Thailand is becoming one of the top strategic gateways for trade, investment, and logistics in the region. This includes improving connectivity between areas domestically and internationally[1], as well as developing digital and logistical infrastructure to facilitate trade and investment. By doing so, Thailand aims to be a key player in the global marketplace[2] and attract more international business.[1] Customs brokers act as intermediaries[3] between businesses and customs to perform customs clearance. They can be a great asset when it comes to international trade. These professionals are experts in navigating Customs' and other government agencies' requirements, which can ensure that any shipments reach their destination smoothly and efficiently. Dealing with complex processes and handling documents with many details is a great challenge that customs brokers face. However, the ability to adapt to these challenges and have even expanded their role in the supply chain by adopting new technologies like digitalization and advanced innovation of customs declaration systems. By doing so, they have been able to make the clearance process more efficient and modern, while still ensuring compliance with customs duties and other laws. It is impressive to see how customs brokers have been able to keep up with industry changes and remain an integral part of the supply chain. The customs brokerage industry plays a pivotal role in international trade by facilitating the seamless movement of goods across borders while ensuring compliance with ever-evolving regulations. This paper introduces a transformative vision for the customs brokerage business, leveraging modern technology to enhance efficiency, accuracy, and customer experience [4]. Modern technology in the “Easy Paste” model marks a paradigm shift in the customs brokerage industry. This visionary approach streamlines operations and positions customs brokers as crucial partners in the global trade ecosystem, fostering economic growth, sustainability, and innovation.
The k-means clustering algorithm calculates the distances between data points within each cluster and determines new centroids for each cluster until the data points remain unchanged within their respective groups. This process affects the iteration of the k-means algorithm, leading to an increase in computation time. Consequently, the overall execution time of the algorithm is prolonged. This research presents a process that utilizes the farthest first and canopy algorithms to reduce the number of iterations in the k-means clustering. This improvement focuses on enhancing the centroid initialization method for each cluster. The research study utilized ten datasets from the UCI machine learning repository to test the effectiveness of improving the k-means clustering algorithm. It compares the number of iterations and the sum of squares error as performance metrics. The results showed that the farthest first and canopy algorithms significantly reduce the number of iterations in the k-means clustering when compared to random centroid initialization. These research findings confirm the hypothesis that the method of centroid initialization for each cluster impacts the iteration process of k-means clustering.
In order to detect defects on semiconductor components, machine learning (ML) and deep learning (DL) approach are introduced into vision inspection system where these ML/DL approach capable to identify object with complex appearance and eliminate the task requirement of placing PCB or electronic component on special fixture and jig. However, there is a small dataset issue occur when implementing ML/DL approach into vision inspection application due to collecting such data is difficult and expensive in industry application. A novel defect classification system using extremely small datasets is proposed, named GENSS, where the GEometric variant property is used to geNerate synthetic data, and then uses Structural Similarity and image hashing approaches to automatically select pre-trained models. Early experiments showed that the suggested method is a practical approach that can achieve high accuracy, at 85.71 %.
Predicting volatility in the Foreign Exchange (FOREX) market is of paramount importance for stakeholders, including portfolio managers, risk managers, and central banks. It significantly impacts risk management, trading strategies, and monetary policy formulation. Traditional finance models, such as Autoregressive Conditional Heteroskedasticity (ARCH) and Generalized Autoregressive Conditional Heteroskedasticity (GARCH), have been extensively employed for modeling and forecasting financial market volatility. In recent years, machine learning (ML) algorithms, including Support Vector Machines (SVM), Neural Networks (NN), and Deep Learning (DL), have emerged as promising approaches for predicting complex financial time series data. This research presents a systematic and performance-based evaluation of traditional finance models and ML algorithms for predicting FOREX volatility, with a specific focus on the EUR/USD currency pair using daily close prices. The comparative analysis assesses the predictive performance of ARCH, GARCH, SVM, NN, and DL models, utilizing the Root Mean Square Error (RMSE) as the primary performance metric. The findings of this study indicated that traditional finance models, including ARCH and GARCH, exhibited reasonable performance in capturing certain aspects of FOREX volatility. However, ML algorithms surpassed these models, with the SVM, NN, and DL models demonstrating superior predictive capabilities. The insights derived from this analysis hold significant value as a guide for decision-making processes for various stakeholders. Additionally, this research contributed to the existing body of knowledge on FOREX volatility prediction, paving the way for future research in this domain.
Due to the digitization of healthcare, EHR systems are commonly used to maintain vast amounts of personally identifiable information and protected health information. There are, however, significant concerns regarding the security and privacy of EHR systems due to the dangers of unauthorized access, data intrusions, and identity theft. This study endeavors to identify methods and techniques that can be utilized to safeguard patient information contained in electronic health records. Examining the current security measures for EHR systems, highlighting the issues they cause, and proposing novel countermeasures against these attacks. Our research concentrates primarily on EHR system encryption techniques, authentication processes, and auditing procedures. In addition, the study also discusses the challenges associated with deployment and regulatory requirements, as well as the impending developments in EHR system security. By prioritizing the security and privacy of their electronic health record systems, healthcare providers can protect patient data, maintain regulatory compliance, and earn the trust of patients and other stakeholders.
This research aims to create a model for measuring the supply chain management performance and sustainability of community product in the mountainous region of northern Thailand in order to improve performance activities. The indicators were first studied by reviewing literature and collecting the data by surveying from organic Gymnema tea factory in Chiang Mai province. The SCOR model technique was used to examine 21 appropriate indicators of performance measurement which were verified by the researcher and related experts in that area. Next, the weight of the criteria was calculated. The result showed that the experts focused on society the most, nearly on the environment and followed by the economy. Finally, the performance measurement model of all indicators was generated to evaluate the performance of sustainable supply chain management of the tea product and the related community product in the nearby area.
When a cyber incident occurs, digital forensic is then essential for investigating how hackers compromised the system or how malware functioned. In this paper, we focus on Windows forensics which is one important branch of digital forensics. Windows forensics can be performed using some existing investigation tools that are expensive and require training before using them, while the current number of well-trained staffs in the cybersecurity field is limited. Moreover, in the step of evidence analysis, Windows forensic investigators need to manually extract certain files such as Windows registry and Windows event logs, which is a repetitive and time-consuming task. Therefore, we propose AXREL as an automated Windows evidence extracting application to facilitate new Windows forensic investigators by providing a user-friendly GUI. Our application is developed by Python 3 on the Windows platform. It can automatically extract Windows registry and event logs, which are the primary sources of evidence for Windows forensics.
This paper proposes competitive efficiency study and methods for detecting SQL injection by using rule base and machine learning. Rule-base employs the black-list word technique while machine learning includes Naive Bayes, random forest, and the K-Nearest Neighbor. We also apply SMOTE to improve the prediction rate on a minority class (injection class). As a result, machine learning algorithms that use the random forest and KNN efficiently detect very similar patterns as good as rule-based detection.
In this research, the optional backbone for YOLO model, called SimRepCSP (Simple Re-parameterization Cross Stage Partial module) was proposed, the model aims to address the goals of reducing training and application costs while enhancing model performance on specific dataset. SimRepCSP is constructed by combining Convolution modules and Re-parameterization Convolution (RepConv) modules, similar to YOLOv7, with the Cross Stage Partial (CSP) network serving as the connection between modules. By incorporating SimRepCSP into YOLOv8 as an alternative backbone, the research aims to achieve improvements in both training efficiency and model performance compared to the original backbone. The experiments were conducted using the GlobalWheat2020 dataset, and the results demonstrate a reduction in training and application costs, as well as higher performance metrics compared to YOLOv8 with the original backbone.
The study by the World Health Organization (WHO) reveals that Thailand has the highest road traffic death rate in Asia. Based on the data obtained during 2019–2021, heavy duty truck vehicles (HDVs) had steadily tended to have an accident, causing more economics losses than others. While the statistics lead to a question whether those accidents were caused by reckless road usage or other possible causes, existing studies cannot accurately explain as they were conducted from interviews with offenders and witnesses. Those studies also did not take topography into account while it is indeed different in Thailand, leading to errors in analyses and incorrect solutions. Hence, this study aims to identify and analyse factors from both the actions of drivers and environment. The researchers use GIS to accurately conduct both spatial analysis and locational analysis while taking the factors of the routes whether they are inbound or outbound, the unsafe environment in each region and the data of driver fatigue obtained by using ADAS with DMS into consideration to supplement the existing studies. The researchers expect the analysis to be fully utilized by related authorities to improve unsafe conditions and form a strategy to manage and reduce accidents.
The usefulness and popularity of smartwatches in the marketplace has led to a large variety of brands and models, each utilizing different technologies and processing methods to detect physical activities and measure vital metrics. Consequently, the reported values of widely available metrics of heart rate and step count can exhibit significant variations across these devices compared to standard medical equipment. This research proposes a model to standardize these values using a medical-grade device as a reference for heart rate and a manual clicker for step count. Descriptive statistics reveal that while the detected median values of step count align closely with actual steps, there is a notable oscillation in the overall data distribution. Fitbit devices consistently report lower step counts with high variability, while Garmin devices demonstrate more accurate step counts. Fitbit devices provide more precise heart rate measurements, while Huawei's measurements are less so. A linear regression model is used to effectively refine smartwatch measurements across multiple models, achieving a high level of accuracy as compared to a medical-grade sensor. Gender does not significantly impact the modeling adjustments.
The rapid advancements in technology have brought about significant disruptions in the job market for computer-related fields, necessitating a comprehensive understanding of emerging trends and evolving demands. This research focuses on the analysis of computer-related job advertisements to uncover co-occurrence information that employers seek within the context of disruptive technology and evolving business models. To address this need, we propose CID-CJA (Co-occurrence Information Discovery in Computer-related Job Advertisements), a novel method for extracting valuable insights from job advertisements. CID-CJA empowers job seekers to identify the knowledge and skills that are relevant in the current landscape, while also assisting universities in aligning their curricula with the dynamic needs of the industry. The flexibility of CID-CJA enables users to conduct detailed drill-down analyses on specific elements of information. The effectiveness of CID-CJA is validated through experimental studies conducted with real job advertisements, which demonstrate its ability to reveal valuable information that is beneficial for computer-related professionals.
In regulating gene expression and identifying relevant biomarkers for diseases like cancer, chromatin interaction plays a crucial role. Thus, the volume of HiC data is rapidly increasing as researchers strive to understand chromatin interaction. Most of the work focused on using a genome browser to visualize Hi-C data and genomic features for chromatin conformation. However, comparing chromatin connection and how much similarity exists between different cell types is still unexplored. As a result, this work used the Pearson and Spearman R correlation technique to investigate the similarities between Human ES and IMR90 Fibroblast cell types. By splitting Hi-C data into small subsets, we discovered that both cells have a significant degree of resemblance. Finally, we compared four different normalization methods based on their coefficient of variation results and running time on Hi-C data. Results demonstrate that min absolute scaling is better for Hi-C data as it has a low coefficient of variation than other methods and can perform fast for large Hi-C data.
Concerts or fan meetings usually attract lots of attention, generating huge demand for tickets for these events. But, existing ticket purchase systems are unable to efficiently and transparently accommodate this. We often hear scandals of bot-controlled clients gobbling up tickets within seconds after sales open or celebrities acquiring prime tickets exceeding individual quota. These problems arise because existing systems are centralized. This can be a single point of failure and the controlling authority can ban or give privileges to certain users. This work sets out to remedy these problems using a blockchain-base solution that, by nature, is highly decentralized. We have created and deployed EVM (Ethereum Virtual Machine)-based smart contracts for ticket sales platform on two EVM-compatible blockchains, Ethereum (ETH) and Avalanche (AVAX). These contracts inherit heavily from the ERC-721 standard for NFT (Non-Fungible Token). The platforms on both blockchains engender the desirable transparency. However, the platform on Avalanche is much more efficient and economical to deploy and utilize. We have open-sourced the code for our platform on Github. Visit the following link to fork or check it out: https://github.com/JesperBerben/TicketNFT In addition, you can now interact with our platform on Avalanche which has been deployed and verified at the address linked to below: https://snowtrace.io/address/0x17fd3f6cf6cbff75604b63f1ca12c6db29730a9c
Programming is a fundamental aspect of engineering and science that requires both knowledge and skills in programming. It is generally accepted that the more learners practice programming, the better their skills are. The online learning platform KruCode has been specifically designed using the Minimum Viable Product (MVP) method for programming courses. It is utilized by students to practice their coding abilities online, while instructors employ it to manage tags, coding questions, assignments, and exams in their classes. When code is submitted by students, it is promptly evaluated by KruCode to verify the presence of syntax mistakes. Students' error messages can be displayed in the system, and their work can be immediately guided. This information was beneficial for beginner students to learn coding and debugging. Students will also see their programming results as covered by the given test cases. However, other features should be developed, such as a dashboard and problem suggestion function, and increase safety for exams.
In the field of digital image processing, the image data are required to operate for some enhancement operations such as image filtering, image restoration, image transformation, and so on. The memory allocation of image data is presented in an array and executed by two-dimensional matrix-matrix multiplication. The convolution operation is an essential operation among many filtering algorithms (e.g. Gaussian filter). In this paper, the characteristics of different convolution methods have been examined, and the important role of convolution and advanced parallel version of 2D convolution to improve the performance and instructions per clock cycle on a multi-core architecture by using AVX and OpenMP. Generally, convolution is performed by sliding the kernel over the image and moving the left side direction of the kernel through all data of the input matrix until all values of the image have been calculated. This paper intends to utilize and analyze how 1D, 2D, and 3D convolutions work in image processing and the parallel versions of 1D and 2D convolutions using padding on CPUs are proposed.
The study of asteroids, the moving rocky objects, not only makes feasible prevention of hazardous collisions, but also provides better understanding of the solar system in it's early stage. However, existing software for asteroid detection requires manual parameter setup, which is a sensitive task requiring an experienced person. Moreover, the sequence of images contains only the brightness of each image while the key feature of asteroid detection is its movement. In this research, we propose a contrastive deep learning model to learn the motion representation of asteroids in a sequence of images. The representation is used to classify a sequence of images by investigation of distance calculation using Euclidean distance and cosine similarity. Moreover, simple classifiers including k-nearest neighbors (KNN) and logistic regression (LR) are implemented to evaluate their ability to classify the motion representation. The representation generation model is trained on sky images from the Gravitational-wave Optical Transient Observer (GOTO) survey. For motion representation, the classification results show that the best classifier achieves F1 score of 88.32% on the validation set and 86.60% on the test set. Moreover, k-nearest neighbors model underperforms the best model by -3.62% of F1 score in the test set. As a result, our approach replaces the hand-engineering process and produces more promising performance.
This work represents a fundamental developing strategy for the self-creation of digital-format geologic maps in the form of mobile apps. The digital geologic maps are apparently more convenient to work with than traditional paper-format geologic maps as the maps are user-interactive, self-directed, and can be easily updated. The digital geologic map mobile apps have recently become significant interactive presentation tools for geology students, geoscientists, and even the public. However, lack of technical methods and software development of those apps have yet revealed. In this study, we provide brief guidelines for database and software development methods to build a geologic map mobile app. The software used in this work is open source. The database file including geological information, spatial relationship and distribution, and a base map of Thailand, was prepared through a geographic information system or GIS. The hybrid app development process, which includes two open-source frameworks, namely Ionic and Angular, is key to produce a mobile app representing geologic maps via mobile devices. Additionally, basic programming language skills such as HTML, CSS, and JavaScript seem to be required for software coding. Following these steps, we produced a mobile app, named DooHin, for displaying a digital geologic map of Thailand with a scale of 1:50000 that is available via the app stores. This app is apparently practical and useful in fieldwork as it encourages a user-interactive experience, general geological information, and a user-positioning system.
Oral tissue detection is an important task in many image-based computer aided analyses of oral healthcare. By detecting the shape of oral tissue in the photographic image, one can eliminate other non-essential parts, which can help optimize the further processes in the pipeline, for example, oral lesion detection, cancer classification, oral tissue segmentation, and dental health analyses. Manual labeling of data by humans can be inefficient, time-consuming, and error-prone, due to subjectivity, intra- and/or inter- observer variability between experts. Therefore, in this paper, we aimed to comprehensively evaluate deep learning-based models for detecting the oral tissue in various perspectives of photographic images. We studied four different state-of-the-art object detection models including Faster R-CNN, RetinaNet, SSD, and YOLOv5. Our models provided good results in the oral tissue detection application, with the Intersection over Union (IoU) over 90 % and F1-Score over 97%. We found that the Faster R-CNN with ResNet50 backbone and RetinaNet with ResNet50 backbone were the most accurate models in detecting objects. Alternatively, Faster R-CNN with MobileNetV3 backbone and SSD with VGG16 backbone were the best models in terms of processing speed, making them well-suited for handling large datasets. We hope that our findings here contribute to the improvement of the image-based oral healthcare analysis system in the future.
