Journal of Engineering and Digital Technology (JEDT)

Analysis of Foreign Tourist Review by Natural Language Processing and a Lexicon-Based Sentiment Analysis Tool

Chakkarin Santirattanaphakdi — 2025-12-26

The current behavior of tourists is to search for online information to support travel decisions. Although online reviews have many advantages, analyzing large amounts of data is time-consuming and resource-intensive in extracting important information; therefore, systematic text analysis helps present more comprehensive and targeted information. This research aims to analyze reviews using dictionary-based sentiment analysis tools and to create an analytical report of reviews from foreign tourists toward Khao Yai National Park. A total of 12,035 reviews related to Khao Yai National Park were collected from online platforms. TextBlob, Flair, and VADER were used to classify opinions as positive, negative, or neutral. The results showed that VADER had the highest average accuracy at 76%. However, sentiment analysis also found reviews in which satisfaction scores conflicted with the content, showing the limitation of using sentiment analysis alone to reflect opinions, as opinions are complex and difficult to compare using a single standard. To address this issue, an analytical report was created by analyzing the relationships between key terms and related reviews using cosine similarity and summarizing the text using natural language processing with data visualization to reflect the strengths and limitations of the park. The analysis found that although Khao Yai National Park is praised for its natural beauty, resource richness, and biodiversity, some limitations are mentioned in the reviews, including different fees for Thai and foreign tourists, inconvenient and insufficient public transportation, and the need for valuable tourism experiences and activities suitable for all tourist groups. The results of this research provide basic information for tourist decision-making and are useful for developing guidelines to improve Khao Yai National Park to better meet tourists’ needs, and can also be applied as a model for other tourist destinations to enhance service quality and the competitiveness of Thailand’s tourism industry in the long term.

Exploiting Transformer Network for Nail Diseases Classification and Recognition

Aekkarat Suksukont — 2025-12-26

Diagnosing nail diseases is a complex task due to their similar visual characteristics, often requiring expert dermatologists for accurate assessment. Misdiagnosis can lead to ineffective treatment and prolonged patient discomfort. This study explores the use of a transformer neural network for classifying nail diseases, leveraging its ability to identify intricate patterns and subtle features that may indicate early signs of disease. The research focuses on three nail conditions: psoriasis nails, onychomycosis, and healthy. The model was trained with a carefully optimized set of hyperparameters to improve learning efficiency and classification performance. Experimental results showed that the network achieved a peak accuracy of 99.40%, demonstrating its ability to effectively distinguish between different nail conditions. This approach not only enhances classification accuracy but also has the potential to reduce the workload of healthcare professionals and speed up diagnosis. Ultimately, this advancement could contribute to the development of automated diagnostic systems, leading to improved patient care and treatment outcomes.

Improving Efficiency of Ternary Tree to Support Different Quality of Service Levels

Warakorn Srichavengsup — 2025-12-26

This paper presents three algorithms that improve the performance of Ternary tree to support different quality of service levels. These algorithms are Partial Access type 1, Partial Access type 2 and Adaptive Probability algorithms. For the proposed algorithms, users are divided into two classes, namely class 1 and class 2, with class 1 users given higher priority than class 2 users. In Partial Access type 1 algorithm, class 1 users randomly select one slot from the first two slots, while class 2 users can access all three slots. In Partial Access type 2 algorithm, class 1 users randomly select 1 slot from the first 2 slots, while class 2 users randomly select 1 slot from the last 2 slots. Third algorithm is Adaptive Probability algorithm. In Ternary tree algorithm, each user randomly selects 1 slot out of 3 slots. When viewed in terms of probability, each user randomly accesses each slot with a probability of 1/3. Adaptive probability algorithms use different probability values for each slot. For example, let the probability of accessing slots 1, 2, and 3 be 1/2, 2/5, and 1/10, respectively. Due to the different channel access behavior between class 1 and class 2 users, each class of users has different delay values. Therefore, these three algorithms can be used to support systems that require different quality of service levels. The results show that each algorithm can provide different quality of service, especially Adaptive Probability algorithm, which can adjust its parameters to accommodate different quality of service levels while maintaining an appropriate delay.

Improving the Efficiency of Vehicle Queuing and Product Conveyance through Simulation: A Case Study

Sureeporn Yeanyong — 2025-12-26

This research examines the process of enhancing product delivery efficiency. The objective is to reduce both vehicle waiting times and product conveying times in the warehouse, thereby increasing the overall number of deliveries. Simulation techniques were applied in combination with vehicle queuing disciplines, and warehouse management theory was incorporated into the research. The research, starting from data from the current process of the case study company, was collected and used to develop a computer simulation model in FlexSim, enabling both problem analysis and the exploration of improvement scenarios. Three queuing discipline scenarios, warehouse management from FSN analysis, and a combination of queuing disciplines with FSN analysis were proposed, resulting in seven improvement scenarios. The simulation results indicated that Queuing Approach 1, which prioritizes six-wheeled vehicles in combination with FSN analysis, was the most appropriate method for process improvement. This approach enhances process efficiency compared with the situation prior to the improvement. The average waiting time for all vehicle types was reduced, and the total number of vehicles exiting the system within the specified monthly period increased from 582 to 658. This increase in vehicle throughput enhances the opportunity to sell more products, resulting in a profit of 14,063,037.11 Baht per month for the case study company.

Optimal Allocation and Deployment of Roadside Units in Cloud-Based Internet of Vehicles Framework

Nay Myo Sandar — 2025-12-26

The research focuses on internet of vehicles (IoV) where vehicles are equipped with cameras and sensors to monitor traffic jams, accidents, and locations to ensure safety and comfort for drivers. To process sensor data effectively, cloud computing is used because of vast storage and processing capabilities. However, transferring data from sensors to cloud can be challenging due to bandwidth and memory constraints. Therefore, a cloud-based internet of vehicles framework is proposed incorporating Roadside Units (RSUs). RSUs can buffer video streams from vehicles and send them to cloud services. With RSUs in the framework, total latency for transferring video streams to cloud services can be significantly enhanced. In this research, dynamic programming approaches are applied to determine how many RSUs are needed at the lowest cost and greedy algorithm is implemented to prove the optimal solution from dynamic programming. Furthermore, K-means clustering algorithm is applied to find the best locations for RSUs. According to numerical results, the proposed methods can determine the optimal number of 6 RSUs with the minimum cost and allocation of RSUs to serve video streams across regions.

Optimal 2DOF-PID Controller Design Using Whale Optimization Algorithm

Kittisak Lurang — 2025-12-26

The proportional-integral-derivative (PID) controller was first introduced in 1922. It has been widely accepted in industry for almost a century, because it can improve transient and steady-state responses as well as easily implementation. However, PID controllers tend by nature to excel in one aspect of system performance due to its trade-off. When the PID controller is designed to achieve input tracking, the load regulation performance of the system is then reduced, and vice versa. This problem can be solved by using a two degree-of-freedom PID (2DOF-PID) controller. This paper presents the design of the optimal 2DOF-PID controller by using the whale optimization algorithm (WOA), one of the most efficient metaheuristic optimization techniques for the time-delayed systems having slow responses and the servo systems possessing fast responses. Results obtained by the 2DOF-PID designed by the WOA will be compared with those obtained by the 1DOF-PID controller. From the simulation results, it was found that the 2DOF-PID controller designed by the WOA algorithm can effectively control the time-delayed system and the servo system. A maximum reduction in the IAE has been achieved, with 19.22% for the time-delay system and 17.14% for the servo system. Consequently, faster and smoother tracking and load regulation responses have been satisfactorily obtained once compared to those of the 1DOF-PID controller.

Rainfall Impact on Soil Behavior: Landslide Risk Assessment Using Physical Modeling

Laddawon Dul — 2025-12-26

Landslides are recognized as major geological hazards that can cause significant damage in mountainous areas, where risk factors often include high rainfall, specific soil characteristics, and road construction across elevated terrain. This study considers to analyze soil behavior under continuous rainfall conditions in a high-risk area along Highway No. 1096, Samoeng District, Chiang Mai Province, using physical modeling in a laboratory setting. Two types of soil masses were examined: natural roadside soil with a unit weight of 15.30 kN/m³ and compacted shoulder soil with a unit weight of 18.34 kN/m³, tested under four rainfall intensity levels. The results revealed that the natural soil mass collapsed when rainfall reached 60 mm/hr, while the compacted shoulder soil failed at 160 mm/hr. A landslide intensity index of 0.75 was identified as a critical threshold, marking the onset of continuous failures in natural soil and sudden failure in compacted soil when rainfall exceeded 220 mm/hr. Based on these findings, the study proposes a preliminary warning system using both rainfall thresholds and the landslide intensity index. A Yellow Alert level 60–160 mm/hr is recommended for initial monitoring of soil movement, Dark Yellow for evacuation preparedness, and Orange Alert for road closure when rainfall exceeds 220 mm/hr. Additionally, four distinct types of landslide behavior were identified, providing valuable insights for future prevention and risk management planning in landslide-prone areas.

Solving Factory Maintenance Problem: A Case Study of a Semi-Finished Food Product Manufacturing and Distribution Company in Uttaradit Province

Adul Phuk-in — 2025-12-26

This research focuses on scheduling maintenance for a semi-finished food product manufacturing and distribution company in Uttaradit Province. Heuristic algorithms, neural networks (NNs), and local search (LS) were used to create a mathematical scheduling model that solves the preventive maintenance (PM) problem. This is needed for production to keep going. The researchers tested the developed program with 2³ factorial experiments to find the appropriate parameter values for the answer. The research collected data on both small and large problems, and the program was able to find the answer value for scheduling maintenance efficiently. It obtained a makespan value, which was close to and matched the lower bound, reflecting the efficiency of the neural network. The local search method was employed to solve the problem. In addition, the data collected before and after the research for 6 months found that the total cost decreased to 1,915,062 baht, down from the original 422,396 baht, or 9.93 percent. The mean time between machine failures (MTBF) increased to 83.84 hours or 35.27 percent, showing a decrease in costs in terms of time and maintenance.

Transforming Unstructured Data in IT Project: A Comparative Study of Zero-Shot and Generative AI Text Classification

Cai Tung-lersloy — 2025-12-26

In today's world, we have a lot of messy, unorganized data from things like comments, interviews, and images. This is especially true in IT projects, where there's often too much information to handle easily. Our study looks at how we can turn this messy data into useful numbers and insights using smart computer programs. We tested two main methods: Zero-Shot Text Classification and Generative AI Text Classification. Zero-Shot is like having a smart assistant that can sort information without needing examples first. Generative AI is more like having a creative writer who can come up with new examples to help sort information. We asked 42 participants with experience in working with unstructured data to answer some questions, then used these methods to analyze their answers. We found that Zero-Shot works better for information that has clear patterns, while Generative AI is good at handling more complex or unclear information. Our results show that choosing the right method can make a big difference in how well we understand and use the data. Zero-Shot was about 15% more accurate for well-organized information, while Generative AI was 20% better at dealing with complex, messy data. This research helps companies and researchers choose the best way to make sense of their data, especially in IT projects where there's often too much information to handle manually.