Siriwan Kajornkasirat

Introduction. This study investigates library user clusters to improve book recommendations. Data on daily borrowing behaviours from 2020 to 2025 were collected from the Prince of Songkla University, Pattani Campus library. Method. Users were clustered using K-means. The optimal number of clusters was selected via the elbow method combined with silhouette scores. Borrowing-pattern relationships were analysed using association rule mining with the FP-Growth algorithm (minimum support 15%, minimum confidence 90%). Analysis. The study examined the distribution of borrowed book categories and evaluated clustering performance using precision, recall, F-measure, and hit rate. Association rules were extracted to identify co-borrowing trends among categories. Results. Languages was the most borrowed category, followed by History & Geography and Science. The best user clustering solution was k = 3, achieving precision 82.98%, recall 98.55%, F-measure 0.8815, and hit rate 97.57%. The strongest rule showed that users borrowing Economics [330], Social Sciences [300], and History of Europe [940] were likely to also borrow History of Asia [950]. Associations were also found among life sciences, biology, physics, and chemistry. Conclusions. The study results indicate patterns of interest and knowledge connections derived from borrowing behaviour, which can be used to improve library services and serve as a model for data-driven library management.

This study develops a scalable customer segmentation framework for translating large-scale retail transaction data into actionable marketing and operational insights. The dataset comprised 3,150,173 point-of-sale records collected from 37 branches of a Thai retailer between January 2022 and March 2023. After preprocessing, 3,140,395 valid records were retained and aggregated into RFM profiles for 162,234 customers. Continuous Recency, Frequency, and Monetary variables were standardized using StandardScaler and clustered using K-Means. Candidate solutions from k = 2 to k = 8 were evaluated using WCSS and mean Silhouette Scores. A four-cluster solution, with a Silhouette Score of 0.3381, was retained based on the elbow pattern and business interpretability. The resulting segments were Champions, Potential Loyalists, At-Risk Customers, and Hibernating Customers. Champions represented 20.5% of customers and generated 57.0% of revenue, while Champions and At-Risk Customers jointly accounted for 52.3% of customers and 88.5% of revenue, indicating substantial revenue concentration rather than a strict 80/20 distribution. Product composition, weekday and monthly sales patterns, and branch-level performance provided additional operational insights. Unlike ranked RFM scoring, the framework retains continuous behavioral information and avoids predetermined score boundaries. The findings support segment-specific customer management, although validation across other retailers, longer periods, and alternative outlier treatments is required.

Oil palm is an important economic crop that is widely cultivated, especially in Southeast Asia. Thailand is one of the world’s largest producers and exporters of palm oil. Efficient management of oil palm plantations is crucial for increasing yields and minimizing agricultural losses. This study aimed to develop a smart oil palm plantation and production management system. This system utilizes Internet of Things (IoT) technology and an integrated supervised machine learning model utilizing regression analysis to monitor and control agricultural equipment within the plantation. MySQL database was used for management of sensor data. Python (version 3.9.6) programming and Google Map API were used for data analysis, spatial analysis and data visualization suite in the system. The results showed that the data from the sensors are displayed in real-time, allowing plantation managers to monitor conditions remotely and make informed adjustments as needed. The system also includes data analysis and data visualization tools for decision-making regarding production management. The model attained an accuracy of over 95%, which reflects its reliability in performing the specified prediction task. The system serves as a support tool for automating soil quality monitoring, fertilization, and field maintenance in oil palm plantations. This enhances productivity, reduces operational costs, and improves yield planning.

This study investigates the statistical and machine learning models to improve sales forecasting accuracy in the retail sector by incorporating both transactional and environmental variables. Motivated by the limitations of traditional data handling systems within a case study company, the research proposes an integrated, centralized information system that enhances data accessibility, reduces redundancy, and supports timely decision-making. Multiple forecasting approaches—including SARIMA, SARIMAX, LSTM, Ordinary Least Squares (OLS), and Poisson regression—were evaluated using historical sales and weather data. Results indicate that regression-based models (OLS and Poisson) outperformed time series and deep learning models in terms of model fit and predictive power, emphasizing the effectiveness of simpler, interpretable methods when relevant features are included. The study also demonstrates that weather conditions, such as humidity and temperature, exhibit moderate correlations with sales volume, though their direct predictive contribution is limited when used in isolation. This data-driven framework offers a scalable solution for retail operations, contributing to cost reduction—such as minimizing reliance on third-party business intelligence tools—and promoting sustainable competitive advantage.

This research presents the development of the intravenous drug administration application for pediatric patients using augmented reality (AR) technology, with a primary focus on aiding nursing students in administering medications accurately to reduce the risk of errors. The system architecture encompasses two core components: the creation of medication preparation videos and detailed drug information, and the design of a mobile application featuring medication list display, drug dosage calculation, user satisfaction assessment, and intravenous drug information addition. The system classifies users into administrators and nursing students, allowing administrators to manage user information in the member database. the application seamlessly integrates Visual Studio Code, flutter, dart programming language, firebase database, and AR.js Studio for QR code-linked videos. Operating in four main parts, namely users, mobile application, member database, and results display, the IDA application enables users to log in, access detailed drug information, calculate dosages, and view AR-based medication preparation videos. Tested with 111 nursing students, the system demonstrated functionality, completeness, and accuracy. The Likert scale-based evaluation revealed high satisfaction levels in content, design, functionality, and benefits received, affirming the intravenous drug administration application's effectiveness in pediatric intravenous drug management through AR, offering an innovative solution for nursing education and error reduction.

This research aimed to investigate a suitable algorithm for customer segmentation using as customer behavior indicators the recency, frequency, and monetary (RFM) values of the customers. The clustering algorithms K-means, fuzzy C-means, and self-organizing neural network (SONN) were compared for finding the most appropriate algorithm. The customer segmentation was analyzed using association rule mining with the frequent pattern algorithm (FP-Growth). Data on retail transactions during January 2021 - May 2023 were obtained from Tuenjai Company, Thailand, with a total of 202,469 records. The results from the three algorithms were compared by the silhouette coefficient (SC), Calinski-Harabasz (CH) index, Davies-Bouldin (DB) index, iteration count, and execution time. The results showed that the K-means algorithm was the most suitable algorithm for customer segmentation in this study. K-means clustering grouped the customers into three groups here labeled as “important value”, “general development”, and “lost”, based on the RFM values. There were 38 rules for the important value segment, and two rules each for the general development and the lost groups. These results could be useful to the business organization for improving the customer experiences, increasing sales, preparing or promoting products, and stock management efficiency.

The spread of infectious diseases, such as COVID-19, presents a significant problem for public health and healthcare systems. Digital technology plays an important role in achieving access to healthcare by enhancing device connectivity and information sharing. This study aimed to develop, implement, and demonstrate a tracking and surveillance system to enhance monitoring for emerging infectious diseases, focusing on COVID-19 patient profiling. The system integrates IoT-based wearable devices, an artificial intelligence (AI) camera for real-time monitoring, and a MySQL database for data management. The program uses Charts.js for data visualization and Longdo Map API for mapping, leveraging Jetson Nano boards, webcams, and Python (Version 3.9). We employed a classification technique to categorize patients into two groups: those with a positive mood and those with a negative mood. For comparing accuracies, we utilized three types of models: multilayer perceptron (MLP), support vector machine (SVM), and random forest. Model validation and evaluation were conducted using Python programming. The results of this study fall into three parts. The first part involved testing the monitoring and surveillance system. It was found that the system could receive information from the wearable device, display the received data in graph form, and notify the medical staff when examining symptoms to consider whether the patient should be taken to the hospital. The second part focused on testing the device, and it was found that it could measure body temperature, heart rate, and blood oxygen levels (SpO₂) and send those data to the database. The third part involved an AI camera test, and it was found that the most suitable algorithm to analyze the patient’s facial expressions was Random Forest. The results show that the system supports hospitals in managing COVID-19 and similar diseases by enabling timely interventions through facial expression analysis.

The study aimed to develop a recommender system for dengue prevention using environmental factors and mosquito larvae data. Data were collected from 100 households in Surat Thani, Thailand using mosquito larval survey in January 2020. Data mining techniques: frequent pattern growth (FP-Growth) and Apriori algorithms were used to find association rules and to compare accuracies for selecting a suitable model. The recommender system was designed as a web application. FP-Growth is more suitable for these data than Apriori algorithm. The factors associated with dengue infection, including community area, densely populated area, and agricultural area. Most areas where mosquito larvae are found are community areas and agricultural areas. Aedes larvae were found most in water containers with dark colors and without a lid. Aedes larvae were also found in small water jars, large water jars, cement tanks, and plastic tanks. The recommender system should be useful to dengue vector prevention and to health service communities, in planning and operational activities.