Machine learning always requires a large amount of labeled data, and the test data may have a different distribution than the training data. Transfer learning has proven to be an essential method for solving this problem in many fields. However, achieving successful transfer in graph datasets remains challenging, as the pre-training datasets must be large enough and carefully selected. This research looks at the inherent challenges of data scarcity and the need for robust models to increase the versatility and efficiency of Graph neural networks (GNNs)in various implementation domains. By examining the performance between trained GNNs and non-pre-trained GNNs, which can further demonstrate the generalization of the pre-trained GNN strategy and the significance of transfer learning to graph data.

Transverse confinement is important in enhancing the mechanical performance of ultra-high-performance concrete (UHPC) columns under axial compressive loads. This research offers a detailed review of the axial compression behaviour (ACB) of UHPC columns confined by transverse steel bars, steel tubes, and fiber-reinforced polymer (FRP) sheets, respectively. The findings indicate that using a combination of transverse steel bars, steel tubes, and FRP sheets significantly boosts the bearing capacity and ductility of UHPC columns, leading to increased peak stress and enhanced peak strain in the confined concrete. The ACB of UHPC columns restrained by transverse steel bars is influenced by the strength and volumetric ratio of the stirrups; higher strength and greater stirrup volume result a lot in enhanced performance. Similarly, the strength and thickness of the steel tube are key factors in the ACB of UHPC confined by steel tubes, with performance improving proportionally to these attributes. For FRP sheet-confined UHPC, the mechanical properties, thickness, and winding configuration of the FRP sheets critically affect the ACB. Thicker sheets and increased winding layers correlate with better mechanical properties. The insights provided in this study offer valuable guidance for engineers considering the implementation of UHPC in construction projects, providing a reliable foundation for optimizing ACB through appropriate transverse confinement strategies.

In the modern aviation industry, accurate prediction of complex flow fields is of great significance for optimizing blade design and improving engine performance. Although traditional computational fluid dynamics (CFD) methods can provide high-precision flow field information, they have long calculation time and high resource consumption, making it difficult to meet the rapid response requirements of engineering practice. As an emerging machine learning model, neural networks have gradually become an effective tool for flow field prediction with their powerful nonlinear mapping capabilities and high computational efficiency. This paper aims to combine machine learning technology to construct an efficient and accurate flow field prediction model method, and introduce new theoretical support for the design and optimization of aircraft engines. This paper first explains the basic concepts of machine learning and the shortcomings of current flow field prediction methods. Then, through three cases, the basic principles and application process of neural networks are introduced, including BP neural network, RBF neural network and UNet neural network methods, and the current status and superiority of neural networks in complex flow field prediction are analyzed in detail, providing an important reference for promoting the informatization and intelligent development of the aviation manufacturing industry.

With the increasing demand for garden landscaping, the development and application of small smart lawn mowers have become particularly important. This paper presents a series of innovative design concepts and maintenance strategies through an in-depth study of the modular design and intelligent maintenance systems of small smart lawn mowers. The article first outlines the basic concepts of modular design and, combined with the actual needs of smart lawn mowers, details the specific implementation of the design. Next, the article discusses the construction methods of the intelligent maintenance system, emphasizing technical applications in fault diagnosis and remote monitoring. Through theoretical analysis and empirical research, this paper aims to provide more efficient and reliable smart mowing solutions for modern garden landscaping.

This study presents a hybrid Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) model designed to classify electrocardiogram (ECG) signals into five categories, including normal and abnormal heart activities. The model leverages CNN layers to extract critical local features from ECG signals, while GRU layers capture temporal dependencies in heartbeat sequences. Squeeze-and-Excitation (SE) blocks were incorporated to enhance the model's focus on important signal components. Using the ECG5000 dataset from the UCR Time Series Classification Archive, the model was trained with data augmentation techniques such as time scaling, time shifting, and noise addition to improve its robustness. After training for 20 epochs with an Adam optimizer, the model achieved a test accuracy of 94.19%, demonstrating its effectiveness in distinguishing between different heart conditions. This automated classification system holds significant potential for aiding healthcare professionals in diagnosing heart diseases more accurately and efficiently, offering critical support in clinical decision-making.

The advancement of Geographic Information Systems (GIS) has brought a wide range of decision-making tools to almost all sectors. However, traditional 2D GIS systems often lack the depth and interconnections that are crucial to deep thinking in many applications of GIS, such as urban planning, environmental monitoring and disaster management. This paper describes the use of 3D modelling and Virtual Reality (VR) in GIS platforms as a means of enhancing the full comprehension of complex data. The ability to visualise cities, ecosystems or disaster-prone areas in 3D makes spatial data more intuitive and interactive. VR takes this one step further by allowing stakeholders to move around the virtual environment and interact with data in real time, improving their level of preparedness when making decisions. We discuss the practical applications of these technologies in the fields of urban planning, environmental conservation and disaster management. We also highlight some of the technical challenges involved in building a 3D GIS or VR, such as data processing and user interface design. The paper concludes with some future trends and possible developments in 3D GIS and VR.

Battery Energy Storage Systems (BESS) are the backbone of modern power grids. They allow for the increase of energy storage, peak shaving, or backup power. Due to their complexity and dynamics, BESS require high-advanced management methods to optimise its performance. This paper focuses on the integration of Artificial Intelligence (AI) into BESS, discussing three main pillars: system stability, battery usage optimisation, and predictive maintenance. The emergence of Artificial Intelligence and in particular deep learning, reinforcement learning, and neural networks, brings significant improvements in the modelling of complex reaction mechanisms, the adaptation to real-time data, and predictive maintenance. By analysing large datasets from various sources, AI can increase the precision of State of Charge (SOC) estimation, reduce maintenance costs, and improve the reliability of the system. The comparison with different case studies underlines the potential implementation of AI in real-life applications, which brings cost savings and increased system efficiency. This paper concludes that the power of AI enables new techniques for BESS management, and it would bring major benefits in the construction of more powerful and resilient energy systems as a whole.

Multimodal sentiment analysis (MSA) is an evolving field that integrates information from multiple modalities such as text, audio, and visual data to analyze and interpret human emotions and sentiments. This review provides an extensive survey of the current state of multimodal sentiment analysis, highlighting fundamental concepts, popular datasets, techniques, models, challenges, applications, and future trends. By examining existing research and methodologies, this paper aims to present a cohesive understanding of MSA, Multimodal sentiment analysis (MSA) integrates data from text, audio, and visual sources, each contributing unique insights that enhance the overall understanding of sentiment. Textual data provides explicit content and context, audio data captures the emotional tone through speech characteristics, and visual data offers cues from facial expressions and body language. Despite these strengths, MSA faces limitations such as data integration challenges, computational complexity, and the scarcity of annotated multimodal datasets. Future directions include the development of advanced fusion techniques, real-time processing capabilities, and explainable AI models. These advancements will enable more accurate and robust sentiment analysis, improve user experiences, and enhance applications in human-computer interaction, healthcare, and social media analysis. By addressing these challenges and leveraging diverse data sources, MSA has the potential to revolutionize sentiment analysis and drive positive outcomes across various domains.

Advancements in technology and societal changes have profoundly altered lifestyles, leading to an increased desire for global travel. Alongside this trend, the concept of the 'sharing house' has emerged as a popular alternative to traditional hotel accommodations. Sharing houses offer benefits such as flexible rental periods, a variety of housing options, and competitive pricing, making them increasingly attractive to travelers. Airbnb stands out as a leading platform facilitating this model. Analyzing Airbnb data provides valuable insights for government policy-making, urban planning, travel planning for renters, host profitability, and strategic decisions for Airbnb itself. This study utilizes data from Airbnb's London listings, focusing on seven key attributes. Employing unsupervised learning techniques like K-Means Clustering combined with Principal Component Analysis (PCA), the study identifies three principal components and two distinct clusters, achieving a silhouette score of 0.64. By visualizing these clusters on a map, the research offers guidance to the London government for a deeper understanding of host behaviors and assists renters in selecting more suitable accommodations and hosts on Airbnb.

TODO machine learning has made significant advancements in the field of structural health monitoring, offering flexible and efficient solutions for detecting both local and global damage in various infrastructures. Local damage detection focuses on identifying cracks and spalling in specific areas of concrete structures such as bridges, highways, and tunnels. Techniques such as artificial neural networks (ANNs) and deep neural networks (DNNs) have been successfully employed for surface defect recognition, demonstrating their applicability across different structural contexts. Additionally, low-cost methods using devices like smartphones have been explored for quick road integrity assessments, proving to be both practical and affordable. Global damage detection encompasses the classification of structural collapse modes and damage types, utilizing feature extraction and deep learning models to enhance accuracy in identifying large-scale structural failures. These studies underscore the growing role of machine learning and computer vision in improving the resilience and monitoring of infrastructure systems.