Rice production plays a vital role in the economies of Southeast Asia; however, the increasing adoption of automated and digitally integrated rice milling systems has posed significant challenges for workforce training, particularly for inexperienced operators. To address this gap, this study proposes an Interactive Virtual Reality Learning Platform (IVRLP) designed to support immersive, experiential training for rice milling operations. The study adopts a mixed-methods approach, integrating quantitative pre- and post-training assessments with qualitative user feedback to evaluate the platform’s effectiveness. A total of 31 participants from rice milling cooperatives in Vientiane, Lao PDR, engaged with the IVRLP, which simulates key operational processes such as cleaning, separation, and polishing. Statistical analysis using a paired-samples t-test revealed significant improvements in learning outcomes. Specifically, process familiarity increased from M = 3.12 (SD = 0.84) to M = 4.08 (SD = 0.63), with a paired-samples t-test result of t = 6.85, df = 30, p < 0.001, while operational confidence improved from M = 3.05 (SD = 0.91) to M = 4.15 (SD = 0.58), t = 7.42, df = 30, p < 0.001. In addition, descriptive findings indicate high levels of user satisfaction, with 80.9% of participants reporting enhanced understanding and 78.5% expressing willingness to recommend the system.
Multicultural learning environments are increasingly prevalent in higher education; however, empirical understanding of how cultural context shapes knowledge conversion during collaborative, hands-on learning remains limited. This study examines how knowledge conversion unfolds within LEGO® SERIOUS PLAY® (LSP) activities by integrating Nonaka and Takeuchi's SECI (Socialization, Externalization, Combination, Internalization) model with constructionist learning principles and multimodal interaction theory. A qualitative comparative case-study design was employed. Two graduate cohorts – Thai learners ($n=9$, Thai-medium instruction) and International learners from East and Southeast Asian backgrounds $(n=10$, English-medium instruction) – participated in an identical LSP-based Smart City design workshop under controlled conditions. Data were collected from group presentation videos, individual reflection recordings, and instructor observations, and analysed using a hybrid deductive-inductive coding framework with frequency counts of Smart City-related coded episodes as a structured basis for cross-cohort comparison. The findings suggest distinct, culturally patterned knowledge-conversion trajectories. Thai learners appeared to demonstrate tacit synchrony, metaphor-first externalization, holistic combination, and experiential–emotional internalization. International learners appeared to exhibit explicit verbal coordination, verbal-first externalization, analytical combination, and cognitively oriented internalization. These patterns suggest that SECI processes may not operate uniformly across cultural contexts but appear to be mediated by communication styles, cognitive orientations, and culturally situated meaning-making practices. Based on these findings, the study proposes the Multicultural SECI Flow Model, which explains how constructionist artefacts may function as intercultural mediating objects that support convergence toward shared understanding across divergent cultural learning pathways. The study contributes to Knowledge Management scholarship and provides practical implications, including facilitation guidelines and a practitioner checklist, for culturally responsive LSP-based learning in higher education, corporate training, and digital collaboration contexts.
Crafty tactics like nested request bodies, encoding schemes, and JavaScript Object Notation (JSON) operators are now used by attackers to trick and bypass conventional Web application firewalls. The proposed machine learning-based system for detecting and mitigating SQL injection attacks is designed not just to protect against conventional SQLi attacks but also against JSON-based SQLi attacks, NoSQL injection attacks, hybrid attacks, and conventional WAF evasion techniques. The proposed system utilizes a stacking ensemble of Random Forest, Gradient Boosting, and Logistic Regression classifiers with manually constructed features that represent various properties of queries instead of using conventional static rule-based techniques or deep learning models. The detector is integrated into an application process that facilitates query inspection, batch analysis, decision explanation, and mitigation actions. The application is made available via a Flask-based REST API. To add structural variety, the dataset is constructed from public payload sources and augmented methodically. To support detections and to provide potential WAF rules, feature-level explanations are employed. The proposed work is also extended to incorporate a privacy-preserving federative learning framework to show its efficacy in collaborative environments. The system's overall goal is to provide a modern, API-driven application as a complementary injection attack detection and monitoring layer suitable for quick, easy deployment.
The COVID-19 pandemic claimed numerous lives, and led to the development of new tools for disease treatment. Initial diagnosis using X-rays can now identify COVID-19-infected individuals, and significant research has promoted neural networks for X-ray image classification. Quantum technology is also gaining interest, with intriguing properties such as superposition and entanglement. This study proposed a Quantum Neural Network (QNN) for X-ray image classification to investigate the relationship techniques between layer expansion and quantum circuits, measured by expressibility and entanglement capability. The results showed that both metrics significantly affected model accuracy. Seven circuits were tested, each with six layers, to examine their impact on model performance. The experiments yielded an accuracy of 95%, with highly effective image classification circuits exhibiting a balanced relationship between entanglement capability and expressibility.
Rice production plays a vital role in the economies of Southeast Asia; however, the increasing adoption of automated and digitally integrated rice milling systems has posed significant challenges for workforce training, particularly for inexperienced operators. To address this gap, this study proposes an Interactive Virtual Reality Learning Platform (IVRLP) designed to support immersive, experiential training for rice milling operations. The study adopts a mixed-methods approach, integrating quantitative pre- and post-training assessments with qualitative user feedback to evaluate the platform's effectiveness. A total of 31 participants from rice milling cooperatives in Vientiane, Lao PDR, engaged with the IVRLP, which simulates key operational processes such as cleaning, separation, and polishing. Statistical analysis using a paired-samples t-test revealed significant improvements in learning outcomes. Specifically, process familiarity increased from M = 3.12 (SD = 0.84) to M = 4.08 (SD = 0.63), with a paired-samples t-test result of t = 6.85, df = 30, p < 0.001, while operational confidence improved from M = 3.05 (SD = 0.91) to M = 4.15 (SD = 0.58), t = 7.42, df = 30, p < 0.001. In addition, descriptive findings indicate high levels of user satisfaction, with 80.9% of participants reporting enhanced understanding and 78.5% expressing willingness to recommend the system.
Smartphones have become necessary in everyday life since they make communication, financial transactions, and data access easier. However, their broad use poses serious security risks, especially regarding ongoing user authentication. Traditional authentication techniques, including PINs, passwords, and patterns, only authenticate users at points of entry, leaving devices open to replay attacks, session hijacking, and spoofing. To overcome these constraints, the hybrid authentication approach proposed in this research uses multimodal touch behaviour for real-time identity verification. Using the Touchalytics dataset, this method combines motion sensor data from accelerometers, gyroscopes, and magnetometers with fine-grained touch attributes, including touch area, pressure, finger orientation, and typing dynamics. Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks are combined in the system's deep learning (DL) architecture for sequential touch analysis, and optimization approaches are used to improve model performance. The model captures detailed touch behaviour and motion sensor data, with hyperparameter tuning applied using Particle Swarm Optimization (PSO), Cuckoo Search Optimization (CSO), and Sea-Horse Optimization (SHO). The CNN-LSTM + PSO model outperforms standalone DL models by achieving 99.86% accuracy with a False Acceptance Rate (FAR) of 0.0009, False Rejection Rate (FRR) of 0.0012, and Equal Error Rate (EER) of 0.001, according to extensive assessment on the Touchalytics dataset. For next-generation mobile security, this combination of Swarm Intelligence (SI) and DL provides a strong, flexible, and effective authentication architecture.
Learning about Unity in Diversity can be challenging to implement effectively, as many students become disengaged with conventional lecture-based methods. This study aims to develop and evaluate a game-based learning model for character education that promotes the values of Unity in Diversity among elementary school students in Indonesia. In today's increasingly diverse world, fostering understanding of different cultures and perspectives is essential. The research employs two public schools in Indonesia. Quantitative data were collected through structured questionnaires, while qualitative data were obtained from classroom observations and student feedback a mixed-methods design integrated with the Game Development Life Cycle (GDLC), encompassing six stages: initiation, pre-production, production, testing, beta, and release. The study involved a sample of elementary school students ($N$ = 55) from two public schools in Indonesia. Quantitative data were collected through structured questionnaires, while qualitative data were obtained from classroom observations and student feedback. The developed game integrates core values such as tolerance, respect, cooperation, and empathy within an interactive digital learning environment. Findings reveal that the game-based learning model enhances student engagement and supports the development of diversity-related values, with mobile-based platforms identified as the most feasible and effective medium for implementation. These results suggest that integrating game-based strategies provides a more engaging, accessible, and meaningful learning experience for young learners.
This study aims to solve the problem of video storage and improves the overall efficiency of cameras by adopting real-time anomaly detection, hence informing the user about any suspicious anomalies. The proposed trained model processes live video streams, identifying unusual events and anomalies such as theft, weapons, or violent activities. Simultaneously, a video storage optimization algorithm reduces redundant frames while maintaining movement detected video streams from CCTV surveillance. In addition, if the model detects an unusual event occurring in the live video stream, it immediately notifies the user about the type of anomaly and the location of the event that occurred. Experimental results demonstrate that the proposed system effectively detects anomalies with an average precision–recall score of 0.958 and an F1 confidence score of 0.93, ensuring reliable threat identification and detection. The model is robust and differentiates between normal and anomalous activities as justified by experimental results.
This paper explores a joint method for efficient compression as well as the reliable transmission of images. In image transmission, low dormancy and very fast data transmission are the main requirements of modern wireless communications systems. Efficient and reliable image transmission through wireless networks need to handle several challenges like adverse wireless channel conditions, the need for high power consumption, managing high computational complexity, and low error resilience capability of image compression schemes. This paper deals with the compressed sensing approach in combination with orthogonal frequency division multiplexing (OFDM) to take care of the above challenges. There are three important steps in compressive sensing: sparse signal representation, measurement collection, and sparse recovery. In this process, a measurement matrix is utilized to sample those elements which are significant for accurately depicting the signal in the measurement step. So, the design of precise measurement matrices is crucial for compressive sensing. This paper proposes a Lanczos measurement matrix which significantly improves the quality of the reconstructed image with minimum data. At the same time, the use of OFDM handles multipath fading channels for reliable transmission of the image data. The simulation results also compare the performance of several measurement matrices in terms of image quality through peak signal-to-noise ratio (PSNR) value, the structural similarity index, and the Bit Error Rate (BER) for transmission performance after passing through Additive White Gaussian Noise (AWGN) as well as multipath channels.
This study applied a within-subjects experimental design to examine user engagement, perceived realism, and credibility when experiencing a virtual organic coffee shop through two immersive formats: Virtual Reality (VR) and 360◦ video. The VR system recreated a Hmong hill tribe cafe´ in Chiang Rai, Thailand, while the 360◦ video documented the authentic organic coffee pro- duction process from seed to roast. Forty-one participants experienced both systems in a counterbalanced order. The findings revealed a complex pattern of results: while initial Wilcoxon signed-rank tests found no significant differences between formats when comparing individual matched variables (p > 0.05), comprehensive analysis with Benjamini-Hochberg corrections revealed that VR demonstrated significantly superior presence (Z = −4.10, p < 0.001, r = 0.64), engagement (Z = −4.50, p < 0.001, r = 0.70), and spatial realism (Z = −3.85, p = 0.002, r = 0.60) with large effect sizes. A mixed-effects ordinal regression model confirmed VR’s advantage in purchase intent (β = 0.68, p = 0.011). Crucially, system usability emerged as a critical factor, showing very strong correlations with both overall user experience (ρ = 0.82, p < 0.001) and purchase intent (ρ = 0.70, p < 0.001) regardless of medium. The relationship between perceived authenticity and satisfaction was particularly strong (ρ = 0.55, p < 0.001), while presence (ρ = 0.42, p = 0.006) and credibility (ρ = 0.46, p = 0.003) showed moderate positive associations with purchase intent. These findings demonstrate that while the measurement approach significantly influences detected differ- ences between immersive media, the translation of immersion into consumer behaviour depends on multiple psychological factors, including usability, credibility, and authenticity, rather than the choice of medium alone.
Procuring resilience, resource efficiency, productivity, pest and malady control in agrarian production is imperative when climate change poses a threat. In recent years, hydroponics is considered as an emerging farming technique and is popular in urban areas due to its minimal water use and ability to grow plants without soil. In the Nutrient Film Technique (NFT) based hydroponics system, plants are cultivated by using water content nutrient solutions. Integration of Internet of Things (IoT) technology to NFT based hydroponic systems, many advancements such as minimizing water usage, real-time plant growth monitoring, efficient nutrient diffusion and reduction in human efforts can be achieved. In this work, an IoT based smart hydroponics system using NFT is proposed. Key components of the proposed solution include sensor networks for data acquisition, a robust Machine Learning (ML) framework for data analysis and prediction, as well as actuators for automated control of environmental conditions. The system monitors different real-time environmental parameters and the status of the plant's growth and controls the nutritional value of water in an automated and cost-effective way. A Support Vector Machine (SVM) algorithm is used to predict the pH values with an accuracy of 89.6%, surpassing the Decision Tree (DT) and Random Forest Regression methods.
Path planning generates a shorter path from source to destination based on sensor information acquired from an environment. An obstacle avoidance is an important task in robotics within path planning since the automatic functioning of robots requires reaching the destination without collisions. Moreover, obstacle avoidance algorithms have an important part in robotics. The existing algorithms did not enable robots to navigate their environments effectively, lessening the threat of collisions and preventing obstacles. Here, an Adaptive Spider Wasp Optimizer (ASWO) is introduced for path planning in mobile multi-robots. Initially, the simulation of an environment utilizing multiple robots and targets along with obstacles is accomplished. There-after, multi-objectives namely path smoothness, obstacle avoidance, and path length are considered. Lastly, path planning is conducted employing ASWO by considering fitness parameters such as path smoothness, obstacle avoidance, and path length. However, ASWO is designed by integrating adaptive concept with Spider Wasp Optimizer (SWO). In addition, ASWO achieved maximal value of fitness and path smoothness about 1.795 and 91.121% as well as minimal value of path length about 897.883 km.
This study proposes a data-driven interactive multimedia platform that integrates real-time environmental and demographic data visualization with AI-based facial generation to deliver dynamic and immersive user experiences. The system combines image generation AI with data visualization in a unified framework, enabling real-time, personalized interaction. Real-time weather and demographic data are collected and processed through public APIs provided by Seoul Metropolitan Government, and these data streams are mapped to visual parameters such as sky color, cloud density, and background environments to reflect local conditions dynamically. Facial generation is carried out using a fine-tuned stable diffusion model trained on a Korean facial dataset categorized by age and gender. The generated face meshes are refined using detailed expression capture and animation (DECA) and implemented as MetaHuman characters within Unreal Engine to produce expressive real-time avatars. The platform adopts a client–server architecture and leverages cloud-based asset management to efficiently handle real-time data and 3D resources. This approach demonstrates a novel form of interactive media experience that merges real-time public data with AI-driven personalization and presents new opportunities for interdisciplinary HCI research that bridges art, design, urban data, and artificial intelligence.
Traditional learning methodologies often fall short of accommodating diverse learner needs and adapting dynamically to individual learning paces and styles. This limitation underscores the growing need for personalized learning, which has the potential to significantly improve learning outcomes, foster deeper engagement, and enhance learner motivation. This study introduces a novel personalized recommendation framework (PRF) that leverages large language models (LLMs) and chain-of-thought (CoT) prompting techniques to advance personalized learning. Specifically, it proposes a strategic personalization framework that addresses learner heterogeneity by incorporating both preference-based and performance-based features. CoT prompting is integrated to simulate human-like sequential reasoning in LLMs, thereby improving the framework's adaptability and effectiveness. A case study was conducted in a computer programming course, a domain that requires both conceptual understanding and practical problem-solving, to evaluate the proposed framework. The assessment involved 15 expert reviewers who examined the framework's effectiveness and overall satisfaction. Experimental results showed that the proposed PRF generated recommendations perceived as significantly more satisfactory than those produced by the non-PRF system (M = 4.50 ± 0.30 vs. 3.73 ± 0.21, p < 0.001). In addition, the experts strongly agreed that the framework effectively identified students in urgent need of support, provided timely recommendations, and delivered personalized learning experiences aligned with individual learner needs.
The diagnosis of amyotrophic lateral sclerosis (ALS) experiences critical delays averaging 9-12 months, limiting therapeutic interventions. We propose a novel architecture that integrates deep learning with blockchain technology for secure and auditable speech ALS detection. Our CNN-BiLSTM architecture with an attention mechanism processes the acoustic characteristics of 217 participants (133 ALS, 84 controls) in the VOC-ALS and Minsk datasets. The model achieves 96.5% accuracy, 95.3% sensitivity, and 97.8% specificity, outperforming traditional approaches. The blockchain implementation on Optimism Layer-2 ensures data integrity through immutable audit trails, IPFS off-chain storage, and smart contract-governed access control. This hybrid approach addresses both diagnostic accuracy and critical data governance challenges in multi-institutional ALS research, demonstrating feasibility for clinical deployment while maintaining patient privacy and regulatory compliance.
Full search (FS) motion estimation provides high accuracy but at high computational cost, while fast search methods introduce irregular, hardware-unfriendly patterns. This work proposes four adaptive FS-based VBSME algorithms that preserve FS regularity while reducing complexity through direction-driven search, adaptive block sizes, stationary block revalidation (SBR), and early termination using a spiral pattern.On CIF sequences, vector-driven VBSME with early termination reduces SAD computations by 53.9–-94.5%, while SBR improves PSNR by up to 1.39 dB. For 1080p video, SBR-based FS VBSME achieves higher PSNR than conventional FSBME, while SBR-based vector-driven VBSME delivers nearly 60% fewer SAD evaluations.
Choosing which employees to promote is a complex task that demands both fairness and effectiveness. Machine learning has significantly improved promotion decisions by providing data-driven insights and automation. Clustering-based promotion models are popular, but their performance is often hindered by deficiencies in the input data, which are typically noisy, class imbalanced, and high dimensional. These problems can be alleviated through data preprocessing. Accordingly, this study introduces an analytical framework for employee promotion clustering that incorporates feature engineering – both feature extraction and feature augmentation – to enhance clustering performance and generalizability. The principal contribution is the development of the generative performance feature (GPF), an augmented representation that amplifies the influence of performance-oriented features extracted via principal component analysis (PCA). The GPF captures the intrinsic structure of the original dataset and is formulated as an additive composite feature. The PCA-transformed dataset combined with the GPF is then used to construct a clustering model. The proposed framework was evaluated on two public datasets with both K-means and fuzzy C-means (FCM) clustering models. The GPF led to significant improvements across key evaluation metrics, namely, the Rand index, the mutual information score, the V-measure, and the Fowlkes–Mallows index. K-means clustering demonstrated superior performance to FCM clustering across all evaluation metrics for both datasets.
Online fraud and social engineering tactics frequently use phishing websites as platforms. Phishers often modify the source code of the web pages they exploit in their attacks to create the illusion that alterations were made to authentic websites. A solitary response is insufficient to mitigate phishing due to the many methods employed in its execution. This study examines machine learning algorithms and evaluates their efficacy when trained on datasets including attributes that differentiate secure websites from phishing sites. Automated algorithms facilitate real-time fraud protection by swiftly detecting suspicious URLs, domain names, and website content. This study aims to identify the optimal method for detecting a prevalent category of cyberattacks. This would enhance the security and privacy of all internet users by facilitating the identification and blocking of malicious websites. Nonetheless, there is an urgent desire for automated models that provide rapid and precise detection. This research introduces a regression-based assessment method for phishing detection to address this demand. Our approach employs a whale optimization algorithm for feature selection. An AutoML framework subsequently utilizes the selected feature subsets as input. The model showed good accuracy in its predictions with very small errors on the test data, shown by an RMSE of 0.1079, an MSE of 0.0116, and an $\mathrm{R}^{2}$ value of 0.9534. These results demonstrate the reliability of our feature selection and modeling methods.
Ambulance response time is a critical factor in saving lives and is heavily influenced by ambulance placement strategies. Currently, Nang Lae Subdistrict is served by only one ambulance, yet the managing agency does not systematically analyze high-frequency accident zones or areas with frequent emergency calls. This oversight raises concerns about whether the current ambulance station provides optimal coverage. To address this, call data from all ambulance dispatches over four months (July-November 2023) were analyzed. The findings revealed 96 cases within a 5-kilometer radius of Nang Lae Subdistrict Municipality and 18 cases within a 10-kilometer radius. Cluster analysis was conducted to determine optimal ambulance placement, identifying two potential locations approximately 1.55 km apart. These results can inform strategic improvements in ambulance deployment, particularly during high-demand periods such as festivals, where accident rates surge and faster response times are crucial. Additionally, the study observed instances of ambulances servicing areas beyond their designated zones, suggesting a need for better resource allocation.
Conventional collaborative filtering methods struggle to understand the subtle connections between users and evolving short video content, often resulting in imprecise or broad recommendations. Additionally, the absence of interpretability, usually displayed as unchanging lists or tables, restricts user confidence and involvement. To overcome these challenges, this research suggests a tailored short video recommendation system that integrates a CNN-BiLSTM hybrid model with a Multi-Head Attention (MHA) approach and visualization techniques. CNN is utilized to obtain visual characteristics from brief videos, BiLSTM identifies temporal relationships in video sequences and user actions, while MHA improves feature weighting for tailored significance. To address the transparency concern, the system incorporates real-time visualization methods like heat maps and interactive charts, enabling users to grasp the reasoning behind each suggestion. Experimental findings from the MicroLens dataset indicate that the proposed model achieves a hit rate of 0.94 at k = 15, outperforming conventional methods such as ItemCF by 0.16. This method greatly enhances the accuracy of recommendations, transparency, and user engagement in digital media contexts
