practical empowerment cultural immersion identity construction. Through layered integration of cultural content, innovative immersive teaching methods, multi-dimensional practical platforms, and interdisciplinary faculty teams, it deeply embeds Chinas outstanding traditional culture, revolutionary culture, and advanced socialist culture throughout the entire teaching process. A diversified assessment system combining quantitative and qualitative measures, with equal emphasis on process and outcomes, is established across three dimensions: professional competence, cultural identity, and practical achievements. Practice demonstrates that this model effectively achieves the coordinated development of professional skill cultivation and cultural literacy shaping, providing a replicable teaching paradigm for nurturing journalism and communication talent in the new era.
  Keywords: cultural identity construction; journalism and communication studies; curriculum and instruction; model innovation; practice evaluation.
  DOI: 10.1504/IJRIS.2026.10081488
   
  
• Research on reinforcement learning-based algorithms for generating personalised financial education pathways  
  by Xiaorui Xue, Haiyan Liu 
  Abstract: This paper addresses the limitations of traditional one-size-fits-all financial education models by proposing a reinforcement learning-based approach for generating personalised financial education pathways. The research models financial knowledge graphs and learners dynamic states such as knowledge levels, interest preferences, and learning objectives as Markov decision processes. By designing reward functions that balance knowledge logic with learning experience, the method guides agents to automatically generate and continuously optimise personalised learning sequences through interactions with the learning environment. The paper focuses on addressing key challenges in financial education scenarios, including sparse state representations, delayed rewards, and path interpretability, while comparing the adaptability of different reinforcement learning algorithms for this task. Simulation experiments demonstrate that this method achieves significant improvements in knowledge mastery and learning efficiency compared to traditional recommendation systems. This research provides a feasible algorithmic framework and practical reference for constructing adaptive, scalable intelligent financial education systems.
  Keywords: education pathways; reinforcement learning; generating personalised; LightGBM.
  DOI: 10.1504/IJRIS.2026.10081580
   
  
• Research on fluctuation analysis of new energy sources in power systems and power balance optimisation based on big data  
  by Wei Liu, Gang Wu, Fuyao Deng, Ruiguang Ma, Tiannan Ma, Chuan He, Hao Luo, Fengjiao Xu 
  Abstract: A multi-source data analysis platform integrating power and meteorological data is first established, enabling refined modelling of wind and photovoltaic output characteristics through time-series decomposition and clustering algorithms. Based on the forecasted data, a power balance optimisation model is developed that incorporates demand-side response and energy storage coordination, achieving multi-objective optimisation of system balance and operation cost. Simulation results demonstrate that the proposed method effectively mitigates the impact of renewable fluctuations on power dispatching, enhances system economy and reliability, and provides strong technical support for the development of modern power systems.
  Keywords: big data technology; renewable energy fluctuation; power balance; optimisation scheduling.
  DOI: 10.1504/IJRIS.2026.10081581
   
  
• Adaptive inertias and voltage support technology for microgrid enhanced by synergy of virtual energy storages and distributed energy resources  
  by Hui Luo, Xinyi Huang, Yuqian Mo, Jia Liu, Li Tao, Yu Wang 
  Abstract: To address the issues of reduced microgrid inertia and increased voltage fluctuations caused by high penetration of distributed energy resources, this paper proposes an adaptive inertia and voltage support technology based on multi-source coordination. First, the physical mechanisms of distributed power source grid-connected converters in inertia response and voltage regulation are analysed. Second, an adaptive control strategy considering state-of-charge and output margin is designed, enabling energy storage, PV, and wind turbines to dynamically adjust output power based on system frequency change rate and voltage deviation, thereby simulating the inertia and reactive power support characteristics of synchronous generators. Finally, validation is conducted through a microgrid simulation model incorporating multiple types of distributed energy resources.
  Keywords: distributed energy; synergistic enhancement; microgrid adaptability; inertia and voltage support.
  DOI: 10.1504/IJRIS.2026.10081807
   
  
• Anomaly detection and resilience optimisation in electricity spots markets based on spatiotemporal-graph convolution and learning application  
  by Miao Wang, Xiaoming Yang, Wentao Sun, Yanyun Zhang, Yubo Jia 
  Abstract: This paper proposes an intelligent detection and optimisation framework integrating spatio-temporal graph convolutional networks with meta-learning to address the increasingly complex operational environment of electricity spot markets. The method first employs spatio-temporal graph convolutional networks to capture complex spatio-temporal correlation features among multiple entities in the power market, enabling precise identification of abnormal behaviours such as price anomalies and market manipulation. Building upon this foundation, a meta-learning mechanism is introduced to allow the model to rapidly adapt to dynamic market changes based on limited historical samples, thereby enhancing its generalisation and adaptive capabilities. Furthermore, an anomaly detection-based market resilience optimisation model is constructed. By adjusting market clearing strategies and risk control mechanisms, this model enhances market resilience and operational efficiency in response to abnormal disturbances.
  Keywords: Spatio-temporal convolutional networks; meta-learning applications; electricity spot markets; anomaly detection; resilient optimisation.
  DOI: 10.1504/IJRIS.2026.10081808
   
  
• A study on methods for the automatic statistical analysis and precise assessment of distribution network reliability for low-voltage direct connection customers  
  by Mingli Shi, Jianwei Ye, Yongyi Yang, Zihang Jiao, Haodong Bai, Tingting Sun, Kunyang Gao, Yang Xu, Pengfei Wang 
  Abstract: In response to the lack of automatic identification of low-voltage individual households (i.e., low-voltage single-phase/three-phase users) in distribution network reliability statistics, and the difficulty in accurately matching household-level power outage events, this paper proposes a method for automatic reliability statistics and precise assessment based on the fusion of multi-source data from the electricity consumption data acquisition system and the distribution automation system. Firstly, an automatic identification model for power outage events at low-voltage individual households is constructed, utilising dynamic correction rules based on voltage sags, outage duration and the relationship between households and transformers to achieve accurate extraction of household-level power outage events. Through validation using actual transformer substation case studies, this method has been shown to effectively enhance the completeness of low-voltage household reliability statistics and the precision of assessment results, thereby providing decision-making support for targeted investment and operation and maintenance of distribution networks.
  Keywords: low-voltage residential customers; distribution network reliability; automatic statistics; multi-source data fusion; combined weighting.
  DOI: 10.1504/IJRIS.2026.10081981
   
  
• Research on interpretable risk identification of power grid dispatching instructions-integrating domain knowledges and reinforcement  
  by Zhihua He, Peng Zhang, Beijie Liao, Qiang Lu, Juelong Li, Shudi Liu, Rong Hu, Qian Gao 
  Abstract: To address the security risks posed by the black-box decision-making of deep reinforcement learning in power grid dispatch, this paper proposes an explainable risk identification framework that integrates domain knowledge with reinforcement learning. First, physical laws of the power grid and expert dispatch rules are embedded into the reinforcement learning model to construct a knowledge-guided decision space, constraining the agents exploration direction. Second, an attention-based risk attribution module is designed to dynamically assess vulnerabilities under fluctuating renewable energy output and load conditions, thereby quantifying operational risks. Finally, simulations of typical scenarios validate that this approach enhances the robustness of dispatch strategies while providing clear explanations for decision-making, offering dispatchers reliable auxiliary information.
  Keywords: domain knowledge integration; reinforcement learning; RL; explainability; grid dispatch.
  DOI: 10.1504/IJRIS.2026.10081982