Journal of Computer and Creative Technology

A Principal Component Analysis of Key Factors Influencing Enrollment Decisions in a Bachelor of Business Administration Program

2026-07-12T20:48:10+07:00

This study aims to identify and extract the key factors influencing students’ enrollment decisions in higher education by adopting a data-driven analytical approach. Principal Component Analysis (PCA) was employed within the CRISP-DM framework to systematically analyze quantitative data collected from 571 prospective students across three provinces in Thailand. The survey instrument comprised 45 observed variables. Data suitability tests indicated strong adequacy for factor analysis, with a Kaiser–Meyer–Olkin (KMO) value of 0.963 and a statistically significant Bartlett’s Test of Sphericity. The PCA results, using Varimax rotation, revealed five latent components shaping enrollment decisions: financial readiness, institutional support and safety, student identity formation, technological and curriculum relevance, and peer guidance. The findings demonstrate that enrollment decision-making is multidimensional and hierarchical, extending beyond academic reputation to include economic, psychosocial, and experiential considerations. Academically, this study contributes by integrating PCA with educational data mining to provide a structured interpretation of student choice behavior. Practically, the results offer evidence-based insights to support curriculum development, targeted communication strategies, and enrollment policy design in higher education institutions.

Development of a Mobile Application System for Promoting Tourism in Phra Nakhon Si Ayutthaya Province

2026-07-17T12:41:23+07:00

This research aimed to: 1) develop a mobile application system to promote tourism in Phra Nakhon Si Ayutthaya Province, and 2) assess user satisfaction with the developed system. The participants consisted of 250 individuals, including Thai tourists, international tourists, local entrepreneurs, and officials from relevant agencies, selected using purposive sampling. The research instruments included a mobile application system developed for the Android operating system using Visual Studio Code, a system performance evaluation form, and a user satisfaction questionnaire. The research findings revealed that the overall performance of the mobile application system developed to promote tourism in Phra Nakhon Si Ayutthaya Province was at a high level (M = 4.37, SD = 0.58). In addition, overall user satisfaction with the system was also at a high level (M = 4.29, SD = 0.61). These findings indicate that the developed mobile application effectively enhances access to tourism information, improves usability, and enriches the overall user experience for promoting tourism in Phra Nakhon Si Ayutthaya Province.

Deep Learning to Enhance the Accuracy and Lag-time of Time Series Data Forecasting

2026-08-24T08:32:38+07:00

Time series forecasting is crucial in the present, applied in areas such as weather forecasting, finance, and beyond. Gold price forecasting is highly complex due to significant influences from economic factors. Although current studies have explored gold price forecasting using deep learning models, issues of prediction lag persist, affecting the accuracy of directional forecasts. To address this problem, this research focuses on developing a gold price time series forecasting model with improved accuracy and reduced prediction lag. This is achieved by combining the strengths of the LSTM (Long Short-Term Memory) model, which excels at learning long-term dependencies in data, and the CNN (Convolutional Neural Network) model, which efficiently extracts spatial features from data. Additionally, a data grouping method based on price trend characteristics is employed, allowing the model to learn specific patterns of each trend type appropriately. This reduces lag caused by delayed responses to actual data changes and enhances the accuracy of gold price forecasts in both numerical value and directional movement. Consequently, the model demonstrates strong potential for effective application in real-world financial trading decisions. This study uses gold price data from January 2014 to July 2025, totaling 2,943 days. Price trends are classified into three groups uptrend, sideways and downtrend. Four models are trained, which are LSTM, CNN, LSTM_G3, and CNN_G3. The most suitable model for each group is then selected to create the LSTM+CNN_G3 model. The results show that the LSTM+CNN_G3 model achieves the lowest overall MAE (Mean Absolute Error) at 17.68 US dollars and the highest overall MPM (Movement Prediction Metric) at 54.08%, outperforming the baseline LSTM and CNN models that do not use data grouping. This demonstrates that the grouping method effectively reduces prediction lag. In a gold trading simulation, the LSTM+CNN_G3 model yields the highest profit of 1,870 US dollars, representing a 101.87% return on capital. These findings indicate that grouping data by price trends combined with selecting appropriate architectures can improve accuracy and reduce lag, enabling the model to be effectively applied in real-world financial markets.

Design and Implementation of a Low-Code Geofencing Architecture for Location-Based Attendance Management in Local Government Organizations

2026-07-21T14:36:46+07:00

As public organizations undergo digital transformation, efficient workforce mobility management is imperative. This study addresses a critical gap in the integration of spatial tracking capabilities within low-code platforms by designing and implementing a low-code geofencing architecture at Thasao Subdistrict Municipality, Thailand. The proposed three-tier architecture facilitates real-time data processing using Google AppSheet and Google Sheets. The system's business logic implements the Haversine formula for geodesic distance calculation and incorporates error-handling mechanisms to mitigate intermittent GPS signal instability. Performance evaluation via black-box testing demonstrated a 100% success rate across predefined functional test cases under controlled conditions, rather than absolute system accuracy. Field observations further revealed that real-world performance may vary depending on network latency and GPS signal quality. Empirical data collected through purposive sampling from 20 off-site personnel indicated a high level of satisfaction (Mean = 4.39, S.D. = 0.57), with the evaluation instrument reaching high reliability (Cronbach's Alpha = 0.85). This research contributes a scalable and resource-efficient architectural framework that reduces administrative workload and supports practical digital transformation in local government contexts, while also highlighting implementation constraints relevant to real-world deployment.

Development of an Intelligent Air Quality Control System for Shoe Cabinets to Prevent Odor Accumulation

2026-07-26T09:19:19+07:00

This research aimed to develop an intelligent shoe care cabinet capable of reducing factors that contribute to unpleasant odors from footwear. The system focuses on controlling environmental conditions related to accumulated humidity and microbial growth, which are primary causes of odor, to promote good hygiene and reduce the burden of shoe maintenance for users. The system integrates a heater, UV-C (253.7 nm), a disinfectant misting system, and a ventilation fan, operating in conjunction with Internet of Things (IoT) technology via an Arduino UNO R4 Wi-Fi board and the ThingsBoard Cloud platform. From system testing, the system reduced relative humidity from 43%RH to 30%RH within 30 minutes of heater operation, and TVOC levels decreased from 184 ppb to 25 ppb within 5 minutes of ventilation. UV-C irradiation at 253.7 nm reduced bacterial counts from 6.80 × 10³ to below 30 CFU/ml and fungal counts from 1.04 × 10⁴ to below 30 CFU/ml within 10 minutes. Overall user satisfaction was rated at the highest level ( = 4.65, S.D. = 0.50).

Comparative Analysis of Pull and Push Monitoring Models for Information Services in CGNAT Environments

2026-08-22T10:16:36+07:00

This research investigates the development of a monitoring system for information environments operating behind Carrier-Grade NAT (CGNAT), a significant constraint that prevents direct access from a central server to client nodes. The study conducts an in-depth comparative analysis of two architectures: (1) a Pull Model operating over an overlay network utilizing Tailscale technology, and (2) a traditional Push Model transmitting data via a Pushgateway. Experiments were performed on a live system deployed on Microsoft Azure, with network conditions simulated using pfSense. The results reveal significant trade-offs between the two approaches. The Pull Model via Tailscale demonstrates superior performance regarding alerting latency, real-time data continuity, and enhanced security, as it eliminates the need to expose public ports. Conversely, the Push Model is distinguished by lower system overhead on client nodes and greater configuration autonomy. These findings provide guidelines for architectural selection based on specific use cases. The Pull Model is recommended for scenarios prioritizing high security where specialized configuration expertise is limited; Tailscale’s underlying Zero Trust Network Access (ZTNA) simplifies deployment through single-account authentication without complex setup. Alternatively, the Push Model is ideal for resource-constrained devices, such as IoT, provided that expert personnel are available to ensure secure configuration.

Automated Detection and License Plate Recognition of Motorcycles on Sidewalks Using Deep Learning

2026-08-24T11:50:37+07:00

Motorcycle riding on sidewalks is a common traffic violation in urban areas and poses significant risks to pedestrian safety. However, existing approaches remain limited in processing surveillance images under real-world conditions, particularly those captured from top-view CCTV cameras with constrained resolution. In addition, motorcycle license plates are significantly smaller than automobile plates, resulting in fewer pixels available for character recognition and increasing the difficulty of optical character recognition (OCR). This study proposes an automated system for detecting motorcycles on sidewalks and recognizing license plates from CCTV images. The system employs deep learning techniques for object detection and OCR for extracting textual information from license plates, with a focus on real-world deployment using top-view camera perspectives. The proposed system consists of three main stages: motorcycle detection, license plate detection, and character recognition. A YOLO-based model is trained on annotated datasets for detection tasks, and an image enhancement process is applied prior to OCR to improve recognition performance. Experimental results show that the system achieves average confidence scores of 0.80 for motorcycle detection and 0.84 for license plate detection. In addition, OCR accuracy improves from 32% to 46% after image enhancement. The results demonstrate that the proposed system can effectively operate in real-world environments for detecting illegal motorcycle behavior on sidewalks and has the potential to be further developed into an automated traffic law enforcement support system.

Enhancing Silkworm Feeding Efficiency at Each Larval Stage Using Chatbot Technology to Improve Silk Production Capacity in Surin Province

2026-08-24T08:12:15+07:00

In Surin Province, silkworm farming runs on memory. Feeding recommendations vary by larval stage, yet most rely on memory and informal knowledge passed between generations. This study developed and evaluated MaiKham AI, a domain-specific chatbot built on a Retrieval-Augmented Generation (RAG) architecture, designed to deliver stage-specific feeding guidance through text, voice, and image. The study had two objectives: assessing technical performance of the chatbot, and measuring its effect on sericulture productivity and technology acceptance. A quasi-experimental design was used. Fifty silkworm farmers in Surin Province were purposively selected 25 used the chatbot, 25 continued conventional practice. The RAG system drew on a 52-week silkworm rearing manual loaded into Google AI Studio’s Knowledge Base, with Gemini 1.5 Pro generating responses grounded in that domain knowledge. Expert panels rated the system 4.58 out of 5.00 (S.D. = 0.24). Chatbot users harvested heavier cocoons and higher-grade silk than the control group (p < 0.001), with yield gains of 70–90% across rearing cycles. On the Technology Acceptance Model, farmers scored Perceived Usefulness at 4.38 (S.D. = 0.78) and Perceived Ease of Use at 4.46 (S.D. = 0.87); elderly farmers who had initially resisted the phone adapted more quickly once they could speak their questions aloud. After a single structured training session, 94% used the system without assistance. RAG based chatbots appear capable of filling the knowledge gap in smallholder sericulture, even where farmers are older and digitally inexperienced.

Applying AI to Control Air Conditioning System Based on Occupancy

2026-08-30T22:07:48+07:00

This research aimed to develop and evaluate a prototype of an occupancy-driven air conditioning control system. The system integrated human detection technology using the YOLOv8n algorithm with a Raspberry Pi processing unit an edge computing solution that is cost-effective, low-power, and compact—for real-time occupancy monitoring and counting for controlling air conditioning units via infrared signals. The system's performance was evaluated through 791 recorded events, focusing on detection accuracy under real-world indoor environmental lighting variations varying environmental conditions and control responsiveness. The results indicated high robustness to lighting variations, achieving an overall average accuracy of 92.21% with a Mean Absolute Error (MAE) of 0.44 people. The primary cause of detection error was attributed to person-to-person occlusion during high occupancy periods rather than lighting conditions. Furthermore, the system effectively adjusted air conditioning operations based on real-time occupancy, significantly reducing unnecessary energy consumption when spaces were unoccupied. These findings demonstrate the potential of the developed system for integration into smart buildings and sustainable energy management applications.

The Development Model for Promoting Media Literacy Skills for Aging Society through Family Participation

2026-08-24T18:41:32+07:00

This research aimed to develop a model for promoting media literacy skills among the elderly through family participation in Sang Tho Subdistrict, Khueang Nai District, Ubon Ratchathani Province. The study was conducted in three phases: (1) studying the current conditions, problems, and needs; (2) developing a model and operational manual; and (3) examining the effects of implementing the model through participatory learning processes. The research instruments included an interview form, a test, a skill assessment form, a learning record form, and a satisfaction questionnaire. Data were analyzed using content analysis and basic statistics, namely mean, standard deviation, and paired t-test. The results revealed that (1) the elderly tended to use online media for extended periods and lacked skills in verifying information, indicating a strong need for developing media literacy skills with family support; (2) the developed model comprised six components and a seven-step learning process (SONGSOT Model), which experts evaluated as highly appropriate; and (3) after participating in the program, participants’ post-test scores were significantly higher than their pre-test scores at the .05 level. Overall, the achievement of media literacy skills reached 79.46%, and all four skill dimensions met the evaluation criteria of 70% or higher. Participants reported a high level of satisfaction, with the highest ratings in support systems and learning activities that promoted deep, accurate, insightful, and critical understanding. The findings demonstrate that the developed model effectively enhanced media literacy skills among the elderly through active family participation.

Time-Lagged Backpropagation Neural Network-Based Machine Learning for Loop Location Identification in Layer 2 Switch Network

2026-08-28T09:40:47+07:00

This research aimed to study and analyze loop problems in Layer 2 switching networks and their impacts on system performance, to develop a machine learning model using a Time-Lagged Backpropagation Neural Network to identify the locations of network loop problems, and to evaluate and compare the performance of the developed model based on defined metrics. The research methodology consisted of eight main steps: 1) network simulation, 2) loop problem simulation, 3) traffic data collection, 4) data preparation and cleaning, 5) training and testing data splitting, 6) model development using various algorithms, 7) parameter tuning, and 8) model performance evaluation using a confusion matrix in conjunction with statistical metrics. The results indicated all five developed models successfully identified the locations of loop occurrences. Among these, the Time-Lagged Backpropagation Neural Network delivered the best performance, achieving an Accuracy of 91.46%, Precision of 95.27%, Recall of 91.46%, an F1-score of 93.11%, and an Area Under the Curve (AUC) of 0.986. These findings demonstrated that the proposed model possessed high efficiency in accurately detecting and locating network loop problems.

Application of Simplified QFD for Selecting 3D-CAD FOSS in Mechanical Engineering Education: A Case Study in Private Educational Institution

2026-08-28T09:40:45+07:00

The aim of this research was to present a selection method for “Free and Open-Source Software (FOSS)” type of three-dimensional computer-aided design (3D-CAD) by using the Simplified Quality Function Deployment (Simplified QFD). The research was conducted into 4 tasks: searching for the alternative 3D-CAD FOSS which have been popular used among makerspace communities, extracting appropriate criteria used for selecting 3D-CAD FOSS from research literature and makerspace communities, brainstorming to select the most suitable software by using QFD matrix among 11 stakeholders from top executive down to student representative of Rajapark Institute, and validating the selected software by real teaching experiment. Teaching experiment was performed on a sample of 8 undergraduate students who enrolled in the elective subject of “Computer for Mechanical System Design” from academic year of 2025. With three alternative 3D-CAD FOSS (namely FreeCAD, Blender and Autodesk 123D Design) that were brainstormed and selected by using the simplified QFD, FreeCAD is the most suitable software with a maximum score of 31. Results from student questionnaires after teaching experiment pointed out that students were more positive on FreeCAD than Autodesk 123D Design and Blender respectively.

Design and Development of a 3D Cultural Exhibition Model for the Creative Cultural Park at Rajamangala University of Technology Isan

2026-08-30T16:46:35+07:00

This study aimed to (1) analyze museum and exhibition design principles suitable for the local cultural context of the cultural exhibition space at the Creative Cultural Park, Rajamangala University of Technology Isan, (2) develop a 3D cultural exhibition model to communicate local cultural identity, and (3) evaluate the suitability of the developed 3D model and synthesize guidelines for applying AR/VR technologies to enhance the exhibition. The study employed a Research and Development (R&D) design using an exploratory sequential mixed-methods approach Phase 1 comprised document analysis, site analysis, and semi-structured interviews with five experts. Phase 2 developed and refined the prototype using SketchUp and Blender. Phase 3 evaluated the model with 370 faculty members, staff, and students selected through convenience sampling, together with expert evaluation (n = 5) and behavior mapping. Data were analyzed using content analysis and descriptive statistics, including Mean, Standard Deviation, frequency, and percentage. The findings showed that the integrated use of contextual learning, constructivist learning, participatory design, and user-centered design provided an appropriate basis for the cultural exhibition design. User evaluation indicated a very high level of overall suitability ( = 4.83), with cultural interpretation and local identity receiving the highest mean score ( = 4.94). Expert evaluation also indicated a very high level of overall suitability ( = 4.83), while the technology dimension received a comparatively lower but still high rating ( = 4.00). Behavior mapping further showed that 95.7% of participants discussed exhibition objects with others, reflecting active social interaction. The study contributes an integrated Input–Process–Output framework that connects local cultural context, learning theories, participatory and user-centered design, 3D exhibition development, and AR/VR application guidelines for educational and community cultural spaces.

Development of a Learning Management Model Based on Engineering Design Processes Integrated with Artificial Intelligence Technology to Enhance Creative Thinking and Problem-Solving Abilities of Grade 8 Students

2026-08-29T14:59:46+07:00

This research aimed to (1) investigate guidelines for developing a learning management model based on the engineering design process integrated with artificial intelligence technology, (2) develop the learning management model, and (3) examine the effects of implementing the model. The sample consisted of 30 Grade 8 students, selected through purposive sampling. The research instruments included an expert interview form, lesson plans, a learning management model, learning support tools, a creativity assessment form, a problem-solving ability assessment form, a learning achievement test, and a satisfaction questionnaire. Data were analyzed using mean, standard deviation, reliability coefficients, and t-tests. The findings indicated that (1) the development of the learning management model should integrate the engineering design process with artificial intelligence to support students’ thinking, analysis, design, and product development through project-based learning and authentic problem-solving situations with ethical considerations. (2) The developed SUTLUM Model comprised six stages: Survey, Understand, Think, Launch, Upgrade, and Manifesto, along with four key components: teachers, learners, AI tools, and learning modules. The model demonstrated a very high level of Quality (M = 4.65, S.D. = 0.47). (3) The implementation results showed that students’ creativity was at a very high level (M = 31.90 out of 36), and their problem-solving ability was also at a very high level (M = 21.60 out of 24). In addition, students’ post-test achievement scores were significantly higher than their pre-test scores at the .05 level, and their satisfaction with the learning management was at the highest level (M = 4.68, S.D. = 0.46).

Development of an Integrated Data Decision Support System for Area-Based Targeting and Assistance Tracking of Poor Households: A Case Study of Nakhon Ratchasima Province

2026-08-29T17:14:28+07:00

This study set out to (1) build an integrated-data Decision Support System that targets, screens, and tracks assistance for poor households; (2) examine what happened once it was deployed area-wide; and (3) gauge user satisfaction alongside the system quality and suitability judged by experts. Development followed a system-development approach within Participatory Action Research and the System Development Life Cycle. The system linked three databases, TPMAP, ThaiQM, and PPP-Connext, through an API tied to the Government Data Catalog and the provincial City Data Platform, then added a standardized two-track verification mechanism and a three-level monitoring and feedback process (tambon, district, province). Two instruments were used: a five-point satisfaction questionnaire for 56 trained officer-users, and an expert-assessment form for five information-technology experts. Five experts confirmed content validity (IOC 0.80–1.00), and internal-consistency reliability was Cronbach’s = 0.89 and 0.93. Once deployed across 657 agencies covering 21,521 households, the redesigned process replaced scattered offline records with a single real-time platform. Assistance reached 4,613 households, and every identified hardest-hit (“Living-with-Difficulty”) household was referred into state welfare. Users rated overall satisfaction high ( = 4.34); experts rated overall quality and suitability high ( = 4.44), though processing speed scored lowest ( = 4.00). Two caveats matter. The study did not measure changes in household income or poverty status, and the values describe the instruments, not how the system performs. Even so, the evidence suggests that pulling fragmented poverty data into one DSS can sharpen targeting, shorten the time needed to reach information, and make assistance more transparent, giving other provinces a model they can adapt where data readiness is comparable.

Using Generative Artificial Intelligence with Reflective Activities to Enhance Critical Thinking among Undergraduate Students

2026-08-29T17:14:26+07:00

This research aimed to (1) to compare students' critical thinking before and after the intervention within each group; (2) to compare critical thinking between the experimental group and the control group after the intervention; (3) to examine the effectiveness of the learning activities; and (4) to examine students' satisfaction with the learning activities in the experimental group. The sample comprised 42 undergraduate students obtained through two-stage cluster random sampling. In the first stage, eight academic programs each offering two parallel class sections were identified; in the second stage, one program was randomly selected, and its two intact sections were randomly assigned as the experimental group (n = 22) and the control group (n = 20). The research instruments consisted of lesson plans, a critical thinking test, and a reflective thinking assessment form. Data were analyzed using mean, standard deviation, t-test, and the effectiveness index (E.I.). The findings revealed that (1) the posttest critical thinking scores of the experimental group ( = 26.32, SD = 1.36) were significantly higher than their pretest scores ( = 12.59, SD = 1.44) at the .05 level (t = 31.04, df = 21, p < .05); (2) the posttest critical thinking scores of the experimental group ( = 26.32, SD = 1.36) were significantly higher than those of the control group taught through conventional instruction ( = 17.65, SD = 2.28) at the .05 level (t = 15.14, df = 40, p < .05); (3) the learning activities implemented with the experimental group yielded an effectiveness index of 0.5008, higher than that of the control group (0.1722), indicating that GenAI combined with reflective thinking activities effectively promotes students' critical thinking; and (4) students in the experimental group reported overall satisfaction with the learning activities at the highest level ( = 4.60, SD = 0.62). These results demonstrate that applying Generative AI together with reflective thinking activities, as a tool supporting reflection, can enhance the quality of learning and constitutes an appropriate approach to instructional design in higher education in the digital era.