To investigate the mechanical properties and durability of UHPC bridge structures under varying environments, this study designed factorial experiments covering three curing temperatures, two sand types and mineral admixtures, then coupled accelerated durability tests (chloride penetration, sulfate attack, freeze–thaw) with on-site monitoring data. A DOA-RF model was developed by optimizing Random Forest parameters through the Dream Optimization Algorithm and validated against experimental results. Compressive strength reached 150–200 MPa, tensile strength 10–20 MPa; high-temperature curing increased flexural strength by ≈8 % yet reduced chloride permeability by 23 %, demonstrating superior marine durability. The DOA-RF algorithm achieved the lowest MSE and highest R² (0.75) among all benchmarked methods. Beyond engineering performance, the study proposes the first framework integrating UHPC with cultural-creative industries for accelerated marketization, branding and internationalization, offering a novel pathway for sustainable infrastructure innovation.

The goal of this research is to enhance road safety by delivering an affordable, non-intrusive, and effective drowsiness detection solution suitable for integration into mainstream vehicle systems. Driver drowsiness is a critical factor contributing to road accidents, often resulting in severe injury or death. In Malaysia alone, it is estimated that drowsiness causes between 2,000 and 3,000 traffic accidents annually. Conventional vehicle-based drowsiness detection systems, which rely on steering behavior or lane departure, often fail to detect early physiological signs of fatigue. Recent advancements in artificial intelligence (AI), particularly in computer vision and machine learning, offer new opportunities for developing low-cost, real-time drowsiness monitoring systems. This research proposes a driver drowsiness monitoring system (DDMS) that utilizes deep learning and visual behavior analysis to detect signs of fatigue through real-time monitoring of eye and mouth activity. A customized Convolutional Neural Network (CNN) is developed to classify eye states (open vs. closed), while yawning is detected using facial landmark analysis and Mouth Aspect Ratio (MAR) computations. The system is trained and evaluated using the MRL Eye Dataset, which consists of 4,000 annotated images, with data preprocessing and augmentation applied to enhance robustness. Through systematic experimentation and hyperparameter tuning, the model achieves a peak accuracy of 98%, with equally high precision, recall, and F1 Scores, using the Adam optimizer, a learning rate of 0.001, and 50 training epochs. The system also demonstrated strong real-time performance across varied lighting conditions, though challenges remain in scenarios involving occlusion (e.g., sunglasses) and extreme head positions. The results indicate that the proposed method is not only feasible but highly accurate, marking a significant advancement in proactive traffic accident prevention technologies.

With the development of artificial intelligence technology, text-driven image generation has gradually become a research hotspot. In terms of the application of traditional Chinese painting generation, due to the particularity of traditional Chinese painting themes, techniques, and artistic conception, existing models have problems such as cross-modal alignment deviation. To enhance the model's understanding of text semantics and improve the matching degree between text and generated Chinese painting images, a multimodal dataset of Chinese paintings is studied, and the Vector Quantized Generative Adversarial Network (VQGAN) is improved. A new multimodal text generation method is constructed by combining transformer-based bidirectional encoder representation, Transformer decoder, and contrastive language-image pre-training models. The results showed that when trained for 100 rounds on the general object dataset and Flickr30k dataset in context, the Inception Score (IS) values of this method were 39.6 and 5.1, respectively, and the Fréchet Inception Distance (FID) values were 2.3 and 1.7, respectively, which were better than models such as VQGAN. In the ablation experiment, the IS was 5.8, the FID was 10.3, and the cosine similarity was 0.82. The convergence was achieved after 60 rounds, which was better than other variants. The method proposed in the study has good adaptability to the generation of traditional Chinese painting. Although the expansion of complex scenes is slightly weak, the overall performance is excellent, filling the gap in traditional artistic semantic mapping and providing support for the digital dissemination and innovation of traditional Chinese painting.

Aiming at the problem of insufficient note-level attribute modulation and the difficulty of fusion of cross-genre musical elements, the study proposes a hierarchical conditional embedding mechanism and a symbolic feature conditional diffusion method. Through the dynamic gated fusion of note-structure and adaptive acoustic modulation guided by symbols, it optimizes the millisecond precision generation of melodic rhythm and the synergistic control efficiency of high-fidelity audio synthesis. This enables fine-grained controlled generation of cross-cultural heterogeneous music. The experimental results indicated that the model achieved 96.2% note localization accuracy in cross-cultural scenarios, which was 12.8% higher than the benchmark. The minimum value of beat synchronization deviation was 1.7 ms, which was 52.9% lower than the optimal comparison model. The average value of polyphony duration was 70.6%, an improvement of 9.8%. Differential scale fusion reached 12.5 tone level, breaking through the limit of twelve equal temperament. The peak memory occupation was 198.3 MB, and the energy consumption of a single song was as low as 0.142 kWh, reducing energy consumption by 29.4% compared with the traditional solution. The professional composition evaluation revealed that the cultural coordination degree of the heterogeneous style fusion fragment was 92.1%. The real-time generation delay stabilized at 2.8 ms, and the generation quality improved by 38.7% compared to the industrial standard. These results proved the model's comprehensive advantages in cross-dimensional control and artistic expression. This model can be integrated into digital audio workstations (DAWs) as either a plug-in or a cloud API. It provides creators with real-time, interactive generation and style transfer capabilities. It provides intuitive control over both macro-level structure and micro-level acoustic details via natural language commands or symbolic input. This significantly lowers the barrier to high-quality, AI-assisted creation. This drives the popularization and application of cross-cultural music fusion exploration.

This study aims to develop an enhanced Statistical Process Control (SPC) framework for the effective monitoring and forecasting of global temperature trends, addressing the limitations of traditional control charts in handling autocorrelated environmental data. The primary objective is to construct a modified Exponentially Weighted Moving Average (EWMA) chart capable of detecting both abrupt and gradual shifts in time series exhibiting seasonal and stochastic dependencies. The proposed approach models global temperature data using a Seasonal Autoregressive Moving Average process with exponential white noise (SARMA(1,1)L) to capture temporal patterns and residual variability. An analytical formulation for evaluating the one sided Average Run Length (ARL) is derived, enabling quantitative assessment of chart performance. The simulation analysis demonstrates that, by applying an optimized smoothing parameter, the proposed EWMA control chart outperforms the traditional EWMA and CUSUM charts, particularly in detecting small shifts with significantly lower ARL1 values. The findings confirm that the proposed model effectively tracks and predicts global temperature anomalies with high sensitivity and accuracy. The novelty of this research lies in integrating SARMA-based modeling with SPC design to improve detection reliability under autocorrelated and nonstationary conditions. This data-driven framework offers a promising tool for real-time environmental monitoring and climate forecasting applications.

Early and accurate breast cancer diagnosis remains critical yet challenging in routine practice. This study proposes a simple, reproducible pipeline that combines deep feature extraction from pre-trained CNNs (ResNet50, VGG16, EfficientNet-B0, DenseNet121, MobileNetV2) with classical machine-learning classifiers (logistic regression, SVM, k-NN, decision tree, random forest, gradient boosting, XGBoost, LightGBM, Naïve Bayes, and MLP). Features are computed after standardized preprocessing; class imbalance is addressed with SMOTE when present. We evaluate three image datasets (binary and multiclass) using accuracy, precision, recall/sensitivity, F1, and confusion matrices, and apply paired statistical tests across cross-validation splits. Findings: EfficientNet-B0+MLP and ResNet50+MLP achieve peak accuracies up to 99.6% on high-quality, balanced data, while DenseNet121+MLP with SMOTE attains 97.8% on imbalanced multiclass data. SMOTE yields substantial gains on imbalanced data and negligible effect on balanced sets; decision trees underperform consistently. Novelty/Improvement: Rather than a monolithic end-to-end network, we provide a modular, resource-aware blueprint that (i) disentangles feature extraction from classification, (ii) quantifies when imbalance correction matters, and (iii) reports clinically relevant error types. We further outline explainability with Grad-CAM/SHAP and discuss inference-time trade-offs and real-world workflow integration, offering an interpretable and deployment-friendly alternative to heavier end-to-end models.

This study developed an efficient object detection model for indoor environments, addressing common challenges such as occlusions, varying lighting, and cluttered scenes. We evaluated several YOLOv8 variants—ranging from nano to extra-large—and introduced an optimized YOLOv8x model. Our approach combines structured pruning, quantization-aware training, and advanced data preprocessing techniques, including augmentation and noise reduction, to improve model performance while reducing computational demands. The models were developed and evaluated using a carefully selected indoor object detection dataset featuring ten common categories. Performance was measured through key metrics like precision, recall, and mean average precision (mAP). Among them, the fine-tuned YOLOv8x clearly outshined the baseline models, reaching a training precision of 0.577, a recall of 0.572, and an mAP@0.50 of 0.537. When tested on new data, it demonstrated even better generalization, delivering a precision of 0.502, a recall of 0.528, and an mAP@0.50 of 0.480—proving robustness and reliability in real-world scenarios. These results demonstrate that pruning and quantization can significantly reduce model complexity without sacrificing accuracy, which helps to detect indoor objects. In essence, it is optimized for indoor object detection, offering promising applications in smart environments, surveillance, and robotics.

Objectives: This paper aims to assess the agricultural ecological and economic efficiency of the Yangtze River Economic Belt by using the data envelopment analysis (DEA) model to evaluate the regional agricultural level. Methods: Relevant data from 11 provinces and cities in the Yangtze River Economic Belt from 2010 to 2020 was collected from statistical yearbooks. Then, the agricultural eco-efficiency and economic efficiency were evaluated using the slack-based measure (SBM) model in the DEA model. Findings: The evaluation result of agricultural eco-efficiency was consistently higher than that of ecological efficiency. From a regional perspective, the eco-efficiency of the downstream area was higher than that of the middle and upper reaches. From the perspective of group division, only Guizhou and Chongqing had a high eco-efficiency. Improvement: The findings suggest that the overall agricultural eco-efficiency in the Yangtze River Economic Belt is low, and there is still a large space for development. It is necessary to further reduce agricultural carbon emissions and non-point source pollution and improve agriculture through technological innovation and other means.

The fourth industrial revolution has considerably enhanced technology, resulting in disruptive developments across various industries, including higher education. Generation Z, known as digital natives, has specific digital preferences, making mobile applications essential for improving their educational experience. The present study aims to investigate the digital transition in higher education, emphasizing using mobile applications as campus support services for Generation Z students. The study investigates the factors influencing mobile app acceptance and usage, intending to enhance educational support services' efficiency and quality among 100 students from different universities. By performing multiple linear regression, the study revealed that perceived usefulness is the most critical factor driving students' intention to use mobile apps in higher education. In contrast, other elements such as ease of use, competence, accessibility, and data privacy were not deemed significant concerns by the students. The findings are intended to advise higher education institutions on integrating mobile apps to assist Generation Z better, eventually leading to increased student engagement and satisfaction. This study emphasizes the significance of mobile technology in current educational contexts and offers practical insights for universities looking to use digital technologies to improve campus support services. However, the outcome may vary for students from different demographic and socio-economic backgrounds.

Personalized tourist attraction recommendations present a challenging problem in intelligent travel planning. Bangkok, the capital of Thailand, is a popular tourist destination offering a convenient metro system that enables travelers to plan their journeys easily. Leveraging this infrastructure, this study proposes a deep learning-based model designed to classify tourists into five categories: Nature Tourists, Cultural Tourists, Shopping Tourists, Historical Tourists, and Industrial Tourists. The model employs Neural Collaborative Filtering (NCF), utilizing deep neural networks to capture complex, non-linear patterns between users and destinations, surpassing the limitations of traditional matrix factorization methods. It integrates both user-related data, such as tourists’ opinions on destinations, and location-based data from the attractions themselves. To evaluate the model, data were collected from 30 stations along Bangkok's Pink Line, covering the northern part of the city and Nonthaburi province, and 31 tourist attractions along the route. Experimental results demonstrate high classification accuracy across tourism types: 96.26% for Nature Tourists, 80.59% for Cultural Tourists, 93.78% for Historical Tourists, 70.35% for Industrial Tourists, and 97.66% for Shopping Tourists. Furthermore, the study proposes three optimized travel routes tailored to tourist preferences: one for Nature and Cultural Tourists, another for Cultural Tourists, and a third for Historical and Cultural Tourists. By categorizing tourists based on their interests and recommending destinations accordingly, the model supports more informed and personalized travel decision-making. However, this current study serves as a prototype model and can be further applied to problems related to public transportation systems, such as deployment in mobile applications and integration with GPS positioning systems to enhance convenience and accuracy in providing tourist destination recommendations.