This paper introduces a novel mixed-frequency intertemporal network asset pricing model (MFIN-APM) that captures the intertemporal propagation of mixed-frequency information in firms’ dynamic supply chain networks and outperforms the state-of-the-art deep learning models for asset pricing. To achieve this, we integrate a mixed data sampling (MIDAS) module and a long short-term memory (LSTM) module into the graph convolutional network (GCN). Applied to the Chinese A-share and U.S. NYSE stock markets, we demonstrate how the MFIN-APM model exploits mixed-frequency data, extracts intertemporal and high-order topological features through dynamic supply chain structures, and improves the performance of both the aforementioned models and traditional asset pricing models. Monte Carlo simulations further confirm the generalizability of our empirical results.

The National Equities Exchange and Quotations (NEEQ) in China is a key platform for small and medium-sized enterprises (SMEs) to access public capital markets. However, their credit evaluation is challenging due to financial opacity and information asymmetry. Given that conventional credit evaluation methods mainly rely on financial statements, news reports, and transaction history, often overlooking the complex relationships that affect SMEs’ performance, we propose a KG-AttRGCN-XGBoost model to evaluate enterprise credit effectively. This model uses a knowledge graph (KG) to construct enterprise relationship networks, utilizes contextual embeddings of a relational graph convolutional network (RGCN) with a constrained attention mechanism to model complex inter-enterprise connections, and transmits the extracted features to XGBoost for credit evaluation. Experimental results show that our model significantly outperforms several popular graph neural network-based credit evaluation methods, better handling multi-relational data and achieving higher precision.

Despite the growing attention given to the impact of ESG performance on corporate finance, such as debt financing costs, little is known about the underlying mechanism. Hence, we aim to answer this question from the supply chain perspective. We use 19,121 enterprise-year observations of 3585 Chinese A-share listed enterprises over the 2013-2022 period and employ panel data regression models with fixed effects estimation. The empirical results confirm a negative relationship between ESG performance and corporate debt financing costs even after a series of robustness tests. We conclude that good ESG performance effectively reduces debt financing costs with a marginal effect of 0.002. This reduction effect can be achieved through two channels: supplier stability and customer concentration. Heterogeneity analyses further demonstrate that the reduction effect is more pronounced in non-state-owned enterprises and enterprises with low industrial competition. Overall, our findings enrich the understanding of how ESG performance reduces debt financing costs and highlight the importance of supply chain management for enterprises.

In intelligent fault diagnosis, fault severity classification with class imbalance remains a tremendous challenge. To simultaneously consider relative natural order in fault severity and class imbalance, we propose the adaptive ordinal sample-weighted meta residual network (AOSW-MRN). The AOSW-MRN model uses a weighting network and a meta-model cloned from the residual network to create a nonlinear weighted mapping. It adaptively learns sample weights from a balanced and clean-label meta-dataset, training a model robust to imbalance and ordinal relationships. We validate its effectiveness in two real-world case studies with different imbalance rates. Experimental results demonstrate that our model outperforms several existing Start-of-the-Art models regardless of classification and regression performance since it considers the ordinality of samples in the feature space.

Despite growing attention to the impact of ESG performance on corporate finance, particularly debt financing costs at the firm level, spillover effects at the supply chain level remain underexplored. Using data from Chinese A-share listed companies over the 2012-2023 period, we build supply chain networks to examine the spillover effects of core enterprises' ESG performance to node enterprises. The empirical results confirm that core enterprises' good ESG performance can significantly reduce node enterprises' debt financing costs even after a series of robustness tests. This reduction effect is achieved through increasing the supply chain liquidity or reducing supply chain disruption risk. Heterogeneity analyses demonstrate that the reduction effect is more prominent when core and node enterprises share consistent characteristics and are located in different provinces. Overall, our findings advance the understanding of ESG spillover effects within supply chain networks and highlight supply chain ESG management's important role in corporate finance.

We develop a penalized U-MIDAS-Mlogit model by introducing the group LASSO penalty into the unrestricted MIDAS multinomial logit model. This penalized U-MIDAS-Mlogit model can implement multinomial classification in a high-dimensional mixed-frequency data environment. We apply it to credit ratings for listed companies in China over the period 2008-2023. The penalized U-MIDAS-Mlogit model can extract pivotal information from high-frequency financial variables and low-frequency internal and external governance indicators. It outperforms several competing models in predicting credit ratings.

We, the Editor-in-Chief and Publisher of Disability and Rehabilitation: Assistive Technology, have retracted the following articles, which were published as part of a planned Special Issue titled "AI empowered Assistive Technology" and edited by Uzair Aslam Bhatti: Liu, Y., & Long, T. (2025). How does social media shape the duration of formal care for disabled elderly? Moderating effects based on digital competence and psychological expectancy. Disability and Rehabilitation: Assistive Technology, 20(7), 2146-2159. https://doi.org/10.1080/17483107.2025.2525539Lu, X., & Guo, W. (2025). Utilizing artificial intelligence to enhance social connections - the alleviating effect of emotionally intelligent chatbots on loneliness. Disability and Rehabilitation: Assistive Technology, 1-11. https://doi.org/10.1080/17483107.2025.2540494Xuanfeng, D. (2025). Gesture recognition and response system for special education using computer vision and human-computer interaction technology. Disability and Rehabilitation: Assistive Technology, 1-18. https://doi.org/10.1080/17483107.2025.2527226Ren, H., Lv, J., Ren, L., & Guo, R. (2025). Optimizing the spatial allocation of pension resources in downtown Shanghai: a dynamically updated genetic algorithm approach. Disability and Rehabilitation: Assistive Technology, 1-14. https://doi.org/10.1080/17483107.2025.2536173Kadiyala, B., Nippatla, R. P., Boyapati, S., Vasamsetty, C., Alavilli, S. K., & M, T. (2025). Prevention and health care intervention of common injuries in long-distance running for college teachers. Disability and Rehabilitation: Assistive Technology, 1-20. https://doi.org/10.1080/17483107.2025.2564371Zhang, L., Wang, G., & Wang, M. (2025). Impacts of virtual reality-aided physical training on the balance ability and energy expenditure of children with moderate intellectual disability. Disability and Rehabilitation: Assistive Technology, 1-16. https://doi.org/10.1080/17483107.2025.2546026Zhang, S., & Meng, Q. (2025). Intelligent sports rehabilitation: integrating deep learning and real-time monitoring to achieve personalized rehabilitation. Disability and Rehabilitation: Assistive Technology, 1-15. https://doi.org/10.1080/17483107.2025.2559187Gao, P., Wei, D., Lu, W., & Zhai, Y. (2025). Unveiling the mechanisms influencing service efficiency in Chinese hospital portal sites: evidence from DEA and fsQCA analysis. Disability and Rehabilitation: Assistive Technology, 1-16. https://doi.org/10.1080/17483107.2025.2554956Chen, J., Chen, G., & Wang, X. (2025). Configuration analysis of mobile phone addiction among college students: findings from fsQCA method. Disability and Rehabilitation: Assistive Technology, 1-14. https://doi.org/10.1080/17483107.2025.2548858Ye, L., Du, J., & He, Z. (2025). Willingness to communicate with AI chatbots in English: the role of personality and trust among Chinese college students. Disability and Rehabilitation: Assistive Technology, 1-16. https://doi.org/10.1080/17483107.2025.2572532Wan, L., Gao, W., Xi, P., Xu, K., Li, T., Wu, J., & Wu, D. (2025). AI-Empowered assistive technology for optimizing specimen submission in obstetrics and gynecology: integrating DeepSeek with the ADDIE model. Disability and Rehabilitation: Assistive Technology, 1-11. https://doi.org/10.1080/17483107.2025.2561248Liao, Z. (2025). Artificial intelligence-enabled integration of "Fu" culture into the moral education system for special education across K-12 and higher education: a soft computing approach. Disability and Rehabilitation: Assistive Technology, 1-13. https://doi.org/10.1080/17483107.2025.2555541Cen, X. (2025). Multimodal deep learning methods for speech and language rehabilitation: a cross-sectional observational study. Disability and Rehabilitation: Assistive Technology, 1-13. https://doi.org/10.1080/17483107.2025.2551708Wang, Y., Li, H., Du, Y., Zhang, P., Shi, S., Ma, Y., & Wang, S. (2025). The relationship between job competence, demographic characteristics and professional misconduct among medical staff in the AI era. Disability and Rehabilitation: Assistive Technology, 1-12. https://doi.org/10.1080/17483107.2025.2561247Xiao, B., Sun, H., Lin, S., Li, J., & Yu, Y. (2025). Virtual classrooms, real resilience: how AI tutors enhance cognitive rehabilitation in pediatric cancer survivors. Disability and Rehabilitation: Assistive Technology, 1-14. https://doi.org/10.1080/17483107.2025.2549898Li, M. (2025). Evaluating the role of social support systems in enhancing the well-being of elderly individuals with disabilities. Disability and Rehabilitation: Assistive Technology, 1-21. https://doi.org/10.1080/17483107.2025.2568945Xiao, X. (2025). Choral harmony: the role of collective singing in ritual, cultural identity and cognitive-affective synchronisation in the age of AI. Disability and Rehabilitation: Assistive Technology, 1-17. https://doi.org/10.1080/17483107.2025.2556025Maashi, M., Alhefdhi, A., Alanazi, F., & Rizwanullah, M. (2025). Enhancing communication for people with hearing disabilities through robust sign language recognition using deep learning and the internet of things. Disability and Rehabilitation: Assistive Technology, 1-19. https://doi.org/10.1080/17483107.2025.2562454Cheng, W., Guo, J., & Lu, D. (2025). CHATWELL: an AI-enabled adaptive tutoring system for improving mandarin composition skills in L2 students with learning difficulties. Disability and Rehabilitation: Assistive Technology, 20(7), 2358-2374. https://doi.org/10.1080/17483107.2025.2554955Zhang, R. (2025). Personalised sports rehabilitation analysis using a fitness enhanced model based on big data and deep learning. Disability and Rehabilitation: Assistive Technology, 1-14. https://doi.org/10.1080/17483107.2025.2561926Xue, Q., Yang, J., & Liu, X. (2025). An AI-integrated spatial auxiliary decision-support model for historic and cultural blocks based on the spatial triad. Disability and Rehabilitation: Assistive Technology, 1-17. https://doi.org/10.1080/17483107.2025.2554331Huang, R. (2025). Between intimacy and surveillance: elderly users' adaptation and privacy negotiation in human-robot interaction. Disability and Rehabilitation: Assistive Technology, 1-20. https://doi.org/10.1080/17483107.2025.2566379Wang, P., Mi, B., Lu, J., & Zheng, F. (2025). Evaluating AI tutor feedback in medical education: a case study of computer basics and applications course for undergraduates. Disability and Rehabilitation: Assistive Technology, 1-15. https://doi.org/10.1080/17483107.2025.2560673Liu, J., Liu, L., Wu, Y., Wang, Z., & Li, X. (2025). MRI feature engineering and SVM framework for schizophrenia recognition. Disability and Rehabilitation: Assistive Technology, 1-25. https://doi.org/10.1080/17483107.2025.2569801Zhu, D., Xu, Q., & Jiang, C. (2025). Good intentions may be wrong: evidence from anonymous donation behaviour on Chinese charitable crowdfunding platform. Disability and Rehabilitation: Assistive Technology, 1-14. https://doi.org/10.1080/17483107.2025.2565405Xiao, B., Lin, S., Yu, Y., Li, J., & Li, X. (2025). AI-Enhanced assistive interventions for adolescent cyberbullying: a gender-sensitive moderated mediation approach. Disability and Rehabilitation: Assistive Technology, 1-19. https://doi.org/10.1080/17483107.2025.2570890Xiong, P., & Zhang, Y. (2025). Artificial intelligence as an assistive technology in language education: investigating TPACK, self-efficacy, and organisational support. Disability and Rehabilitation: Assistive Technology, 1-19. https://doi.org/10.1080/17483107.2025.2561927Yu, Z., Xu, Z., & Qi, J. (2025). Disability-oriented data protection in AI-enabled assistive technologies: bridging gaps in China's legal framework. Disability and Rehabilitation: Assistive Technology, 1-25. https://doi.org/10.1080/17483107.2025.2568940Xiao, B., Xing, Q., Ye, J., & Nie, Y. (2025). The impact of social media on adolescent conformity: mechanisms, individual moderators, and implications for AI-Empowered assistive technologies. Disability and Rehabilitation: Assistive Technology, 1-19. https://doi.org/10.1080/17483107.2025.2570889Yan, X., & Chen, H. (2025). From cognitive alignment to technological adaptation: understanding rural digital governance through AI-augmented human-data collaboration. Disability and Rehabilitation: Assistive Technology, 1-24. https://doi.org/10.1080/17483107.2025.2564370Sun, H., Tang, X., & Jiang, Y. (2025). Dual-channel conduction modulated by digital literacy: examining technology shock awareness and informal digital learning through an AI-enabled human-computer interaction lens. Disability and Rehabilitation: Assistive Technology, 1-17. https://doi.org/10.1080/17483107.2025.2573221Liu, Z. (2025). Hierarchical attention mechanism in deep learning improving music therapy rehabilitation through context aware emotion mapping. Disability and Rehabilitation: Assistive Technology, 1-15. https://doi.org/10.1080/17483107.2025.2568164Following publication, the Publisher identified concerns regarding the peer review process. An investigation by the Taylor & Francis Publishing Ethics & Integrity team in full cooperation with the Editor-in-Chief concluded that the articles included in this Special Issue were not peer-reviewed appropriately in line with the Journal's peer review standards and policy. As the stringency of the peer review process is core to the integrity of the publication process, the Editor-in-Chief and Publisher have decided to retract all of the articles within the above-named Special Issue. The Editor-in-Chief and the Publisher have not confirmed if the authors were aware of this compromised peer review process. The authors have been informed of this decision. We have been informed in our decision-making by our editorial policies and the COPE guidelines. The retracted articles will remain online to maintain the scholarly record, but they will be digitally watermarked on each page as 'Retracted'.

Given the volatile operational environments, rotating machinery is subject to a high rate of failures and expensive maintenance costs. This calls for the imperative undertaking of formulating innovative methodologies for diagnosing faults in rotating machinery. Industrial fault diagnosis faces two prevailing challenges in the real world: fusing one-dimensional vibration signals from two different perspectives and assessing fault severity. We employ a two-branch network for processing one-dimensional vibration signals from both time and frequency domains. Additionally, we incorporate a unimodal binomial distribution and a weighted cross-entropy loss to account for fault severity. Amalgamating ordinal regression with two-branch networks, a model termed the ordinal two-branch networks with a unimodal binomial distribution (OTBN-UBD) is developed for fault severity classification. We validate the efficacy of the OTBN-UBD model by using the failure case data collected from real industrial rotating machinery in operation. The OTBN-UBD model, incorporating ordinal information, shows superior multi-classification and ordinal regression performance, as indicated by our experimental results. In practical engineering applications, the OTBN-UBD model demonstrates considerable flexibility, with the potential to mitigate extreme misclassifications through adjusting weights within the loss.

High-frequency macro-financial environment variables provide more useful information and are efficient in predicting the low-frequency GDP growth rate. To this end, we extend the traditional Growth-at-Risk (GaR) into a high-frequency GaR (HF-GaR). In this extension, we construct three high-frequency macro-financial environment indices using a mixed frequency dynamic factor model and then use a mixed data sampling-quantile regression method to measure China’s daily GaR from Jan 1, 2000, to Sep 30, 2024. The evidence shows that our HF-GaR has favorable prediction performance, with quantile mean absolute error and quantile root square error values less than 0.1 and is significantly superior to the traditional GaR at the 1

The existing equipment maintenance methods mainly separate the two related phases of prediction and predictive maintenance (PdM) by looking at remaining useful life (RUL) prediction without considering maintenance or optimizing maintenance schedules based on the given prediction information. To address this issue, we propose a framework based on a multi-stream attention fusion network with quantile regression model and deep reinforcement learning (DRL). In this novel framework, the multi-stream attention fusion block is used to comprehensively capture the operating status of industrial equipment and eliminate the extracted duplicate information. Quantile regression is employed to obtain the probabilistic RUL prediction results with uncertainty expression. We further formulate the PdM problem for DRL, where maintenance actions are triggered based on the estimates of the RUL distribution. To illustrate the superiority of our framework, we compare it with some state-of-the-art models using a public data and our private data. The experimental results indicate that our method exhibits high accuracy and stability in bearing RUL prediction and predictive replacement.

Multivariate time-series (MTS) collected from multiple sensors on industrial pumps often exhibit concept drift and noise contamination due to variable working conditions and complex environments. To detect anomalies in such MTS, we propose a novel model called spectral residual with self-attention variational autoencoder (SR-SAVAE). Specifically, the spectral residual operation is used to mitigate concept drift, while the variational inference combined with a total variation regularization is used to address the issue of noise contamination. Experimental results on three public datasets indicate that the SR-SAVAE model achieves good anomaly detection results for general MTS. More importantly, compared to other state-of-the-art models on a private dataset about pumps, the results illustrate the superiority of the SR-SAVAE model in anomaly detection for MTS with concept drift and noise contamination. Finally, ablation studies on the SR-SAVAE model detail the efficacy of each component.

We develop a social media Q & A text semantic similarity (QATSS) measure to distinguish the quality of management responses on the Shanghai Stock Exchange E-interaction (SSEEI) platform, and examine its role in identifying corporate fraud. We find robust evidence that firms with higher QATSS are less likely to commit corporate fraud. Further analyses show that the negative relationship between QATSS and fraud is more pronounced in less visible firms, non-state-owned firms, and firms with lower audit quality. Overall, our results suggest that the semantic similarity between management responses and investors' questions on social media is a value-relevant signal for fraud detection.

In real industrial processes, machines usually run under variable working conditions, which impose challenges for anomaly detection. To complete anomaly detection for machines under variable working conditions, we develop a reconstruction-based autoencoder called clustering-based contrastive learning autoencoder (CBCL-AE). It integrates clustering-based contrastive learning (CBCL) to perform clustering in the feature space and enhance the differentiation of features from different working conditions, thereby achieving adaptive working condition recognition. Considering the crucial role of the clustering of CBCL, we theoretically and experimentally demonstrate its convergence property during the training process, which directly determines the effectiveness of CBCL-AE. CBCL-AE's superiority has been validated on three public datasets and two private datasets collected from an actual industrial process. These validations highlight its superiority over five state-of-the-art models in unsupervised anomaly detection.

We propose a novel latent factor pricing model to extract latent pricing factors and corresponding factor loadings from multi-source heterogeneous information through a deep learning architecture. Notably, we pioneer the extraction of policy pricing factors from China’s national strategies (“Five-Year Plans”, “Government Work Reports”, and “Monetary Policy Reports”) using natural language processing and a dynamic topic model. The proposed mixed-frequency deep factor asset pricing (MIDAS-DF) model learns from mixed-frequency heterogeneous data and captures nonlinear joint patterns between inputs and outputs, providing more nuanced insights into asset pricing. The empirical analyses of the Chinese A-share market from January 1, 2003 to July 31, 2022 show that the MIDAS-DF model outperforms competing models in pricing individual stocks, various test portfolios, and investment portfolios. The results also demonstrate that low-frequency policy information anchors long-term pricing trends, while high-frequency market and sentiment information refine short-term pricing accuracy. They work together to enhance the pricing performance.

In real-world scenarios, fault severity data follows an imbalance distribution, meaning that normal and low-level faults constitute a large portion of the training data. In contrast, high-level fault instances are relatively scarce. This imbalance leads to the decision hyperplane of classifiers to skew in favor of higher fault levels. Typically, reweighting methods ensure that hyperplanes fairly divide the feature space by assigning weights based on the number of instances. However, we found that the number of instances fails to represent the effective area within a fault level in the feature space. To tackle this issue, we first analyze the distribution and their correlations of different instances within each class. Second, we calculate the effective area by treating instances of the same class as identically distributed random variables. Third, we introduce a novel weighting approach, with weights inversely proportional to the effective area. This method results in more flexible weights, allowing for adaptive adjustments based on feature optimization during the training process. Experimental results demonstrate that the proposed method outperforms the comparative methods on a real-world fault severity dataset and enhances fault diagnosis performance under class imbalance.

The ordered logit (OLogit) model is a regression model for an ordinal dependent variable. For a conventional time series OLogit model, both the dependent variable and the independent variables are required to be observed at the same frequency. However, this requirement is violated under the circumstance of mixed frequency data. To this end, we introduce the unrestricted MIDAS (U-MIDAS) method into the OLogit model and develop a novel U-MIDAS-OLogit model, in which high-frequency covariates are used to predict a low-frequency outcome with ordinal categories. The U-MIDAS-OLogit model enlarges the application of OLogit and enables to produce timely forecast. To verify its effectiveness, we conduct extensive Monte Carlo simulations. The numerical results show that the U-MIDAS-OLogit model is superior to several typical OLogit models in terms of prediction performance. We then apply the U-MIDAS-OLogit model to predict credit ratings of listed companies in China and the US, respectively. The empirical results also confirm its promising in practical applications.

The recently implemented environmental protection tax (EPT) policy in China provides the opportunity to conduct a quasi-natural experiment to empirically evaluate environmental regulations’ impact on corporate practices. We adopt the difference-in-differences method to analyze the effect and mechanisms of the EPT on corporate environmental, social, and governance (ESG) greenwashing based on Chinese A-share listed companies from 2015 to 2021. The empirical results show that the EPT drives companies to engage in ESG greenwashing, significantly increasing ESG greenwashing by approximately 13.16 %. The mechanism tests demonstrate that the EPT exerts governance pressure but does not create more resources, making companies more likely to achieve compliance through greenwashing. Furthermore, the effect of the EPT on ESG greenwashing is more pronounced for large companies and those located in regions with high economic development and strong regulatory enforcement. Our study provides solid evidence with valuable implications for improving the EPT policy to achieve green and sustainable development.

Existing self-supervised multivariate time series anomaly detection methods struggle with interference among variables during reconstruction. They also tend to miss capturing critical anomaly information, resulting in unsatisfactory performance, especially in scenarios with strong mechanistic contexts. To this end, we propose a targeted anomaly detection algorithm called inference stacked recurrent autoencoder (ISRAE). Its key contribution lies in the design of a specific inference kernel, derived from specialist knowledge, which captures the strong mechanistic relationships among variables. This kernel is then fused with the multidimensional anomalies predicted by the SRAE, which mitigates interference among variables through the stacking technique. Furthermore, a novel differential constraint is introduced into the loss function, which not only highlights anomaly reconstruction errors, but also smooths the reconstructions, enhancing overall detection performance. Comprehensive comparison experiments and ablation studies show that ISRAE achieves superior anomaly detection performance under strong mechanistic contexts and highlight the importance of each key module in ISRAE.

Emerging intelligent fault diagnosis models based on domain adaptation can resolve domain shift problems produced by different working conditions. However, the prerequisite of obtaining target data in advance limits the application of these models to practical engineering scenarios. To address this challenge, a deep mixed domain generalization network (DMDGN) is proposed for intelligent fault diagnosis. In this novel model, data augmentation is applied to both class and domain spaces, adversarial learning is employed to introduce adversarial perturbations, and a domain-based discrepancy metric is used to balance intra- and interdomain distances. The model can effectively learn more domain-invariant and discriminative features from multiple source domains to perform different generalization tasks for different working loads and machines. The feasibility of the DMDGN model is verified on two public datasets and one private dataset collected from practical production processes. Empirical results show that the DMDGN model outperforms several state-of-the-art models.

By considering the effect of long- and short-run correlation (LS) networks, we propose an LS network-augmented parametric portfolio selection model (LSNA-PP). First, we combine the dynamic conditional correlation-mixed data sampling (DCC-MIDAS) model with the planar maximally filtered graph (PMFG) method to construct LS networks and extract network topological characteristics. Second, we design portfolio weights as a function of these topological characteristics to construct the LSNA-PP model. Third, we apply the model to construct an international portfolio from 2010 to 2021. The empirical results illustrate the efficacy of the LSNA-PP model in two ways. First, the LSNA-PP model clarifies the economic interpretation of topological characteristics in portfolio selection, such as the positive effect of the long-run correlation network and the negative effect of the short-run correlation network on the weights. Second, the LSNA-PP model performs well in terms of return expectations, risk diversification, and attractive risk-adjusted returns, which are especially useful for stakeholders such as regulators, managers, and investors.