This research aims to develop and evaluate the Lak Hok virtual tour navigation system to promote sustainable cultural tourism by showcasing Thai wisdom through immersive digital experiences. The system utilized 360-degree panoramic images hosted on a web server and supported accessibility via laptops, smartphones, and virtual reality (VR) headsets. Both subjective evaluations and objective performance metrics were employed to assess the system’s usability, aesthetic appeal, and content quality (CQ). User satisfaction, measured through a survey of 87 participants, demonstrated consistently high ratings (mean scores: 3.59-3.77 for ease of use (EU), 3.32-3.95 for design aesthetics, and 3.62-3.70 for content knowledge). Objective tests revealed an average system response time of 1.45 seconds, a false interaction rate of 4.2%, and a navigation accuracy of 98.5%. Statistical analysis showed no significant differences in user satisfaction across gender, age, or region, highlighting the system’s broad accessibility and usability. Unlike prior systems, this study formalizes satisfaction modeling via equation-based analysis. This virtual tour system provides a scalable and engaging platform for preserving and promoting cultural heritage, offering a sustainable solution for modern tourism development.
This paper presents a sub-threshold complementary metal-oxide semiconductor (CMOS) temperature sensor core for ultra-low-power applications, with the key advantage of reliable operation over an exceptionally wide temperature range from –100 °C to 100 °C, which is rarely reported in existing CMOS-based designs. The proposed architecture operates entirely in the sub-threshold region and is evaluated using circuit level simulations, with validation through comparison to a previously reported temperature sensor. Simulation results show excellent linearity across the full temperature range, achieving a coefficient of determination of R² = 0.99997 and a sensitivity of approximately 2.41 mV/°C. At a supply voltage of 1.4 V and 25°C, the sensor core consumes only 22 nW, highlighting its suitability for energy-constrained applications. These results demonstrate the potential of sub-threshold CMOS temperature sensing for wide-range, ultra-low-power sensing systems.
The rapid expansion of industrial internet of things (IIoT) adoption in Industry 4.0 has improved automation and real-time control yet simultaneously increased security risks in operational technology (OT) environments, where device integrity and system reliability are critical. Existing attestation approaches such as SAFEHIVE, SEDA, CRA, and ERASMUS provide scalable verification capabilities but still lack continuous hardware-rooted validation and adaptive access control required for real-time industrial systems. To address this gap, this study proposes a hybrid cybersecurity framework that integrates IoT device-trusted remote attestation (ID-TRA) based on trusted platform module (TPM) with zero trust architecture (ZTA) to ensure continuous device trustworthiness in brewery operations. The framework was implemented on an industrial testbed with programmable logic controllers (PLCs), edge devices, and industrial switches, and it was evaluated through measurements of attestation latency, false positive rate, communication overhead, and TPM resource utilization. Experimental results show that the framework achieves an average attestation latency of 250 ms, a false positive rate below 2%, and a communication overhead of only 1.1%, while TPM resource usage remains within acceptable bounds (62% CPU and 48 MB RAM). These outcomes demonstrate that the proposed solution can reliably detect unauthorized firmware modifications, prevent compromised devices from accessing critical network zones, and maintain compatibility with real-time control processes. Overall, the integration of ID-TRA and ZTA enhances device-level assurance and strengthens industrial cybersecurity resilience against firmware tampering, replay attacks, and unauthorized lateral movement.
Student behavior and activity play a crucial role in shaping the classroom atmo sphere and influencing the quality of a learning session. Recently, vision-based student activity recognition has gained significant attention. However, recog nizing student activities from classroom videos presents unique challenges due to the nature of the classroom environment, such as the presence of multiple students and severe occlusions. As a result, research in this area has often over looked these challenges. This study provides a detailed and comprehensive re view of student activity recognition from classroom videos. First, we formalize the problem of student activity recognition from videos and categorize existing methods into three distinct approaches: frame-level, clip-level, and continuous recognition. We then provide a detailed analysis of representative methods for each approach. In addition, we present a comprehensive overview of publicly available datasets for student activity recognition and discuss key open chal lenges, together with potential future research directions. Our analysis reveals that: (1) Most existing studies focus on frame-level recognition, while clip-based and continuous activity recognition remain relatively underexplored; (2) there is still a lack of large-scale, standardized benchmark datasets for vision-based stu dent activity recognition; and (3) existing research primarily emphasizes recog nition accuracy, whereas real-time performance and computational efficiency are rarely addressed.
This paper investigates the control performance of a solar-powered cable driven robot using MATLAB simulations, comparing a conventional proportional–integral–derivative (PID) controller with a hybrid fuzzy PD controller. The study adopts a simplified kinematic model to focus on control behavior under cable-induced nonlinearities and time-varying power availability due to solar energy, while neglecting full dynamic and cable tension effects. Both controllers were systematically tuned to ensure a fair comparison. Performance was evaluated in terms of speed and position tracking under fluctuating solar conditions. Simulation results show that the hybrid fuzzy PD controller provides superior performance, with lower RMS tracking errors, reduced overshoot, and faster settling times compared to the PID controller. These findings highlight the potential of energy-aware intelligent control strategies for improving the reliability and accuracy of solar-powered cable-driven robots operating under variable renewable energy conditions.
Social media has emerged as an important part of societal discourse on feminism and gender equality, especially in Bangladesh. Nevertheless, any feminist debate on social media in Bengali polarizes reactions, highlighting the need for automated sentiment analysis. This paper introduces one of the earliest multi-class feminist sentiment classification schemes of the Bengali social media with a manually annotated dataset of 6,830 comments categorized as positive, neutral, or negative. The framework uses term frequency-inverse document frequency (TF-IDF) based n-gram feature representations utilizing traditional machine learning algorithms, with a majority voting ensemble to determine optimal robust models. The data was divided into 80% and 20% for training and testing, respectively. Models were evaluated on the basis of accuracy, precision, recall, and macro-F1 to correct on imbalance of classes. Multinomial naive bayes (MNB) has the best accuracy of 84.74% and macro-F1 of 84.66, which is 4-7 times higher than other models. The ensemble method improved feature strength. Such results indicate that lightweight machine learning models based on TF-IDF features and ensemble models can be useful to detect feminist sentiment in Bangla social media and serve as a guideline in the field of domain-specific sentiment analysis in low-resource languages and help monitor online feminist discourse.
This study investigates user sentiment towards the Mobile JKN public health application by applying text classification models based on deep learning. Two approaches were compared: a multi-layer perceptron (MLP) with TF IDF features and long short-term memory (LSTM) with Word2Vec embeddings. The dataset consists of 114,364 Indonesian-language user reviews collected from the Google Play Store. To address class imbalance, we applied random oversampling. Each model was evaluated using 5-fold stratified shuffle split cross-validation. The results showed that MLP models achieved higher accuracy (up to 83.90%), while LSTM models demonstrated better recall and precision on minority classes such as neutral sentiment. However, statistical validation using the Wilcoxon signed-rank test revealed that the performance differences between models were not statistically significant (p > 0.05). These findings suggest that both models are viable for sentiment analysis, with trade-offs depending on the evaluation metric of interest. Future work may explore hybrid architecture and larger datasets for improved performance and statistical confidence.
Soil moisture monitoring is essential for precision agriculture to optimize irrigation and increase crop productivity. Traditional conductivity-based sensors often face limitations such as low sensitivity, slow response, and measurement instability. This study presents a simple and effective enhancement method by applying a graphene coating on copper electrodes using the drop casting technique. Experimental evaluations were conducted on natural soil samples at varying moisture levels. The graphene-coated sensor exhibited a significantly higher sensitivity of 23.0 Ω/% compared to 12.0 Ω/% for the uncoated sensor, a faster response time of approximately 5 seconds, and improved measurement consistency with a reduced standard deviation of ±15 Ω. Graphene's superior electrical conductivity and strong water affinity are key factors contributing to this performance improvement. These findings indicate that graphene-coated sensors offer a promising solution for reliable, cost-effective soil moisture monitoring in smart farming systems.
This paper addresses the optimal integration of wind turbines into distribution networks with the aim of reducing active power losses and improving voltage stability. Two metaheuristic optimization methods genetic algorithm (GA) and particle swarm optimization (PSO) are applied to determine the optimal siting and sizing of wind turbines in the IEEE 14-bus system. The problem is formulated as a multi-objective function combining loss minimization and voltage profile enhancement under standard network constraints. Simulation results using MATLAB/PSAT show that both algorithms improve system performance compared to the base case, with PSO providing superior loss reduction and voltage stability. Wind variability is represented through a Weibull distribution to reflect realistic operating conditions. The study demonstrates the effectiveness of metaheuristic optimization for renewable integration and highlights PSO’s stronger robustness. The work contributes a comparative evaluation of GA and PSO, supported by stability analysis and realistic wind modelling.
Accurate detection of network intrusions remains challenging under severe class imbalance, where rare attacks such as remote-to-local (R2L) and user-to-root (U2R) are poorly represented. Although many learning-based intrusion detection systems achieve high overall accuracy, conventional loss functions often bias training toward majority classes, leading to weak minority-class performance. This paper introduces a smooth margin-reciprocal loss (MRL), inspired by distance-weighted discrimination (DWD), which emphasizes samples with small or negative margins while rapidly attenuating penalties for well-classified instances. Unlike probability-based focal loss, MRL operates directly on the signed margin and enables stable optimization with first-order methods. Experiments conducted on the NSL-KDD benchmark using linear and shallow multilayer perceptron models show that MRL consistently improves macro-F1 and per-class precision–recall AUC compared with hinge, logistic, and focal losses, with notable gains on minority attack classes.
Parkinson's disease (PD) is a degenerative neurological disease, and at present there are no reliable laboratory tests for it. So how does this happen when people go to identify PD? vocal biomarkers, combined with machine learning (ML), seem to be an option for noninvasive diagnostics. In our work, we used a voice recording dataset which consisted of 26 different feature sets mined by various techniques. When using the extreme gradient boosting (XGBoost) method, out of all these models tested, an accuracy of 91.79% was achieved. As can be seen from its high precision, recall and F1- score, XGBoost performed very well in differentiating PD cases from non-cases. The study concludes that the application of ML, particularly XGBoost, to the diagnostic process can establish a valuable tool for early screening of PD, which will facilitate more speedy and correspondingly cost-effective clinical evaluations. This paper represents an important contribution to the rapidly developing fields of artificial intelligence-based on diagnosis of neurological diseases and digital health.
Temperature forecasting is important for industries affected by climate, especially in semi-arid regions where the weather can change quickly and is hard to predict over time. Many studies have examined various deep learning models, including long short-term memory (LSTM), gated recurrent unit (GRU), convolutional neural networks (CNNs), and transformer-based hybrids. However, their performance in data-limited semi-arid environments is often unclear and inconsistent. This study compares six deep learning methods for predicting daily maximum temperatures in Settat, Morocco. It uses 11 years of ground-observed meteorological data. The models examined include a baseline artificial neural network (ANN) and five hybrid structures: ANN-LSTM, ANN-GRU, ANN-CNN, ANN–random forest (RF), and ANN-transformer. The results indicate that the ANN performs the best overall, with MAE = 0.0432, root mean square error (RMSE) = 0.0543, and R² = 0.8820. It surpasses all hybrid models. When using a relative improvement metric, the ANN shows accuracy gains of 32% to 42% compared to the recurrent, convolutional, and attention-based hybrids. These results suggest that in semi-arid climates, where maximum temperature mainly depends on the same-day atmospheric conditions, simpler feedforward models work better than more complex temporal models. The study underscores the need to match model complexity with climatic factors and dataset size, offering a useful benchmark for temperature forecasting in regions with limited data.
The rapid growth of cloud computing demands efficient task scheduling strategies capable of handling heterogeneous resources, dynamic workloads, and multiple conflicting objectives. Existing approaches often optimize a single criterion, limiting their effectiveness in large-scale distributed systems. This paper proposes hybrid Bat–Whale optimization algorithm (BWOA), a hybrid scheduling algorithm combining the Bat algorithm and Whale optimization algorithm, enhanced with Lévy flight-based exploration, adaptive crossover, and a smart local search mechanism. The framework balances global exploration and local exploitation while preserving population diversity and intensifying search around promising solutions. A problem-aware local search reallocates long-duration tasks to high performance virtual machines and selectively swaps tasks with poor response times. Experiments on a heterogeneous cloud environment with 300 tasks and 50 virtual machines, using min–max scaling for workload normalization, demonstrate that BWOA outperforms classical methods, including first come, first served (FCFS) and Min-Min scheduling algorithms, achieving superior makespan (≈32.77 s) while maintaining competitive utilization, throughput, and energy efficiency. These results highlight the effectiveness of hybrid metaheuristic approaches integrating multiple optimization strategies for multi-objective task scheduling in large scale cloud systems, providing a robust and scalable solution for both academic research and practical deployment.
Following the COVID-19 pandemic, online learning platforms have become vital for supporting distance education. This work presents LABTEC, an online Experimental Platform for electronics education that enables students to manipulate real hardware through a learning management system (LMS). The platform allows remote execution of experiments with electronic circuits and instruments, such as oscilloscopes, providing hands-on practice over the Internet in real time. The main contributions of this work are threefold: (i) a hybrid Flask–Django server architecture, where flask manages instrument-level control and Django provides secure and scalable web services; (ii) the use of a Raspberry Pi gateway as a cost-efficient and versatile hardware interface; and (iii) an open-source remote laboratory framework experimentally validated to support real-time interaction with average end-to-end latency below 50 ms, stable multi-user access, and low resource utilization. Experimental results demonstrate reliable operation under concurrent user scenarios, achieving consistent measurement visualization and control with reduced deployment cost compared to proprietary and institution-centric remote laboratory platforms. Performance evaluation shows a control latency below 50 ms for closed-loop tasks, a success rate above 98% under multi-user access, and average CPU and RAM usage of 35% and 420 MB on Raspberry Pi 4B during peak load. These results demonstrate that the system is responsive, reliable, and suitable for concurrent experiments. Although validated with a single instrument type, the proposed approach offers a scalable and replicable solution that can significantly enhance electronics education and lower laboratory infrastructure costs.
Video classification is essential in computer vision, enabling automated understanding of dynamic content in applications such as surveillance, autonomous systems, and content recommendation. Traditional long-term recurrent convolutional network (LRCN) models, however, often struggle to capture complex spatio-temporal patterns, limiting classification performance across diverse video datasets. To address this limitation, we propose an enhanced LRCN with architectural refinements, optimized filter sizes, and hyperparameter tuning, improving both temporal modeling and spatial feature extraction. Experimental results on three benchmark datasets DynTex, UCF11, and UCF50 demonstrate that the proposed model achieves accuracies of 0.90 on DynTex (+26.8% over standard LRCN), 0.92 on UCF11 (+19.5%), and 0.94 on UCF50 (+1.1%), consistently outperforming ConvLSTM, LRCN, and other state-of-the-art approaches. These findings indicate that the enhanced LRCN effectively captures spatial and temporal dynamics in video sequences, setting a new benchmark for video classification. The study highlights the impact of architectural innovation and parameter optimization, providing a solid foundation for future research on scalable and efficient deep learning models for dynamic content analysis.
Efficient and environmentally friendly energy use for base transceiver stations (BTS) in remote areas is essential for telecommunication network development. This study simulates and compares two BTS configurations: a conventional grid-powered system and a hybrid solar-grid system, focusing on energy efficiency, operational cost, and carbon emissions. The simulation was conducted over a one-year operational period using Python-based modeling with realistic input parameters. The results indicate that the hybrid system can supply approximately 74% of the annual energy demand using solar power, achieving 24.4% operational cost savings and reducing carbon emissions by 73% compared to the grid-only system. These findings confirm that the hybrid BTS system is a feasible and sustainable solution to support telecommunication expansion in remote areas with lower cost and environmental impact.
This article optimizes a decentralized system for collaborative tourism in Alge ria using blockchain, smart contracts, and proof of reputation (PoR) consensus. The system matches services into organized trips, manages reservations, and automates payments to ensure transparency and autonomy without centralized authority. This work opens the door to exploring dual-blockchain architectures. Building on a previous work, we enhanced node interactions, automated con tract execution, and introduced a dual-blockchain structure to reduce latency while improving scalability and security.
This paper presents a performance evaluation of a fourth-generation (4G) cel lular network under adverse weather conditions in a tropical region. While the impact of rainfall on frequencies above 10 GHz is well documented, this study addresses the research gap concerning 4G LTE performance (sub-6 GHz) in high-precipitation environments such as Nigeria. Using a drive-test approach with TEMS Investigation software (v16.3), measurements were collected over 48 days between July and September 2025 along a fixed 15 km route in the Lagos metropolis on the MTN Nigeria network. Samples were recorded at 1 second intervals. Four critical key performance indicators (KPIs)—reference signal received power (RSRP), reference signal received quality (RSRQ), signal to-interference-plus-noise ratio (SINR), and received signal strength indicator (RSSI)—wereanalyzedtodeterminetheir influence on the network performance index (NPI). Correlation analysis revealed that while RSRP exhibits no sig nificant correlation with NPI during rainfall (rs = 0.009), SINR and RSRQ demonstrate strong positive correlations (rs = 0.828 and rs = 0.824, respec tively). Despite these high correlations, average performance values remained low (mean SINR = 23.72%), indicating significant rain-induced degradation. These findings provide a novel empirical basis for the development of weather aware adaptive algorithms in tropical 4G network deployments.
Defect detection plays a pivotal part in the manufacturing process of semiconductors. Defects can be rooted in the product on its own, as well as the tools used to process and make the product, particularly the equipment and machinery used. Defect detection is crucial in semiconductor manufacturing, where even minor flaws can compromise product performance. Defect detection in the backend process of semiconductor manufacturing, specifically in die attach and die bonding, is critical for ensuring product quality and reliability. Die attach involves securing semiconductor chips onto substrates, while die bonding involves connecting wires to the chip. Detecting defects during these processes is vital to prevent issues such as misalignment, inadequate bonding, or contamination, which can lead to malfunctioning chips or devices. Various techniques such as visual inspection, automated optical inspection (AOI), and X-ray imaging are utilized to identify defects like cracks, voids, or irregularities in bond formation. By employing rigorous defect detection measures, manufacturers can uphold stringent quality standards and produce reliable semiconductor devices for various applications.
Token type identification lies at the core of named entity recognition, allowing models to distinguish named entities from non-entity tokens and thereby better capture sentence meaning. This paper presents a deep learning approach for the Arabic named entity recognition task, leveraging deep neural networks and pretrained language models. The proposed model is a combination of the AraELECTRA language model with the bidirectional long short-term memory (BiLSTM) neural network. We utilize the WojoodNER dataset, which provides fine-grained annotations of Arabic text across 21 entity types. The results of this approach are encouraging, with an accuracy of 98.29% and an F1-score of 87%.
