Urban brand visual systems shape city image and user experience across physical and digital touchpoints. However, existing methods are often static and weak in connecting user feedback with actionable visual optimization. To address this issue, this study proposes UXMap-BrandOpt, a user experience map-based dynamic optimization framework. Specifically, it constructs a dynamic user experience map, fuses visual, textual, emotional, spatial, and brand-perception features, and learns a requirement–visual element mapping to recommend optimization actions. A multi-objective decision module balances user satisfaction, brand recognition, cultural fit, visual consistency, and optimization cost. Experiments on public datasets and the Chengdu-UXBrand Dataset show that the proposed method outperforms eight baselines, achieving F1_pain=0.891, S_BR=0.842, S_VC=0.821, Acc_act=0.813, and a utility gain of 0.236. The Chengdu case demonstrates its ability to identify weak touchpoints and provide interpretable visual optimization strategies.
As the new energy vehicle (NEV) industry is increasingly characterized by personalized customization and green manufacturing, conventional centralized production scheduling approaches have become less effective in coping with dynamic disturbances and the coordination requirements of distributed manufacturing resources. To address this issue, this study develops a five-layer collaborative scheduling decision framework consisting of the physical, perception, negotiation, learning, and optimization layers, thereby operationalizing a closed-loop logic of “perception–negotiation–learning–optimization.” Specifically, the perception layer acquires real-time production-state information through a digital twin environment. These findings demonstrate the functional completeness and operational feasibility of the proposed framework for dynamic scheduling in distributed manufacturing environments and provide both theoretical support and practical guidance for intelligent scheduling system design in smart factories.
This paper proposes MCAD-Net, a multi-context collaborative decision framework for marine emergency management. The framework integrates context representation learning, risk-aware modeling, and stakeholder-aware decision fusion to support coordinated multi-stakeholder decisions under uncertain maritime conditions. By jointly encoding environmental, operational, and resource states, MCAD-Net dynamically estimates emergency risk levels and generates coordinated response strategies. Experiments on multiple marine datasets show that MCAD-Net outperforms optimization-based and reinforcement learning baselines, achieving a decision utility of 0.812 compared with 0.753 for the strongest baseline. These results demonstrate its effectiveness in improving coordination efficiency and decision reliability in marine emergency management.
Accurate shelf monitoring and category preference estimation are critical for modern retail management, yet existing pipelines largely focus on improving detection accuracy while neglecting calibration, structural alignment, and downstream decision relevance. This limitation becomes acute when deploying heterogeneous detectors such as Faster R-CNN, SSD, YOLO, or transformer-based models in real-world shelf environments characterized by occlusion, crowding, and promotional fluctuations. These results demonstrate that beyond boosting detection accuracy, aligning detectors with planogram constraints and uncertainty-aware aggregation directly enhances retail decision quality. This study highlights a paradigm shift from detector-centric benchmarking to decision-centric evaluation, offering both theoretical insights and practical tools for robust retail shelf management.
Online tourist reviews, a major form of user-generated content (UGC), are often short and unstructured, complicating the identification of tourist experience dimensions and their relationships. This study presents an integrated framework combining Topic-RoBERTa, topic-level semantic network analysis, and Graph Attention Networks (GAT) to extract experience topics and model their semantic associations. The authors validated the framework on 186,429 reviews from the OD-TripM TripAdvisor review dataset released by the Data Science and Computational Intelligence (DaSCI) research group on GitHub and the Yelp Open Dataset, identifying 12 tourist experience dimensions and constructing a community-structured semantic network. The results show that the proposed method can extract interpretable topics, reveal fine-grained attention-weighted associations, and provide a more structured understanding of tourist experiences. The findings offer valuable insights for tourism management and service optimization.
In the VUCA environment, emergent crises frequently inflict disruptive damage on new ventures. Given their inherent resource constraints, post-disruption organizational recovery constitutes a critical survival challenge for these ventures. However, self-reliant approaches prove insufficient to overcome such bottlenecks. This study proposes entrepreneurial ecological orientation as a novel strategic mindset that enables ventures to acquire exogenous recovery momentum through ecosystem embeddedness. Building on the strategic fit perspective between orientation and institutional environment, the authors employ fuzzy-set qualitative comparative analysis (fsQCA) to investigate the causal complexity underlying new venture resilience. The research advances crisis management literature by introducing an ecosystem lens to resilience studies, while providing actionable insights for constructing ecosystem-based risk mitigation strategies.
Visual search systems often rely on image-only embeddings, limiting semantic understanding—especially with visually similar but semantically distinct items. To overcome this, the authors propose a unified multimodal search framework that integrates generative AI-based captioning and image understanding for improved retrieval. Using Qwen2-VL-7B, the system generates rich captions from images, then fuses visual and textual embeddings via cross-attention. A re-ranking module refines the results. Evaluated on COCO Captions and Fashion-Gen, the model achieves BLEU scores of 0.78 and 0.75, CIDEr scores of 1.12 and 1.08, and retrieval mAP of 0.75 and 0.73. It also records NDCG@10 of 0.91 and 0.89, outperforming baselines like FineCaption, SuperCap, and MM-Transformer by up to 9%. These results validate the approach in bridging the semantic gap between images and text, enabling more accurate, context-aware search in applications such as e-commerce, multimedia, and large-scale retrieval.
This study investigates Information Systems (IS) success from an IT Business Value (ITBV) perspective by integrating constructs from the IS Success Model (ISS) and the Resource-Based View (RBV). An extended research model is developed to explain how core organizational IT resources, namely IT infrastructure and human capital, shape system perceptions, including performance expectancy, system quality, and information quality, which in turn influence IS continuance usage and ultimately firm performance. Rather than focusing on specific emerging technologies, this study emphasizes foundational IT capabilities that enable sustained system success and organizational value creation across different technological contexts. To capture multi-perspective data at the organizational level, the study employed a structured, role-based multilevel sampling strategy and tested the proposed model using covariance-based structural equation modeling (CB-SEM).
Ensuring real-time safety in smart manufacturing environments increasingly relies on vision-based object detection systems deployed at the edge. However, achieving a balance between inference accuracy and low-latency performance remains a key challenge, particularly when deploying deep learning models on resource-constrained industrial edge devices. To address this, the authors propose Factory-Aware YOLOv11(FA-YOLOv11), a lightweight object detection framework designed for latency-aware safety decision-making in factory settings. The method introduces a pruned and quantized YOLOv11 backbone, augmented with multi-scale feature fusion and a latency-gated decision mechanism to suppress unstable detections. Temporal filtering further enhances precision under noisy conditions. This study highlights the importance of latency–precision co-optimization for edge intelligence in smart factories and provides a deployable solution to improve factory safety through fast and accurate perception.
Enterprise strategic decision-making plays a critical role in guiding organizational development, particularly in the context of digital transformation where enterprises must respond to rapidly changing technological and market environments. However, traditional decision support systems often rely on isolated data analysis or statistical models, which struggle to capture complex relational dependencies among heterogeneous enterprise entities such as resources, technologies, and market factors. To address these challenges, this paper proposes SKD-Net (Strategic Knowledge-Driven Deep Decision Network), a unified framework that integrates enterprise knowledge graphs with deep neural decision modeling for strategic decision support. Overall, the results indicate that integrating enterprise knowledge graphs with deep decision networks provides a promising direction for developing intelligent strategic decision support systems in enterprise digital transformation scenarios.
Enterprise performance management increasingly requires analytical systems that move beyond correlational prediction toward intervention-oriented decision support. Existing approaches, however, often address hidden confounding, non-random sample selection, incomplete outcome observation, continuous treatment intensity, and policy optimization in isolation rather than within a unified framework. To address this limitation, this study proposes PRISM-Policy, a unified causal machine learning framework that integrates proxy-assisted confounding correction, observation-aware adjustment, continuous-treatment modeling, and policy learning within a shared latent architecture. Empirical evaluation on four benchmark datasets, together with a Criteo-derived continuous-treatment benchmark, shows that PRISM-Policy delivers the strongest overall empirical profile among the compared methods across estimation accuracy, robustness, continuous-treatment prediction, and policy-oriented evaluation.
This study investigates Information Systems (IS) success from an IT Business Value (ITBV) perspective by integrating constructs from the IS Success Model (ISS) and the Resource-Based View (RBV). An extended research model is developed to explain how core organizational IT resources, namely IT infrastructure and human capital, shape system perceptions, including performance expectancy, system quality, and information quality, which in turn influence IS continuance usage and ultimately firm performance. Rather than focusing on specific emerging technologies, this study emphasizes foundational IT capabilities that enable sustained system success and organizational value creation across different technological contexts. To capture multi-perspective data at the organizational level, the study employed a structured, role-based multilevel sampling strategy and tested the proposed model using covariance-based structural equation modeling (CB-SEM).
EMoGuideNet is a deep-learning model that integrates user emotion and behavior in multi-turn interactions, overcoming the current limitation of treating them as separate processes. By simultaneously streaming emotional conditions and behavioral responses, it captures their interactive dynamics. The model processes multimodal inputs (textual/acoustic) through a recurrent emotion tracking mechanism, enabling temporal emotion modeling, and uses multi-task learning to predict behavior. Evaluated on four benchmarks (MELD, IEMOCAP, DailyDialog, ReDial), it outperforms CoCoLM, SER-Meta, and BERT4Rec, achieving a statistically significant 3.9% improvement in emotion-recognition accuracy and 4.6% in behavioral Hit@10 on MELD (p < 0.01). The model also demonstrates robustness to noise and generalizes well to unseen settings, highlighting the effectiveness of emotion-conditioned behavior modeling. These advancements enhance adaptive dialogue systems by enabling responses with greater emotional relevance and situational awareness.
This study investigates the transformative role of digital technologies in driving structural economic change across manufacturing, services, and agriculture in emerging markets, with China as a central case. Using a mixed-methods approach that combines empirical sectoral data, policy analysis, and comparative case studies, the research uncovers how artificial intelligence (AI), the Internet of Things (IoT), and platform economies contribute to productivity gains, labor market restructuring, and inter-sectoral synergies. Findings reveal that while digitalization enhances competitiveness in manufacturing and enables service-sector expansion through fintech and e-commerce, agriculture remains digitally marginalized due to infrastructural and institutional deficits. The paper highlights how policy asymmetries, digital literacy gaps, and regional inequalities deepen digital divides, impeding inclusive development. A conceptual framework is developed to illustrate how digital infrastructure and institutional support mediate sectoral transformation.
The proliferation of digital services has led to a surge in cyber crimes, while traditional detection methods face bottlenecks such as labeled data scarcity, data imbalance, and poor adaptability to zero-day attacks. To address these issues, this study proposes CRAD-Defender, an end-to-end framework integrating self-supervised learning (SSL), graph neural networks (GNNs), and reinforcement learning (RL). It fuses MoCo v3 and GraphSAGE-GAT to capture individual behavioral features and cross-entity relational risks, adopts One-Class SVM for initial anomaly detection, and leverages DQN for closed-loop optimization of detection parameters. Experimental evaluations on UNSW-NB15 and CSE-CIC-IDS2019 datasets show CRAD-Defender outperforms baseline models with an F1-score of 0.91 and 0.89, ultra-low FPR of 1.1% and 1.3%, and low detection latency. Ablation studies verify the synergistic effectiveness of each core module, demonstrating the framework’s superiority in accurate, real-time cyber crime risk classification and anomaly detection.
Amid global carbon neutrality consensus and rising environmental awareness, new energy vehicles (NEVs) have become the core of green mobility, making green consumer continuous behavioral indicator prediction and targeted marketing optimization critical for NEV enterprises. Existing models struggle with integrating temporal behavioral and unstructured textual sentiment data, lacking adaptive fusion mechanisms, and dynamic prediction capabilities. To address these gaps, this study proposes a multimodal deep learning model TransBERT-X integrating Transformer’s temporal modeling proficiency and BERT’s sentiment extraction strengths. Trained and validated on four datasets (UCI Online Retail, CEVC, UCI Car Evaluation, Sentiment140), TransBERT-X outperforms state-of-the-art models (AI-Driven, LSTM, SEM, BiLSTM) across metrics (MSE, MAE, MAPE, R2). Results show it enhances consumer behavior identification accuracy, establishes an integrated “continuous demand prediction + precision marketing” framework, and provides a feasible technical paradigm for NEV industry marketing transformation.
The rapid digital transformation of the retail industry has generated increasing demand for intelligent business intelligence systems that can translate natural language managerial inquiries into actionable analytical and predictive decisions. However, existing enterprise analytics solutions either rely on rigid dashboard-based querying or isolated AI models, and thus struggle to jointly achieve knowledge grounding, executable analytical reasoning, and decision-oriented synthesis within a unified framework. RIDE constructs knowledge-grounded analytical contexts from enterprise repositories, employs multi-step reasoning–action planning to orchestrate tool-based analytics, generates grammar-constrained executable SQL for reliable database access, and integrates predictive forecasting to support decision synthesis. Overall, this study establishes a new paradigm for large language model–driven business intelligence systems that bridge human-like inquiry and machine-executable analytics, providing a robust foundation for intelligent and trustworthy retail digital management.
Cold chain logistics plays a vital role in preserving the quality of perishable goods during last-mile distribution. However, optimizing delivery routes in such scenarios remains a significant challenge due to the need to balance multiple conflicting objectives, including cost, timeliness, freshness preservation, and environmental sustainability. To address these complexities, this study proposes an enhanced multi-objective evolutionary algorithm, HA-NSGA-II-RL (Hybrid Adaptive NSGA-II with Reinforcement Learning), tailored for cold chain vehicle routing problems. The algorithm integrates three core innovations: forward-biased mating pool selection to improve diversity, destructive mutation to avoid local stagnation, and a reinforcement learning-guided reproduction scheduler for adaptive operator control. Overall, this research provides a robust, scalable solution for multi-objective logistics optimization in temperature-sensitive delivery networks.
Enterprise customer relationship management (CRM) increasingly relies on data-driven precision marketing to improve targeting efficiency and long-term customer value; however, existing CRM approaches face two critical challenges: accurately identifying customers with true incremental marketing impact and effectively translating predictive results into value-aware decisions under budget constraints while accounting for retention risk. To address these challenges, this paper proposes AI-CRM-PS, a unified AI-driven framework that integrates causal uplift estimation, customer lifetime value modeling, and churn-aware risk assessment into an end-to-end decision optimization process. Overall, this study highlights the effectiveness of decision-aware AI in CRM and provides a practical solution for achieving systematic and sustainable improvements in enterprise precision marketing.
The rapid digitization of museum collections demands intelligent systems capable of automatically classifying and managing large-scale artifact inventories. Existing methods often address classification and retrieval separately and fail to satisfy practical management requirements such as multimodal integration and system-level consistency. A deep learning-based system is proposed for the automatic museum classification and management, which integrates artifact localization, robust visual representation learning, multimodal semantic fusion, and management-oriented indexing within a single framework. Experiments conducted on large public museum datasets demonstrate that the proposed system outperforms existing methods in classification accuracy, reliability, and end-to-end automation performance. These results highlight the effectiveness of a unified, management-aware design in real-world museum applications.
