JOURNAL OF INFORMATION SCIENCE AND TECHNOLOGY

The Classification of Children with Autism using Deep Learning-Based Image Analysis

2025-07-18T16:16:15+07:00

Currently, the diagnosis of autism in children still relies on behavioral observation and psychological tests, which may have limitations in terms of accuracy and speed. This research therefore applies machine learning techniques, especially convolutional neural networks (CNN) with the ResNet50 model, to analyze and classify drawings of children with and without autism. The samples were divided into groups of 5–8 and 9–12 years old, and the results were evaluated with Accuracy, Recall, Specificity, F1-Score, and Confusion Matrix. The experimental results showed that the model could classify drawings accurately, with an Accuracy of 81.9% and 89.5%, and an F1-Score of 0.83 and 0.91 for the 5–8 and 9–12 age groups, respectively. After using Data Augmentation, the accuracy increased to 87.7% and 91.1%, with an F1-Score of 0.89 and 0.93. However, the research still has limitations in terms of the size and variety of data, which should be expanded in the future to increase the accuracy of the model.

Apply Monte Carlo Simulation to Synthesize Data

2025-07-15T16:19:52+07:00

This research presents the data synthesization by using monte carlo simulation. Six datasets were synthesized and categorized into two types: (1) datasets with more categorical variables than numerical variables, and (2) datasets with more numerical variables than categorical variables. Synthesize data 1500 rows for each dataset then compared between real data and data synthesization using 1) The Kolmogorov-Smirnov Two-Sample Test, 2) T-test, 3) Cosine Similarity Test, 4) Multiple Linear Regression Analysis, and 5) Direct Data Comparison. The results showed that the Monte Carlo method was the most efficient for synthesizing data, especially for categorical variable data. Based on the coefficients of determination, the Monte Carlo simulation was 60.47% more efficient than Generative Adversarial Networks (GANs) and 52.41% more efficient than Variational Autoencoders (VAEs). Additionally, the Monte Carlo simulation method allows for adjustments to better represent the population in cases where the sample group does not fully cover it.

Comparative Study of ML and DL in Optical Transceiver Failure Diagnosis

2025-11-07T15:52:18+07:00

High-speed optical transceivers require robust failure analysis methods to ensure production reliability in modern communication systems. This study systematically evaluates machine learning algorithms (Random Forest, XGBoost) and deep learning approaches (Fully Connected Neural Networks) for optical transceiver failure analysis across two operational scenarios using real manufacturing data from 6,446 units. In a comprehensive data analysis (Scenario #1), both Random Forest and XGBoost achieved exceptional performance (MSE: 0.0000, MAE: 0.0001), while FCNN demonstrated comparable results (Loss: 0.0002, MAE: 0.0002). In a focused analysis of failed units (Scenario #2), XGBoost outperformed other models with the lowest error metrics (MSE: 0.0091, MAE: 0.0165) compared to Random Forest (MSE: 0.0125, MAE: 0.0399) and FCNN (Loss: 0.1571, MAE: 0.2987). SHAP analysis consistently identifies influential features across both scenarios, providing actionable insights for quality control optimization. These findings establish a quantitative framework for selecting optimal AI approaches for optical transceiver failure diagnostics, with machine learning models recommended for datasets under 10,000 samples and deep learning for larger datasets. The proposed methodology advances AI-driven failure diagnostics in optical transceiver manufacturing.

The Gamified Voice AI Agents: Enhancing User Engagement and Interaction Through Playful Design

2025-12-08T19:17:47+07:00

This research examines the design of a Voice AI agent enhanced with gamification elements such as points, rewards, challenges, storytelling, and personalization to increase user engagement and satisfaction. Two hypotheses guided the study: (H1) users of gamified Voice AI systems demonstrate higher engagement and motivation than those using non-gamified systems; (H2) demographic factors, particularly age and prior experience, influence preferences for gamification features. Experimental testing and user feedback showed that the most frequent interactions involved game-based activities, childlike conversations, and psychology-related queries, with consistently high satisfaction and predominantly positive or curious sentiments. These findings support both hypotheses, highlighting the motivational benefits of gamification and the moderating role of user characteristics. Future work should investigate how gamification shapes long-term engagement, how demographics moderate playful feature effectiveness across contexts, and which combinations of rewards, personalization, and social presence deliver sustainable improvements in satisfaction and retention.

Development of an Interactive Multimedia Platform Using Cross-Device Virtual Reality

2025-12-15T17:06:51+07:00

This research aims to examine the digital ecosystem supporting the design and development of an interactive multimedia platform using virtual reality (VR) technology, and to evaluate the performance of the cross-device VR-based multimedia platform. Experimental results indicate that the proposed digital ecosystem efficiently supports both VR Desktop and VR Meta Quest 2 environments. Performance testing shows that the VR Desktop configuration achieves an average frame rate of 107.83 FPS, with response time and latency values below 50 ms, whereas the Meta Quest 2 achieves an average frame rate of 86.63 FPS and exhibits slightly higher latency relative to the VR Desktop. Expert evaluations showed that the VR Desktop received the highest score (𝑥̅ = 4.46), reflecting its stability and display quality. Furthermore, the satisfaction assessment involving 60 participants indicated that interaction between the VR Desktop and Meta Quest 2 (360° mode) yielded a mean score of 4.33 (S.D. = 0.71), while interaction between the Meta Quest 2 and Meta Quest 2 (360° mode) yielded a mean score of 4.50 (S.D. = 0.50). In conclusion, the integration of these systems substantially enhanced the effectiveness and performance of the virtual learning ecosystem.

Evaluating Interface Consistency and User Control in Fog and Edge Computing Dashboards: A Heuristic Analysis of Open-Source Platforms

2025-12-02T14:31:46+07:00

Fog and edge computing platforms increasingly become necessary for the management of distributed Internet of Things services, but their management dashboards are not typically very friendly to users. This study discusses a heuristic analysis of two most widely used open-source management dashboards, EdgeX Foundry and FogFlow, in terms of five usability guidelines: consistency and standards, user control and freedom, visibility of system status, error prevention, and help and documentation. Nineteen usability issues were identified and rated according to Nielsen's heuristic model through task-based, expert-led walkthroughs. Major issues included non-consistent terminology, a lack of undo or confirm features for critical actions, inadequate real-time feedback, and a lack of embedded help facilities. These issues reduce operation effectiveness and increase the likelihood of user error, most notably in time-sensitive or high-risk scenarios. The study proposes certain recommendations such as standardizing the components of the interface, improving the options provided to the users for control, improving system feedback, and incorporating contextual help. The incorporation of human-centered design principles during the creation of fog and edge dashboards will significantly enhance usability, enable better decision-making, and enable easier use in real-world implementations.

The Guideline For Rapid Software Game Development On Unity Web-based Ecosystem

2025-12-03T23:48:04+07:00

This study proposes a guideline for rapid content-based game development within the Unity web ecosystem, focusing on processes that clearly separate team roles to improve flexibility, usability, and collaboration efficiency. By comparing traditional CMS-based approaches with a web-driven process, the findings show that the proposed method reduces overall development time by multiple of times, enables independent task execution, and ensures stronger protection of the game’s core source code. The proposed process thus provides a practical and scalable solution for modern game production, where speed, security, and adaptability are critical.

The Development of LPG Gas Detection Applications and Automatic Valve Control With Internet of Things

2025-06-02T13:05:52+07:00

This paper presents the development and implementation of an IoT-enabled gas leak detection and alert system designed to improve safety in environments where LPG and similar gases are used. This system monitors gas concentration in real-time using the MQ-5 gas sensor integrated with the ESP8266 NodeMCU V2 microcontroller. When gas levels exceed safe thresholds, a multi-level alert system is triggered, which includes notifications via LINE Notify and an automatic shut-off mechanism to close the gas valve using a Servo Motor. The system operates autonomously, providing rapid responses to mitigate risks associated with gas leaks. Our results demonstrate the effectiveness of this system in reducing the risks of health hazards and fire outbreaks due to gas leaks.

Factors Affecting the Adoption of AI in Organizations: A Theoretical Study

2025-08-05T22:26:27+07:00

This research aims to study the factors affecting the adoption of Artificial Intelligence (AI) in organizations using theoretical analysis and literature review to examine key factors. The findings reveal that there are four main factors influencing the adoption of AI: 1) Technology and Innovation Readiness, where organizations with strong infrastructure and continuous innovation are better able to implement AI; 2) Employee Training and Development, as providing employees with necessary training and skill development enhances effective AI implementation; 3) Change Management, where planning and executive support facilitate smooth adaptation to AI integration; and 4) Continuous Evaluation and Improvement, which ensures effective use of AI systems while emphasizing security measures. These findings can guide organizations in planning and implementing AI efficiently to enhance competitive advantage and foster new innovations.

Association Rule Mining to Identify Clinical Factors Linked to Disease Severity in Liver Cirrhosis

2025-11-30T10:55:34+07:00

This study applies the association rule mining to identify critical clinical patterns linked to different stages of liver cirrhosis. Utilizing data from the Mayo Clinic primary biliary cirrhosis trial (1974-1984), the research analyzed 5,805 patient records after applying inclusion and exclusion criteria. The Apriori algorithm was used to extract association rules, with measures including support, confidence, and lift. The analysis revealed distinct factors associated with each disease stage. For instance, Stage 1 was strongly linked with normal laboratory values like SGOT and Bilirubin. Stage 3 was predominantly associated with abnormal clinical markers, including high prothrombin, low platelets, and age over 60. A dendrogram further clustered these factors, visually reinforcing the associations. The discovered rules provide valuable insights into the combinations of clinical and laboratory features that characterize cirrhosis progression. These findings can aid in early risk stratification and inform clinical decision-making for targeted patient management.