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 Articles will not be published until the final proofs are validated by their authors.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 (26 papers in press) Regular Issues
Abstract: Existing models for news event evolution analysis and situation prediction struggle to balance event semantic dynamics and spatio-temporal feature complexity. Knowledge graphs static properties cannot capture spatio-temporal evolution patterns, and conventional ST-GCNs insufficient semantic fusion limits prediction accuracy. This study proposes a model integrating DE-KG and semantic-aware ST-GCN; its dual-modal fusion module achieves deep semantic-spatio-temporal feature coupling, improving event evolution analysis and situation prediction. Experiments show the models evolution segmentation F1-score of 0.850 (+20.6% vs. ST-GCN) and situation prediction RMSE of 0.089 (+63.7% vs. ARIMA, lower decay rate), with an optimal RMSE of 0.083 for public health events. Results verify that DE-KGs semantic dynamics fix static graphs spatio-temporal gaps, the semantic-aware matrix adapts ST-GCN topology to event semantics, and dual-modal fusion strengthens feature complementarity greatly improving complex event analysis and prediction performance. Keywords: news events; situation prediction; spatio-temporal evolution; spatio-temporal graph convolutional network; ST-GCN; semantic fusion. DOI: 10.1504/IJICT.2026.10077255
Abstract: The rapid advancement of artificial intelligence faces challenges in training data copyright and privacy compliance. Existing techniques often struggle to balance the quality and security of synthetic data, leading to a dilemma where solutions are either low in utility or high in risk. To address this, this article proposes a novel generative adversarial network incorporating a compliance-aware mechanism. This framework introduces a specialised compliance discriminator to guide the model in generating synthetic data that is both highly realistic and strictly compliant. Experiments on public datasets demonstrate that our method maintains classification accuracy within 0.8% of the original data while significantly reducing sensitive information leakage risk by 42%. Statistical validation confirms that all key metric improvements are statistically significant. This work provides an effective approach to resolving the trade-off between data compliance and utility. Keywords: data compliance; generative adversarial networks; synthetic data; privacy protection; adversarial training. DOI: 10.1504/IJICT.2026.10077256
Abstract: Assessing logical consistency in academic writing is challenging. Traditional methods relying on shallow features struggle to capture deep semantic logic and textual structure. This study proposes using spatiotemporal graph convolutional networks (ST-GCN) for this task. Texts are modelled as graphs where nodes are sentences; spatial edges represent static logical links like semantic similarity, and temporal edges model dynamic sequential dependencies. Trained on 12,000 academic texts from journals and student papers across disciplines, the model captures logical consistency from micro to macro levels. In evaluation, it achieved accuracy rates of 92.5% for global consistency, 89.3% for local consistency, 91.7% for lexical cohesion, and 93.2% for argumentative consistency. It achieved a Pearson correlation of 0.89 with human evaluation and a mean absolute error below 0.15, significantly outperforming baselines and offering an effective path for automated assessment. Keywords: logical consistency assessment; academic English writing; spatiotemporal graph convolutional networks; ST-GCN; automated writing assessment; discourse structure analysis. DOI: 10.1504/IJICT.2026.10077257
Abstract: To enhance the intelligence of pronunciation error detection and feedback precision in online oral English teaching, this study designs a system combining speech recognition and machine learning. Its core detection module uses the MFCCDBN model with feature fusion, and builds an SVMbased multiclassifier. Experimental data comes from the CSTR VCTK Corpus and the speech accent archive, containing 1,610 expertannotated phoneme samples. The model yields high accuracy for both samplesufficient and smallsample error types. Compared with LDASVM and Wav2Vec2.0SVM, it outperforms them in accuracy and standard error. Results prove the fusion models stronger robustness and efficiency with limited data, offering a practical technical approach to boost learners crosscultural communication competence. Keywords: oral English teaching; pronunciation detection; Mel-frequency cepstral coefficient; MFCC; deep belief network; DBN; speech recognition; machine learning. DOI: 10.1504/IJICT.2026.10077258 Abstract: This paper addresses the issues of knowledge tracing and path optimisation in personalised learning path recommendation and proposes an intelligent educational model based on reinforcement learning and self-attention knowledge tracing (RL-SAKT). The model is built upon a self-attention knowledge tracing framework, incorporating reinforcement learning strategy optimisation, a multi-task learning mechanism, and a joint loss function: first, it models the students long-term knowledge mastery through the self-attention mechanism, dynamically adjusting the learning path; second, it introduces a personalised path recommendation strategy driven by reinforcement learning, optimising teaching interventions based on students real-time performance and learning needs; finally, it jointly optimises the knowledge tracing loss and reinforcement learning loss to improve prediction accuracy and learning efficiency. Experimental results show that the proposed RL-SAKT model outperforms traditional methods. Subjective evaluation results also show that this method has significant advantages in predicting students learning progress, knowledge mastery, and optimising learning effectiveness. Keywords: reinforcement learning; self-attention; knowledge tracing; personalised learning path; talent cultivation. DOI: 10.1504/IJICT.2026.10077259
Abstract: This study explores the impact mechanism of AI application on employees emotional labour, aiming to balance AI empowerment and employees mental health. Based on conservation of resources theory and emotional regulation theory, 181 employees from multiple industries were sampled. Data were collected via standardised scales, regression analysis was employed to analyse the relationships between AI automation level, intelligent interaction, organisational support and employees emotional expression intensity, emotional regulation difficulty, and emotional exhaustion perception. The study found that automation level was significantly positively correlated with emotional expression intensity (= 0.220, p < 0.05), but excessively high automation level tended to cause employees emotional exhaustion. Intelligent interaction level was significantly negatively correlated with emotional regulation difficulty (= 0.261, p < 0.05). Organisational support shows duality, positively correlating with exhaustion (= 0.816, p < 0.001) but buffering pressure if humanised. The study reveals AI exerts a double-edged sword effect on emotional labour. Enterprises should optimise human-computer collaboration and provide humanised organisational support. Keywords: artificial intelligence; emotional labour; emotional regulation; workplace psychology. DOI: 10.1504/IJICT.2026.10077260 Abstract: In the context of media convergence and the profound penetration of AI technology, traditional broadcasting and hosting education faces three critical challenges: a shallow professional foundation, disjointed practical pathways, and superficial value recognition. Drawing upon the innovative practice experiences of the Television Program Broadcasting and Hosting course at Xinyu University, this paper proposes a talent cultivation model centred around the core philosophy of self-pleasure and fulfilling others. This approach leverages AI programs to establish a three-dimensional linkage training model. This model integrates the PIRIR five-step cyclical training method, the Joyful Audio intelligent platform, and a four-layer, five-dimensional evaluation system, thereby creating an integrated loop of learning, practicing, creating, and broadcasting. Empirical data reveal that students have garnered 18 provincial-level and above competition awards in broadcasting and hosting, and the total viewership for original video programs has exceeded 300 million. Furthermore, volunteer service has accumulated a total of 12,000 hours, while single-event sales during intangible cultural heritage live broadcasts have surpassed ten million yuan, thereby substantiating the effectiveness of this model. This study offers a practical paradigm for the cultivation of audiovisual communication talents in the new era, deeply intertwining technology and education. Keywords: AI-assisted teaching; broadcasting and hosting arts; audiovisual communication; talent cultivation model. DOI: 10.1504/IJICT.2026.10077261
Abstract: As global energy markets grow more complex and volatile, existing forecasting accuracy and policy effect evaluations remain insufficient. An urgent need exists for a precise prediction model integrating legal policies and multi-dimensional energy market data. This paper proposes an LSTM-based model to predict legal regulation effects on energy prices: it uses semantic feature extraction for policy texts, builds a spatiotemporal fusion framework for multi-source heterogeneous data, and designs a hierarchical memory units dynamic regulation module to adapt to stage-specific policy adjustments. In historical data, the models training-set MAE declined fluctuatingly: 56.72 CNY/ton standard coal (vague initial policies) dropped to 28.9 CNY (refined policies). After tiered pricing, policy-related price volatility fell to 45.6%. Policy clarity correlates positively with model accuracy; refinement reduced LSTMs price trend capture error by 66.6%. Keywords: LSTM neural network; energy price forecasts; legal regulation; multi-source heterogeneous data; spatio-temporal fusion. DOI: 10.1504/IJICT.2026.10077262
Abstract: This study extracts target singers voices from mixed audio with background music and noise, addressing the subjectivity, instability and lack of objective standards in traditional evaluation. An innovative SNN-SpEx+ method is proposed, combining Siamese neural network (SNN) and contrastive learning-based SpEx+. Its parameter-sharing twin architecture unifies the feature space for reference and mixed speech, breaking the bottleneck of feature space dislocation in traditional dual-network structures. Contrastive learning is integrated into vocal extraction to build a separation is learning joint optimisation framework, enhancing adaptability to unknown singers and short reference voices. Experiments on MUSDB18-HQ and NSynth-Singer show SNN-SpEx+ outperforms SpEx++ by 0.85 dB in SI-SDRi and 0.17 in PESQ. For short references (<2 s), its SI-SDRi drops only 3.16 dB (3.4 dB lower than SpEx++), providing an automatic standardised evaluation tool for music education and singer selection with broad prospects. Keywords: singing; sound quality; vocal extraction; Siamese neural network; SNN; contrastive learning. DOI: 10.1504/IJICT.2026.10077263
Abstract: Aiming at the voltage limit violations and line overloads caused by high-penetration renewable energy integration, this paper proposes a comprehensive identification method for weak links in distribution networks that integrates dual features of operational status and topological structure. A hybrid Copula function characterises the spatiotemporal correlation of photovoltaic output, and adaptive importance sampling generates typical scenarios. Probabilistic power flow calculates the violation probabilities of node voltage and line current. State and supply vulnerability indices based on utility functions are constructed to quantify voltage instability risk and nodal topological criticality, respectively. A graph neural network is then employed to integrate topological information for comprehensive weak-link identification. Simulation results based on the IEEE 33-node system demonstrate that the proposed method effectively combines operational and topological information, providing a reference for identifying weak links and optimising energy storage configuration in new-type distribution networks. Keywords: hybrid Copula; probabilistic power flow; graph neural network; GNN; vulnerability identification; state vulnerability; structural vulnerability. DOI: 10.1504/IJICT.2026.10077274
Abstract: Predicting the remaining useful life of automotive components is vital for safety and reliability in dynamic operating environments. Existing data-driven methods often miss the temporal dynamics and evolving patterns in sensor data, limiting prediction accuracy. Propose a novel simulation modelling framework that merges a physics-informed degradation simulator with a deep learning network augmented by multi-head temporal attention. This fusion generates realistic degradation trajectories while the attention mechanism dynamically prioritises key time-based features for precise life estimation. Testing on a public turbofan engine dataset shows model achieves a mean absolute error of 12.8 cycles and a root mean square error of 16.3 cycles, surpassing conventional long short-term memory and convolutional neural network models by 18.7% and 23.4%, respectively. The attention outputs provide interpretable views into critical degradation phases, offering a robust and insightful tool for prognostics and health management in automotive systems. Keywords: remaining useful life; RUL; temporal attention; deep learning; degradation simulation; automotive components. DOI: 10.1504/IJICT.2026.10077316
Abstract: Aiming at semantic disconnection and visual distortion between generated results and actual scenes in environmental art design, this paper proposes a consistent generation and simulation model based on multimodal transformer. Traditional methods have limitations in coordinating complex elements and ensuring spatial logic, hindering design implementation. By integrating multi-source information including text, sketches, and scene images, an end-to-end generation-simulation framework achieves consistent mapping from concept to high-fidelity visual output. Using the public dataset mit ade20k, results show the model achieves significant improvements in visual fidelity (area under the curve 0.92, an increase of 8.2%) and user preference (normalised discounted cumulative gain @10 an increase of 15.7%), with all key indicators being statistically significant (p < 0.01). This confirms the models effectiveness in enhancing automation and usability of environmental art design. Keywords: multimodal transformer; multimodal transformer; environmental art; generative model; consistency simulation. DOI: 10.1504/IJICT.2026.10077317
Abstract: Massive open online courses have difficulty dynamically allocating computational resources and maintaining a good learner experience. Existing methods, using system metrics, do not consider the impact of collective psychological states on demand. This paper puts forward a group psychological state-oriented elastic resource allocation framework. We analyse multimodal behavioural data to build a predictive model of learner states such as engagement or confusion. The prediction dynamically guides cloud resource scaling through a state-aware algorithm. Extensive experiments on the public massive open online courses dataset prove the effectiveness of our approach. The accuracy of recognising the psychological state reaches 89.3%, and it can save about 30% of resources compared with traditional methods. This study shows how mixing psychological knowledge enables better resource management, making them more useful and quicker to react for big online learning places on internet. Keywords: massive open online courses; MOOCs; collective psychological state; elastic resource allocation; machine learning; resource optimisation. DOI: 10.1504/IJICT.2026.10077318
Abstract: Hospitals worldwide face critical challenges in optimising workforce allocation while maintaining high standards of patient care. This study presents the development and real-world evaluation of an AI and ICT-enabled hospital human resource management decision support system (HRM-DSS) designed to enhance staffing efficiency, improve staff satisfaction, and support patient care outcomes. Utilising a hybrid deep learning and predictive analytics framework, the system was piloted in a major hospital and demonstrated a substantial reduction in scheduling conflicts, overtime hours, and staffing costs. Significantly, staff satisfaction and perceived fairness of scheduling increased, while patient wait times modestly improved without compromising care quality. A mixed-methods analysis revealed that the systems effectiveness was strongly linked to user trust and the transparency of its implementation. These findings provide robust evidence that integrating AI into HRM can sustainably transform hospital operations, supporting both workforce well-being and patient-centred care. The study offers a scalable blueprint for healthcare leaders seeking data and ICT-enabled solutions to complex HRM challenges. The integration of AI with ICT infrastructures demonstrates how advanced information and communication technologies can transform hospital HRM. Keywords: artificial intelligence; information and communication technology; decision support system; DSS; patient care; human resource management; HRM. DOI: 10.1504/IJICT.2026.10077319
Abstract: In cloud computing environments, accounting data faces severe challenges such as tampering, loss, and verification difficulties. Traditional centralised auditing solutions carry single-point-of-failure risks and suffer from inefficient verification processes. To address this, this paper proposes a smart accounting data integrity assurance system that synergises blockchain and cloud computing. The system leverages blockchains immutability to construct a distributed auditing framework and employs optimised cryptographic algorithms to achieve efficient data verification in the cloud. Experiments on the public university of California Irvine dataset demonstrate that compared to traditional approaches, this system accelerates data verification by approximately 30% while keeping storage overhead below 15%. It effectively resolves the core conflict between trustworthy storage and efficient verification for cloud accounting data, providing modern enterprises with reliable data security assurance. Keywords: blockchain; cloud computing; data integrity; smart accounting. DOI: 10.1504/IJICT.2026.10077360
Abstract: Financial anomaly detection constitutes a critical pillar of corporate risk management. However, stringent data privacy regulations increasingly restrict centralised multimodal learning paradigms. To address this challenge, this paper proposed a novel multimodal federated learning framework deployed in a cloud environment. This system enables enterprises to process proprietary numerical and textual financial data locally. An attention-based cross-modal fusion module is introduced to effectively integrate heterogeneous features. Only model updates are transmitted to cloud servers for secure aggregation, thereby preserving data confidentiality. Furthermore, the framework incorporates adaptive client selection and differential privacy mechanisms to address non-IID data distributions and enhance security guarantees. Extensive evaluations on a real-world financial dataset demonstrate that our approach achieves 94.76% accuracy, 87.45% precision and 86.53% F1-score, substantially surpassing conventional unimodal and federated baselines while providing verifiable privacy assurances. This work presents a scalable and regulatory-compliant solution for collaborative financial risk intelligence. Keywords: multimodal federated learning; financial anomaly detection; privacy protection; cloud-edge collaboration; attention mechanism. DOI: 10.1504/IJICT.2026.10077361
Abstract: To solve the problems of data sparsity, unstable demand and multi-stage coordination in the cross-border e-commerce supply chain optimisation bottleneck, this paper puts forward an end-to-end decision-making framework based on multi-source hierarchical transfer learning and deep reinforcement learning. By performing multi-source adversarial adaptation in feature space to learn domain-invariant representations and introducing a meta-learning weighting mechanism at the sample level to refine the selection of beneficial knowledge, this approach systematically mitigates model cold-start and negative transfer issues. Experiment shows that the framework greatly improves the forecast accuracy of demand from 4.83 to 5.76 compared to the basic model. In simulated cross-border supply chain networks, it cuts down the overall operation expenses by 7.2%, and shortens the time taken to fulfil orders by 13.5%. Keywords: transfer learning; cross border e-commerce; supply chain; deep reinforcement learning; DRL. DOI: 10.1504/IJICT.2026.10077362
Abstract: Traditional approaches to assessing corporate internal control effectiveness often depend on isolated data sources or simplistic fusion methods, resulting in limited accuracy and generalisability. To address this, this paper introduces a multi-view contrastive learning network framework that integrates financial statements, managerial narrative disclosures, and market sentiment data as complementary views of a firms underlying risk profile. The model first learns aligned, robust representations across modalities through self-supervised intra-view and cross-view contrastive pre-training. It then fine-tunes on labelled data with an attention-based fusion mechanism for internal control weakness classification. Experiments show that the proposed framework achieves an area under the curve of 0.842 and a precision of 0.502, exceeding the best baseline by 2.9 percentage points in area under the curve and 10.3% in precision. These results demonstrate that the multi-view contrastive learning network framework significantly enhances the performance, robustness, and interpretability of automated internal control evaluation. Keywords: internal control evaluation; contrastive learning; multi-source data fusion; neural networks; financial risk prediction. DOI: 10.1504/IJICT.2026.10077363
Abstract: With the rapid growth of e-commerce, accurately predicting consumers purchase intentions is crucial for effective marketing. Existing approaches primarily rely on static or coarse-grained features, which fail to capture the dynamic and complex nature of user decision-making. This paper proposes a dynamic modelling framework based on fine-grained user behaviour sequences. It employs time-aware sequence encoding and a dynamic interest state network to learn and update the users purchase intention probability in real-time. This integrated strategy captures both short-term local patterns and long-term global dependencies within behavioural sequences. Experiments on public datasets demonstrate that our model achieves a 1.52% improvement in the area under the receiver operating characteristic curve and a 2.18% gain in F1-score over strong baselines, with statistically verified improvements. This study provides novel theoretical and methodological insights for understanding and predicting dynamic customer behaviour. Keywords: fine-grained behavioural sequences; dynamic modelling; purchase intent prediction; deep learning; sequence analysis. DOI: 10.1504/IJICT.2026.10077364
Abstract: To address the challenges of strong nonlinearity and uncertain disturbances in electromagnetic suspension systems, this paper proposes an RL-Fuzzy adaptive control architecture that integrates reinforcement learning with fuzzy logic. The core innovation involves utilising fuzzy rules to dynamically adjust the exploration rate of the deep deterministic policy gradient algorithm and incorporating Lyapunov stability constraints to suppress current overshoots. Validated using the Springer Nature real vehicle bench dataset, the proposed method reduces the root mean square of body acceleration to 1.15 m/s2 (surpassing the ISO 2631 high-comfort threshold) and full optimisation of key indicators: safety (tyre displacement variance 1.89 mm2), energy consumption (current rms 1.28 A2), training efficiency (42.3% reduction in steps). This approach provides a computationally efficient robust control framework for intelligent suspension systems. Keywords: electromagnetic suspension system; RL-Fuzzy adaptive control; deep deterministic policy gradient; Lyapunov stability constraint. DOI: 10.1504/IJICT.2026.10077365
Abstract: As the user base expands, recommendations under high concurrency conditions cannot meet the low latency requirements. In order to improve the accuracy of recommendations and the real-time performance of the system, a big data-driven interactive English learning system is designed. The hardware part deployed a distributed server cluster (16 core CPU/128 GB memory nodes x 8), equipped with NVIDIA Tesla T4 GPU accelerated deep learning inference, and implemented elastic resource scheduling through Kubernetes containerisation platform; at the software level, a microservice architecture is adopted, with an adaptive interactive interface built on React in the front-end and Spring Cloud framework in the back-end. Constructing dynamic learning profiles through multi-source data fusion, combined with an improved K-means clustering algorithm to achieve precise learner clustering. At the recommendation algorithm level, collaborative filtering is integrated with cognitive ability resource difficulty matching (DSM) and preference resource type correlation (RSM) to generate TOP-N recommendation sets through weighted standardisation, achieving teaching resource recommendation. The experimental results show that this scheme effectively solves the problem of recommendation quality degradation in large-scale user scenarios, while ensuring real-time interactive performance, providing practical reference for the technical optimisation of online education systems. Keywords: big data-driven; interactive; English learning; resource recommendation; user portrait. DOI: 10.1504/IJICT.2026.10077366
Abstract: Aiming at the problem that the precise identification and traceability methods for false e-commerce reviews are difficult to comprehensively capture the complex semantics and potential features in the reviews, this paper first uses a text convolutional neural network and a pre-trained model to extract features from multi-dimensional text semantics respectively. By further integrating the features of the reviewers, the models understanding of the overall semantics is further enhanced. The images posted by users in the comments are subjected to feature extraction using the residual network to obtain the corresponding visual semantic features. Subsequently, multimodal semantic feature fusion will be carried out to identify false comments. The experimental results on large-scale datasets show that the recognition and traceability accuracy rates of the proposed method have increased by at least 0.7% and 1.1% respectively, significantly outperforming the existing methods in the recognition and traceability of false comments. Keywords: e-commerce fake reviews; identification and traceability; multimodal semantics; feature fusion; transformer model. DOI: 10.1504/IJICT.2026.10077367
Abstract: Automated polyphonic music generation remains a challenging task due to the difficulty in designing reward functions and modelling long-range dependencies. To address this, this paper proposes a polyphonic music generation model based on generative adversarial imitation learning. Our model employs a Gated Transformer-XL as its core to effectively capture intricate contrapuntal relationships. Experimental results demonstrate that the model achieves superior performance across multiple metrics: it reduces the Earth movers distance to 0.23; increases voice separation mutual information to 0.45, and achieves a 91.2% harmonic rule compliance rate. In subjective evaluations, the model attained average opinion scores of 8.7 for melodic fluency and 8.9 for harmonic richness, significantly outperforming all baseline models. These results validate the effectiveness of our approach in generating high-quality polyphonic music with both technical proficiency and artistic merit. Keywords: generative adversarial imitation learning; polyphonic music generation; music AI; inverse reinforcement learning. DOI: 10.1504/IJICT.2026.10077368
Abstract: This paper examines how US tariff changes have affected consumer inflation over the past four decades using linear and quadratic regression models. Combining headline CPI with sub-indices for automobiles and apparel, and applying a 10% trimmed-sample procedure, the analysis tests for both average and nonlinear effects. Once GDP growth, interest rates and unemployment are controlled for, tariffs have little explanatory power for aggregate CPI. Sector-level regressions tell a different story: in autos and apparel small tariff changes barely move prices, but beyond a threshold further increases are associated with sharply higher inflation, consistent with convex, state-dependent pass-through. The findings suggest that tariffs do not operate as a broad engine of US inflation; they mainly reallocate price pressure toward import-intensive consumer goods, with the strength of the response shaped by supply-chain flexibility and market structure. Keywords: tariffs; inflation; consumer prices index; trade policy.
Abstract: Universities increasingly log dense streams of student interactions, yet most dashboards act only after learning issues have become visible. To enable earlier and cleaner detection, this study introduces a Transformer-GNN hybrid model for data-driven monitoring of teaching quality. A temporal encoder first learns long-range dependencies from sequential event data, while a graph encoder integrates cohort and course context. A reliability-aware gate and per-family calibration then refine alert stability and accountability. Across three academic terms covering six course families, the model improved the area under ROC by 2.4 points and the area under precision-recall by 5.7 points over a strong fusion baseline, reduced calibration error by 45%, and extended mean warning lead time by 1.3 days. The framework remains robust under pacing shifts and provides interpretable, actionable explanations for instructors. Keywords: transformer; graph neural network; teaching quality monitoring; early warning; educational analytics; calibration; interpretability. DOI: 10.1504/IJICT.2026.10077369
Abstract: Museum cultural and creative products often fail to meet audience expectations, leading to low appeal and homogeneity. This study employs a data-driven approach, using NLP and machine learning to analyse 680,000 visitor comments from a provincial museum. Findings show 42% of visitors prefer novel, culturally rich products, yet satisfaction with existing items is low (3.7/5). Data analysis identified the bronze ware gluttonous pattern and ancient book calligraphy as the most valued cultural symbols. Using K-means clustering and A/B testing, new product prototypes were developed. Market tests with 1,000 participants showed a 32% rise in preference, a 26% increase in payment willingness, and a cultural perception score of 4.3. This method effectively enhances product cultural depth and market appeal, offering a viable path for revitalising traditional culture. Keywords: cultural big data analysis; museum cultural and creative products; tourist behaviour data; semantic mining; design model construction. DOI: 10.1504/IJICT.2026.10077370 |
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:
-
-
