Forthcoming Articles
International Journal of Information and Communication Technology
Forthcoming articles have been peer-reviewed and accepted for publication but are pending final changes, are not yet published and may not appear here in their final order of publication until they are assigned to issues. Therefore, the content conforms to our standards but the presentation (e.g. typesetting and proof-reading) is not necessarily up to the Inderscience standard. Additionally, titles, authors, abstracts and keywords may change before publication. Articles will not be published until the final proofs are validated by their authors.
Forthcoming articles must be purchased for the purposes of research, teaching and private study only. These articles can be cited using the expression "in press". For example: Smith, J. (in press). Article Title. Journal Title.
Articles marked with this shopping trolley icon are available for purchase - click on the icon to send an email request to purchase.
Online First articles are also listed here. Online First articles are fully citeable, complete with a DOI. They can be cited, read, and downloaded. Online First articles are published as Open Access (OA) articles to make the latest research available as early as possible.
Articles marked with this Open Access icon are Online First articles. They are freely available and openly accessible to all without any restriction except the ones stated in their respective CC licenses.Register for our alerting service, which notifies you by email when new issues are published online.
International Journal of Information and Communication Technology (30 papers in press) Regular Issues
Abstract: In order to meet the demand for the digital modelling and intelligent analysis of students learning behaviour, a predictive evaluation framework is proposed, which combines feature engineering with LightGBM. Through the construction of tags and the extraction of training samples from college English course behaviour log, the construction strategy and interpretable output mechanism of the model are discussed. The results show that the proposed model achieves an F1-score of 0.889 on the test set, with an average response latency of less than 230 ms, and a performance fluctuation of less than 1.6% in 5 times cross validation, demonstrating robust deployment stability and generalisation. Keywords: learning behaviour prediction; LightGBM algorithm; educational data mining; Intelligent teaching analytics; learning behaviour modelling. DOI: 10.1504/IJICT.2027.10080149
Abstract: To improve the accuracy of key indicator evaluation in power engineering, this study proposes a comprehensive evaluation method based on K-nearest neighbour (KNN) algorithm. Aiming at the limitations of traditional methods in processing nonlinear and high-dimensional data, an evaluation system including ten key indicators such as engineering cost, construction period, and equipment performance was constructed to conduct experiments on 50 power engineering project data in a certain region. The results show that compared with traditional methods, the KNN model has a 15% improvement in evaluation accuracy, a more reasonable allocation of indicator weights, and can effectively identify key factors affecting engineering quality and efficiency. It also exhibits strong robustness in dealing with nonlinear problems and is suitable for power engineering of different scales and types. This method provides a scientific basis for decision-making in power engineering and has promotional value. Keywords: KNN algorithm; power engineering; index analysis; comprehensive evaluation. DOI: 10.1504/IJICT.2026.10080314
Abstract: In screening close contacts of tuberculosis patients, current approaches classify contacts as close or general based solely on exposure duration, ignoring environmental ventilation, contact distance, and spatial distribution. Many contacts with subclinical lesions are missed while low-risk individuals undergo unnecessary imaging. This paper proposes the Exposure weighted lesion profiling framework, which uses internet of things positioning data to reconstruct individualised exposure doses and guide computed tomography screening. A prospective cohort of 214 close contacts showed that the area under the curve for predicting subclinical lesions using ventilation adjusted exposure doses was 0.91, superior to 0.68 of traditional binary classification. Prioritising screening of the top 20% of contacts by this framework increased lesion detection rate from 8.2% to 23.6% and reduced number needed to screen to detect one positive case from 12.6 to 5.9. This framework fills the gap from binary contact classification to continuous dose-response models. Keywords: contact investigation; exposure dose; subclinical tuberculosis; internet of things geolocation. DOI: 10.1504/IJICT.2026.10080479
Abstract: Traditional agricultural financing models fail to fully integrate multi dimensional data for accurate energy management and risk assessment. This paper proposes a precision agricultural financing model based on digital energy data analysis and risk assessment. We develop an innovative decision support platform that integrates digital twin technology to build a photovoltaic agricultural scenario and a microgrid energy dispatching simulation system. Furthermore, a spatiotemporal attention Transformer model combined with a multi-dimensional risk factor federated learning fusion mechanism is proposed to accurately predict and assess financing risks. Experimental results show that energy costs constitute 49.7% of total operating costs, and the grid's stable power supply time reaches 94.7%. However, due to equipment aging, the actual battery life of energy storage equipment is only 86.0% of its theoretical 40.0 hour capacity. This approach significantly enhances financing decision-making accuracy and microgrid dispatching efficiency. Keywords: precision agriculture; financing model; digital energy; risk assessment. DOI: 10.1504/IJICT.2026.10080605
Abstract: To enable efficient real-time prediction of grouting behaviour in karst cavities, this study proposes a physics-informed framework integrating phase-field multiphase flow simulation with a convolutional neural network (CNN). The phase-field model captures grout-groundwater interfacial evolution and provides physical constraints for data-driven prediction. A dataset of 1,000 samples covering 100 operational scenarios was generated using Latin hypercube sampling, and an encoder-decoder CNN was trained after hyperparameter optimisation. The model achieved a root mean square error of 0.028 and a structural similarity index of 0.942 on the test set. Physical model tests further showed that prediction errors were below 0.2 cm for diffusion radius and 0.5 kPa for pressure field. The proposed framework markedly improves prediction efficiency and accuracy while maintaining physical consistency, providing an effective tool for grouting design and real-time decision-making in karst engineering. Keywords: phase field method; convolutional neural network; cave grouting; multiphase flow simulation; coupling model; diffusion prediction. DOI: 10.1504/IJICT.2026.10080606
Abstract: This study develops a U-Net with transfer learning framework for fine-grained classification and health assessment of urban green spaces using GF-2 remote sensing imagery. The two-level classification includes pixel-level land cover (arbour, shrub, herb, water, impervious surface) and patch-level functional types (park, protective, residential, roadside green space). Health assessment quantifies vegetation coverage, species richness, vitality, landscape pattern, and ecological service potential from vegetation structure to ecological function. The U-Net encoder incorporates MobileNetv2 and an attention mechanism to improve boundary recognition accuracy. Transfer learning from ImageNet pretrained models addresses insufficient samples and cross-domain adaptation, reducing convergence epochs from 120 to 35 and total training time from 84 to 24.5 hours under identical experimental conditions. Comparative experiments on 1,369 preprocessed 256 Keywords: U-Net; transfer learning; green space classification; health assessment; remote sensing imagery. DOI: 10.1504/IJICT.2026.10080607
Abstract: Personalised learning paths are critical in nursing education, yet traditional algorithms struggle to adapt to students dynamic cognitive states. To address this, this study proposes a novel framework integrating recurrent neural networks (RNN) and reinforcement learning (RL). The RNN continuously encodes temporal learning histories to capture evolving knowledge states. Subsequently, the RL agent utilises these latent states to formulate dynamic learning paths, maximising long-term clinical skill retention via a custom reward function. Extensive experiments on a real-world nursing dataset demonstrate that our RL-RNN framework achieves an AUC-ROC of 0.812 and an average cumulative reward of 19.3, outperforming state-of-the-art predictive and RL baselines. Furthermore, it significantly reduces the average instructional steps required to reach clinical mastery to 31.8. Ablation and case studies validate the frameworks practical viability for dynamic curriculum generation, ensuring efficient medical training. Keywords: reinforcement learning; RL; recurrent neural networks; RNN; personalised learning path; nursing education; intelligent tutoring systems; knowledge tracing. DOI: 10.1504/IJICT.2026.10080608
Abstract: The fast-changing digital media systems have enhanced the necessity to have smart visual communication systems that comprehend and react to human sentiments. This paper proposes a new deep sentiment-aware visual communication framework (DSVCF), which combines deep learning, affective computing, and adaptive communication control to improve human-machine communication. The proposed framework works in three interdependent modules: 1) a multimodal sentiment perception network (MSPN), which combines visual, textual and auditory information with the help of a dual-channel transformer convolutional neural network (DT-CNN) to extract emotion-rich representations; 2) a adaptive emotion modulation layer (AEML), which uses a dynamic attention generative mechanism (DA-GM) to control visual content presentation depending on the detected affective states; 3) an optimised communication control module. The DSVCF fills the gap between the understanding of sentiment and visual content optimisation, enabling digital systems to produce emotionally resonant messages and communications that relate to the context. Not only does this framework promote the affective deep learning to applications within digital communication systems, but it also forms a basis of sentiment-based control architecture in multimedia environments. The suggested model is both theoretically and practically valuable as it promotes the responsiveness of sentiment, interpretability, and effectiveness of communication within the intricate digital ecosystem. Keywords: deep learning; visual communication; sentiment analysis; affective computing; reinforcement learning; digital media systems. DOI: 10.1504/IJICT.2026.10080609
Abstract: As the pace of digital transformation accelerates, the demand for graphic design is growing exponentially. Traditional design workflows are highly labour-intensive, inefficient, and ill-suited to meet diverse creative needs. To tackle these issues, this paper introduces a generative artificial intelligence-based model for optimising graphic design workflows. Built upon generative adversarial networks, the model incorporates a self-attention mechanism to enhance semantic relationships between design elements. Leveraging latent space search, this paper presents an iterative framework that optimises design workflows, enabling designers to adjust layouts with constraints on partial attributes. Experimental findings demonstrate that the proposed model lowers the overlap rate to 0.17 and attains an alignment score of 0.64, substantially surpassing the baseline method. Ablation studies confirm the efficacy of the models key components. By effectively boosting design efficiency and layout rationality, this research provides a new technical pathway for the optimisation of graphic design workflows. Keywords: graphic design; process optimisation; generative adversarial network; attention mechanism; latent space search. DOI: 10.1504/IJICT.2026.10080610
Abstract: To address engineering challenges in developing IP derivatives for anime short dramas under mobile internet contexts such as lagging demand mining and subjective design adaptation this paper proposes an intelligent adaptation design framework based on multimodal sentiment computing and generative artificial intelligence (AIGC). First, an unstructured data mining model is constructed using Python web crawlers and natural language processing (NLP) technologies to collect and clean massive user bullet-screen comments and reviews. Through TF-IDF algorithms and K-means clustering, users vague instantaneous emotions are quantified into precise design feature parameters. Second, a visual attention analysis algorithm based on sliding windows is proposed. By employing peak detection, it accurately extracts visual anchors in emotionally weighted segments of short dramas, establishing a feature mapping mechanism from IP morphology to e-commerce product design. Finally, a generative design workflow leveraging stable diffusion and LoRA technologies is developed. Through geometric constraints using ControlNet, it achieves automated, high-fidelity integration of IP visual features with e-commerce product forms. Empirical studies demonstrate that this method significantly improves design-user emotion alignment accuracy (accuracy) and substantially shortens product development cycles, providing a quantifiable engineering pathway for commercialisation of short video IPs. Keywords: anime short dramas; vertical screen visual rhetoric; emotional compensation; e-commerce product adaptation; visual anchors; value co-creation. DOI: 10.1504/IJICT.2026.10080611
Abstract: The generation of tourism and cultural symbols often faces the problem of semantic drift where patterns visually conform to the data distribution but deviate from their original cultural connotations. To address this, this paper integrates graph neural networks with diffusion models, utilising a graph attention network to encode semantic relationships between symbols and using these relationships to guide the diffusion process, thereby ensuring that the generated symbols balance visual realism with cultural fidelity. Tested on 15,327 symbolic images from seven world heritage sites in China, our method achieved area under the curve and accuracy scores of 0.912 and 0.894, respectively - representing improvements of 6.5% and 7.6% over the standard diffusion model and 4.3% and 4.9% over the current state-of-the-art diffusion adapter network. Experiments demonstrate that introducing structured constraints based on symbolic knowledge effectively suppresses semantic drift, providing a viable approach for the intelligent preservation of cultural heritage. Keywords: diffusion models; graph neural networks; GNNs; tourism cultural symbols. DOI: 10.1504/IJICT.2026.10080612
Abstract: Against the backdrop of the rapid development of the global film industry and digital media technology (DGMT), colour aesthetics has evolved into a core medium for film narrative and cross-cultural communication. Addressing the shortcomings of traditional film colour research, which relies on manual annotation and is susceptible to subjective bias, this study proposes an improved semantic segmentation framework (Gabor-HR-DeepLab) that integrates dual-stream weighted Gabor features and high-resolution neural networks. Through refined pixel segmentation and dynamic colour palette extraction of over 20,000 frames from the Eastern film series Ne Zha and the Western Marvel Cinematic Universe (MCU), the differences between the two in dimensions such as colour contrast, brightness variance, and harmony are quantitatively analysed. This research result profoundly reveals the differences between the localised cultural aesthetics representation and the standardised industrial visual strategies. Crucially, this study has made a significant contribution to information and communication technology (ICT) by introducing a robust automated digital media processing procedure; it provides quantifiable empirical evidence by combining computational algorithms with humanistic film analysis; and it offers an expandable computational framework for the digital animation industry. Keywords: deep learning; DL; semantic segmentation; information and communication technology; ICT; multimedia processing; colour language computing; cross-cultural film aesthetics; computer vision; CV. DOI: 10.1504/IJICT.2026.10080613
Abstract: The long-term safety, operational stability and economic efficiency of grid-connected wind power generation systems are of great significance to the accurate and continuous monitoring of wind turbine blade operating conditions. This paper proposed a new wind turbine blade acoustic monitoring scheme based on adaptive noise reduction, ulti-domain feature enhancement, and a two-stage fault detection method for non-contact blade condition assessment. The system directly addresses the critical technical gap of insufficient early fault capture capability and poor anti-interference performance of existing monitoring methods in complex and dynamic wind farm operating environments. Rigorous experimental validation shows that the system achieves 99.2% fault recognition accuracy, 47.6% improvement in anti-noise performance, 0.8-second early fault detection delay, and maintains a low false alarm rate of 1.2% under varying field operating conditions. Keywords: wind turbine blade; acoustic monitoring; fault detection; signal processing. DOI: 10.1504/IJICT.2026.10080614
Abstract: In the process of agricultural digital transformation, field sensor and remote sensing data are often mixed with interfering factors such as soil heterogeneity and management differences, disconnecting the causal chain between digital investment and new quality productivity. This paper proposes a dual causal inference framework, the digital transformation driven new quality productive forces analysis framework, integrating propensity score weighting and double machine learning to isolate the net effect of digitalisation from mixed data. Experiments on two public datasets with over 67,000 samples show that adopting digitalisation increases the new quality productivity index by 21.6% (95% confidence interval 17.8%25.4%). The framework improves causal entropy from 0.42 to 0.67 and reduces neighbourhood confounding degree from 0.31 to 0.13, indicating a residual confounding reduction of approximately 40%58%. This method fills the causal attribution gap and provides a quantifiable path for distinguishing correlation from causation. Keywords: causal inference; digital agriculture; new quality productive forces; double machine learning; DML. DOI: 10.1504/IJICT.2026.10080615
Abstract: High penetration of photovoltaic power poses a dual challenge to asynchronous power grids: insufficient inertia support and output fluctuations. Traditional strategies struggle to coordinate the complementary characteristics of photovoltaic, energy storage, and load resources on a millisecond-to-second timescale. This paper proposes a source-load-storage coordinated frequency regulation strategy: first, photovoltaic fluctuations are isolated using adaptive variational modal decomposition; then, model predictive control is employed to uniformly allocate the load to energy storage, photovoltaic derating, and shiftable loads. Simulation results on an enhanced 39-node IEEE system demonstrate that the peak frequency deviation is reduced from 0.32 Hz to 0.12 Hz, representing a 62.5% improvement; the fluctuation rate of photovoltaic grid-connected power decreases from 15.8% to 12.4%; and the average daily charge-discharge cycles of energy storage are reduced from 12 to 7. This strategy leverages the high inertia characteristics of asynchronous grids to create a time window for slow-responding loads. Keywords: high-inertia asynchronous grid; source-load-storage coordination; model predictive control; variational modal decomposition; frequency damping. DOI: 10.1504/IJICT.2026.10080616
Abstract: Existing design tools can recommend harmonious colour palettes yet fail to quantify the emotions a scheme evokes, leaving designers to rely on intuition rather than measurable feedback. This paper introduces a multimodal deep neural framework that continuously measures the affective qualities of design colours. The model fuses hierarchical visual features from design images with semantic intentions extracted from designer descriptions via a cross-attention mechanism. A perceptual colour-difference regularisation loss ensures physically plausible predictions, enforcing smooth emotional transitions consistent with human perception. Tested on three datasets, our method achieves a root-mean-square error of 0.0873 for emotion regression, a relative gain of 16.2% over the strongest baseline, and yields ratings statistically indistinguishable from those of professional designers. The framework bridges colour physics and design cognition, enabling emotion-aware design assistance. Keywords: colour emotion quantification; multimodal deep learning; design cognition; physical constraints; affective computing. DOI: 10.1504/IJICT.2026.10080617
Abstract: The sinking efficiency of dual-motor driven vibration pile hammers is significantly affected by soil nonlinear hysteretic characteristics, causing synchronisation difficulties. To address this issue, this paper presents a simulation study that establishes an electromechanical coupling dynamic equation based on the Bouc-Wen model and proposes a nonlinear synchronisation control algorithm. The simulation results show that soil nonlinearity causes periodic phase difference fluctuations of +-8 , with synchronisation time prolonged by 32.5% compared to the linear model. The nonlinear synchronisation control algorithm reduces synchronisation time to 3.2 seconds under soft soil conditions, converges the phase difference to within +-2 , and achieves 26% improvement over master-slave control. The sinking displacement decreases non-linearly with increasing soil stiffness, reducing by 41.2% when stiffness increases from 5 MN/m to 20 MN/m. This study reveals the quantitative relationship between soil nonlinearity parameters and synchronisation performance, providing a theoretical basis for intelligent control. Keywords: vibrating pile hammer; soil nonlinearity; dual-machine synchronisation; Bouc-Wen model. DOI: 10.1504/IJICT.2026.10080618
Abstract: The optimisation of enterprise talent structure faces challenges of high-dimensional and multiple constraints. Existing meta-heuristic methods mostly focus on a single objective and are unable to balance skills matching, generational balance, and strategic resilience. This paper proposes a person-job optimisation model based on particle swarm optimisation, which objectively assigns weights using the entropy weight method and introduces a penalty function to handle multi-objective conflicts. On a real human resource dataset from an enterprise, the area under the curve of person-job optimisation model based on particle swarm optimisation reached 0.94, which was 14.6% higher than that of traditional particle swarm optimisation; the talent-job matching degree increased from the initial 0.51 to 0.88, and the generational coupling coefficient reached 0.82. This model fills the gap in methods for the collaborative optimisation of multi-dimensional talent structures under the requirements of new quality productivity. Keywords: new quality productive forces; optimisation of talent structure; particle swarm optimisation match degree. DOI: 10.1504/IJICT.2026.10080619
Abstract: There is a systematic deviation between the disciplinary levels of university regulations and judicial judgements, but existing research lacks a quantitative identification method for the deviation patterns. This paper proposes a deviation degree analysis model based on hierarchical clustering. By constructing 15 indicators such as absolute deviation degree, relative deviation degree, and proportional deviation degree to form a multi-dimensional feature vector, a cluster analysis is conducted on the publicly available administrative judgement data (N = 658). The results show that the deviation patterns can be classified into five types, among which the upward deviation type accounts for 28.3%, with a silhouette coefficient of 0.76. Compared with the K-means clustering method, the method improves the intra-class compactness by 12.4%. This model provides a quantifiable analysis tool for the standardisation of university disciplinary discretion. Keywords: school disciplinary punishment hierarchy; judicial judgement deviation; hierarchical clustering; proportionality; quantitative discretion. DOI: 10.1504/IJICT.2026.10080620
Abstract: Vocal fold fatigue can cause organic lesions for a long time, and monitoring its evolution process is of great significance. The existing research mode single response is slow, static modelling is difficult to tolerate dynamic, poor interpretability and weak generalisation. This paper proposes physics-informed joint simulation and identification framework. The framework uses mathematical modelling of fatigue changes, simultaneous analysis of sound and vibration signals, and introduces the principle of sound to improve the authenticity of the signal, so as to realise the mutual optimisation of generation and recognition. Experimental results show that the signal simulation correlation coefficient of the framework is 0.92, the fatigue prediction determination coefficient is 0.86, and the F1 score is 0.89, which is significantly improved compared with the variational recurrent neural network. Keywords: vocal fold fatigue; timing simulation; multi-modal fusion; physical information deep learning; generative adversarial network. DOI: 10.1504/IJICT.2026.10080621
Abstract: With the wide application of multimedia technology in English teaching, how to effectively correlate the semantic information of cross-media data (such as text, audio and images) has become a key challenge. This paper aims to construct a cross-media semantic association model for college English teaching (ODBA-CMSA) based on output-driven hypothesis and bidirectional attention feature learning to improve the accuracy and interpretability of grading. The model optimises semantic representation through multi-modal embedding, bi-directional attention mechanism, and output-driven loss. The test results show that the model is significantly better than the baseline in accuracy rate (89.5%) and F1-score (89.0%), and its robustness (accuracy rate only decreases by 5.1% under noise) and practicability (user satisfaction 4.4 points) are outstanding, which verifies its effectiveness in complex teaching environment. This study not only provides an efficient semantic association solution, but also enhances the transparency of the model through interpretability design, which provides a new idea for the development of intelligent teaching system. Keywords: output drive; two-way attention; cross-media; college English; semantic association. DOI: 10.1504/IJICT.2026.10080438
Abstract: This study addresses the recognition challenges of Maonan ethnic Nuo masks under complex backgrounds, varying lighting conditions, and multi-scale environments by proposing a 'detection-identification-analysis' multimodal framework. The front-end employs an improved YOLOv10 model (integrating EMA and GELU), while the back-end leverages CLIP and text description augmentation. A MNMD dataset comprising 2,850 images across 28 fine-grained categories was constructed. The enhanced model achieved an accuracy of 84.10% on the test set, with parameters reduced by 40.43% compared to YOLOv8, providing an effective solution for the digital preservation of cultural heritage. Keywords: Maonan ethnic Nuo masks; multimodal recognition; YOLOv10; CLIP; lineage analysis. DOI: 10.1504/IJICT.2026.10080604
Abstract: The precise analysis of singing techniques is of vital importance for intelligent vocal education. The existing methods process singing by segmentation, which destroys the characteristics of continuous changes of techniques, resulting in poor recognition of dynamic techniques such as vibrato. To address this issue, this paper proposes a neural ordinary differential equations waveform modelling network. This framework treats singing as a continuous process and simulates the physical laws of vocal cord vibration, enabling the model to more naturally capture the subtle changes in techniques. Experiments show that the comprehensive recognition accuracy of this method reaches 0.874, and the comprehensive evaluation index reaches 0.932. Compared with the current optimal method, it has improved by 4.2% and 3.1% respectively. The recognition of trills has also improved by 12.3%. This research provides high-precision and interpretable technical support for intelligent vocal education. Keywords: neural ordinary differential equations; singing technique; physics-informed machine learning; continuous-time dynamic modelling. DOI: 10.1504/IJICT.2026.10080439
Abstract: Recent advances in artificial intelligence (AI) have enabled automated support for academic writing; however, many platforms remain 'tool-centric', providing isolated corrections rather than process-aware instructional support. This paper proposes a novel AI-enhanced academic writing instruction platform designed from an ICT perspective, focusing on the specific rhetorical needs of Sino-French and Sino-African cross-cultural academic communication. The framework integrates a three-layer architecture: a multi-level linguistic feature extraction layer using fine-tuned transformer models, a learner modelling layer driven by longitudinal data analytics, and an adaptive feedback engine. To validate the system, a 16-week empirical study was conducted with 342 undergraduate students. Analysis of a corpus comprising 1,250 academic essays reveals that the platform significantly enhances discourse coherence (18.4% improvement) and lexical complexity (12.6% increase) compared to traditional automated writing evaluation tools. Statistical significance was confirmed via a paired-sample t-test (p < 0.01), and effect size calculations (Cohen's d = 0.76) suggest a robust impact on student writing quality. The results demonstrate that embedding evaluation directly into the system operation enables a continuous feedback loop that fosters long-term linguistic development in specialised academic contexts. Keywords: artificial intelligence; academic writing instruction; learning analytics; intelligent educational systems; information and communication technologies; ICTs. DOI: 10.1504/IJICT.2026.10080315
Abstract: Existing gymnastics motion capture and simulation methods struggle with high-dynamic movements, rapid joint changes, and unnatural simulations. To address these pain points, we propose an end-to-end intelligent gymnastics movement analysis framework integrating deep learning spatiotemporal representation with 3D deformable transformers. We introduce a 3D deformable attention mechanism for limb tracking and a modal feature fusion strategy combining RGB and bone pose data. Additionally, a multi-objective optimisation framework with heuristic rules and parallel computing improves simulation authenticity and efficiency. Experimental results demonstrate high accuracy: the average 3D coordinate error for difficult churning is 35.6 mm, key pose recognition is 78%, and swivel tracking accuracy reaches 91.2%. The real-time tracking delay is as low as 12.3 ms, significantly outperforming traditional methods. Keywords: 3D vision transformer; multi-objective optimisation; gymnastics motion capture; action simulation; real-time performance. DOI: 10.1504/IJICT.2026.10080363
Abstract: Traditional business English emails generated by artificial intelligence are often rigid and lack flexibility, making it difficult to meet complex and variable communication needs. To address this issue, this paper explores the application potential of generative adversarial networks (GANs) in this field. It designs a new conditional generative adversarial network architecture, enabling the model to learn and imitate the appropriate style and accurate content of professional business letters. Experiments on an open business email dataset show that, compared with mainstream generation models, the method proposed in this paper improves the practicality of generated emails by 15.7% in human evaluation and increases the accuracy of key terms by 8.2%. This demonstrates that generative adversarial network technology can effectively 'reconstruct' the quality and adaptability of automatically generated business emails, providing a new path for enhancing the professionalism of intelligent communication tools. Keywords: generative adversarial network; GAN; business English email; natural language generation; style control. DOI: 10.1504/IJICT.2026.10080359
Abstract: Traditional manual English writing scoring is inefficient and subjective, while existing automated systems lack semantic understanding and instructional interpretability. This study proposes an intelligent human-computer collaborative scoring system integrating large-scale model representation learning with knowledge reasoning. The system combines representation, reasoning, and collaboration through a fine-grained feature learning algorithm and a domain knowledge map, enabling multidimensional evaluation and traceable scoring decisions. Experimental results on 8,800 English compositions demonstrate that the proposed system achieves 91.7% scoring accuracy (36.8% higher than traditional models), 72% lexical error detection coverage with a 5.2% omission rate, and a human-computer semantic consistency coefficient (SCC) of 43% in evaluating high-level dimensions like semantic coherence and argumentation logic. Keywords: large model representation learning; knowledge reasoning; English writing scoring; human-computer collaborative; explainable artificial intelligence. DOI: 10.1504/IJICT.2026.10080361
Abstract: Effective carbon management requires accurate prediction of energy consumption patterns, yet conventional models struggle with complex temporal dependencies. This study proposes a novel integrated framework combining transformer-based generative adversarial networks with Bayesian optimisation (BO-TransGAN) for energy forecasting and carbon optimisation. Using 2,000 hourly records of energy use and emissions, data were pre-processed via imputation, outlier removal, and normalisation. The TransGAN captures nonlinear temporal dependencies through adversarial learning, while Bobcat optimisation tunes hyperparameters for enhanced convergence and stability. BO-TransGAN achieves 0.987 accuracy, 0.995 R2, minimal losses, and low training (2.12s) and run times (7.35s). It generates realistic synthetic sequences and provides actionable insights for reducing carbon emissions, offering a scalable tool for sustainable energy planning and real-time carbon management. Keywords: carbon management; generative adversarial networks; GANs; energy consumption prediction; sustainable energy planning; energy-carbon pattern modelling. DOI: 10.1504/IJICT.2026.10080316
Abstract: This paper presents an information and communication technology (ICT) enabled adaptive PID power regulation framework for cooling tower pumps. A cyber-physical system (CPS) architecture integrates real-time sensor data through industrial IoT communication links, enabling bidirectional data exchange between field devices and a remote adaptive tuner. The tuner dynamically updates PID gains using streaming operational data under practical network constraints, including time-varying delay and packet loss. Unlike conventional methods, the proposed framework embeds physical knowledge into networked control loops, ensuring reliable performance in non-ideal communication environments. Experimental results demonstrate that compared to state-of-the-art benchmarks, the ICT-based control reduces system response time by 18.5% under variable cooling loads and lowers total pump power consumption by 12.3%. Furthermore, statistical significance analysis confirms the robustness of the proposed approach against both extreme environmental disturbances and adverse network conditions. Keywords: cyber-physical system; CPS; adaptive PID; cooling tower pump; industrial IoT. DOI: 10.1504/IJICT.2026.10080440
Abstract: This paper proposes a student information management platform that integrates support vector machine (SVM) modelling with data mining to improve academic early warning and personalised guidance. The platform adopts a four-layer architecture covering data, processing, application, and presentation functions. Multi-source student data, including academic records, behavioural information, and consumption data, are cleaned through missing value imputation, outlier correction, and feature selection to form model-ready datasets. An SVM model with an RBF kernel is developed for predicting course-failure risk and performance fluctuations, with parameters optimised by grid search (C = 10, γ = 0.1). Using 5,000 undergraduate records, the model is compared with decision tree, KNN, and logistic regression methods. Results show that the SVM model achieves 89.6% accuracy, outperforming decision tree and KNN models. The platform also maintains acceptable response and prediction times under 500 concurrent users. Keywords: intelligent management of student information; support vector machine; SVM; educational data mining; EDM; academic risk early warning; platform design. DOI: 10.1504/IJICT.2026.10080360 |
Open Access
Keep up-to-date
-
Our Blog -
Join us on Bluesky -
Follow us on X -
Visit us on Facebook - Our Newsletter (subscribe for free)
- RSS Feeds
- New issue alerts
- Inderscience is a member of publishing organisations including:
-
-
