Interactive (Poster) Session 2
30 October 2025 (1:30pm - 3:30pm)
30 October 2025 (1:30pm - 3:30pm)
Interactive Session C.1
Room: Foyer of Sipadan
Chairs: Md Pauzi Abdullah (UTM, Malaysia), Fatanah Mohamad Suhaimi (Universiti Sains Malaysia, Malaysia)
#1 Automated Osteoarthritis Classification Using Convolutional Neural Networks on X-Ray Data
Abhinay Bhandekar (IIIT Naya Raipur, India); Akshar Teja Gannoju (International Institute of Information Technology Naya Raipur, India); Kommineni Akhil S R Chowdary and Mallikharjuna Rao K (IIIT Naya Raipur, India)
Degeneration of cartilage and joint structures due to age is a leading cause of knee osteoarthritis (OA), a condition that results in pain and loss of mobility. Accurate diagnosis and early classification of OA are vital for proper treatment and disease control. Diagnosis is normally carried out using X-ray imaging by radiologists, an activity that takes time and is prone to variation. This research explores the use of deep learning, namely convolutional neural networks (CNNs), for OA classification in an automated fashion. A ResNet-18 model is trained on a dataset of Kellgren-Lawrence (KL) graded X-ray images to classify OA severity. Gradient-weighted Class Activation Mapping (Grad-CAM), which finds salient regions influencing predictions, improves interpretability while data augmentation and preprocessing techniques increase model robustness. Although differences in performance across severity levels suggest that more optimization is needed, the results show that deep learning can support OA diagnosis. Future goals include integrating clinician knowledge and improving model generalization.
#2 Enhanced Mechanism for Neighbor Discovery Protocol Table Exhaustion Attacks in IPv6 Networks
Navaneethan CArjuman (CoE for Advanced Cloud, Faculty of Computing and Informatics, Multimedia University, Cyberjaya, Malaysia); Muhammad Ukasyah Bin Md Yusof and Zar Chi Hlaing (Faculty of Computing and Informatics, Multimedia University, Malaysia); Sami Hassan Omer Salih (Al Jouf University, Sudan & Sudan University of Science and Technology, Saudi Arabia)
The adoption of Internet Protocol version 6 (IPv6) has increased due to the requirements for a larger address space and better suitability for modern networks, such as Internet of Things (IoT) networks, but it also introduces new security challenges. The Neighbor Discovery Protocol (NDP) that supports features like the router discovery and address resolution, is not secured due to its stateless and trust-based feature. The most notable one is the NDP Table Exhaustion attack that consists of overwhelming the network with forged Neighbor Solicitation (NS) and Neighbor Advertisement (NA) packets, inducing neighbor cache overload and denial-of-service (DoS). Current defenses, such as Secure Neighbor Discovery (SEND) or machine learning-based detection models, have significant computational overheads and cannot be adapted to resource constrained networks, creating a gap in the overall security of IPv6. To address this problem, this research paper suggests a lightweight, flow-based detection framework that integrates an entropy-based analysis with rule-based thresholds to detect NS and NA flooding attacks. The system gathers traffic into flows, entropy checks on key fields such as source Internet Protocol (IP) and network prefix, and alerts on anomalies when the traffic moves out of the normal operation range. It had the average precision of 0.91, average recall of 0.88, and F1-score of 0.87, but with minimal overhead and highly deployable in both enterprise and IoT networks.
#3 Design of Low-Power Active Inductor-Based Digitally Controlled Oscillator for All-Digital PLL in Zero-Energy IoT Devices
Xi Zhu (University of Technology Sydney, Australia); Jefferson A. Hora (MSU-Iligan Institute of Technology, Philippines); Ricster franuel Sibala (Center for Integrated Circuits Design (CICD) MSU-IIT, Philippines); Haniah L Mohammad (Mindanao State University - Iligan Institute of Technology, Philippines)
This paper presents the design and performance evaluation of a low-power, high-frequency Digitally Controlled Oscillator (DCO) incorporating an active inductor, implemented using 22nm Fully Depleted Silicon-On-Insulator (FDSOI) technology. Tailored for energy-constrained Internet of Things (IoT) devices-particularly those relying on RF energy harvesting-the proposed DCO achieves a frequency tuning range from 44.78 GHz to 45.35 GHz, while maintaining a total power consumption of less than 12 mW. The integration of an active inductor replaces conventional bulky passive inductors, enabling significant area reduction and improved compatibility with standard CMOS processes. This makes the design highly suitable for compact and scalable on-chip RF systems. Although the design exhibits a moderate phase noise of -78.96 dBc/Hz at 1 MHz offset, it offers a well-balanced trade-off between power efficiency, frequency performance, and spectral purity. Overall, the proposed DCO architecture demonstrates a viable solution for next-generation, low-power mmWave transceivers in dense IoT networks and other wireless sensor applications.
#4 Machine Learning-Based Analysis Method for Brain Activity Measurement Data Under Pleasant and Unpleasant Stimuli
Takumi Ishimoto, Rui Takahashi, Shuya Shida and Yutaka Suzuki (Toyo University, Japan)
In this study, we investigated the feasibility of automatic emotion classification by applying the machine learning algorithm Random Forest to time-series brain activity data measured using functional near-infrared spectroscopy (fNIRS). fNIRS is a noninvasive neuroimaging technique capable of monitoring cerebral hemodynamics by detecting changes in oxygenated and deoxygenated hemoglobin concentrations in the cortical surface. It offers practical advantages such as portability, safety, and ease of use, making it suitable for real-world emotion research.
Fifteen healthy male participants were presented with emotionally valenced visual stimuli, consisting of pleasant, unpleasant, and neutral images that had been pre-classified based on previous subjective ratings. During the task, changes in blood flow in the prefrontal cortex were recorded using 22 fNIRS channels. After preprocessing the fNIRS signals to reduce drift and physiological noise, we applied statistical analyses and machine learning techniques to evaluate emotional responses.
Specifically, we conducted paired t-tests to assess significant differences in hemoglobin levels before and during stimulus presentation, and used the feature importance metric from Random Forest to identify which brain regions contributed most to emotion classification. The classification model achieved a remarkably high accuracy of 99.48% using all channels and maintained strong performance (96.04%) even when restricted to the top five most informative channels.
These results indicate that the integration of fNIRS and machine learning is a promising approach for developing objective, reliable, and noninvasive methods for emotion recognition, with potential applications in fields such as affective computing, mental health assessment, and brain-computer interfaces.
#5 Design of Low-Power Digitally Controlled Oscillator for All-Digital PLL Receiver in Zero-Energy IoT Devices
Jacob Anthony M Agito (Mindanao State University Iligan Institute of Technology, Philippines); Xi Zhu (University of Technology Sydney, Australia); Jefferson A. Hora (MSU-Iligan Institute of Technology, Philippines); Haniah L Mohammad (Mindanao State University - Iligan Institute of Technology, Philippines)
The proliferation of Internet of Things (IoT) devices in the era of sixth-generation (6G) wireless communication demands ultra-low power and high-frequency circuit solutions to support sustainable, near-zero energy operation. As IoT devices become more pervasive in applications such as smart environments, industrial automation, and wearable electronics, there is a growing need for compact, energy-efficient frequency generators capable of operating in the millimeter-wave spectrum. This study presents the design and analysis of a Digitally Controlled Oscillator (DCO) operating at 45 GHz, conforming to the IEEE 802.11aj standard, and implemented using GlobalFoundries' 22nm Fully Depleted Silicon-On-Insulator (FDSOI) technology. The oscillator core adopts a complementary Colpitts topology, chosen for its efficiency and suitability at high frequencies. Frequency tunability is achieved through a digitally controlled switched capacitor array, implemented with low-leakage transmission gates. Operating at a low supply voltage of 0.8V, the proposed DCO achieves a power consumption of only 5.959 mW and demonstrates excellent spectral purity, with a phase noise of -99.03 dBc/Hz at 1 MHz offset, making it ideal for ultra-low-power mmWave transceivers in dense IoT networks.
#6 Crowdsourced Geospatial Cellular Data and Fuzzy Rule-Based Inference for Optimizing Network Selection and Radio Coverage Mapping in Maritime Operating Zones
Maria Gemel B Palconit, Jonathan Maglasang, Jayson C Jueco, Mary Nathaline Sevilla, Rau Lance Cunanan and Gabriel B Valenzuela (Cebu Technological University, Philippines)
Reliable and adaptive wireless communication is critical for Maritime Autonomous Vehicles (MAVs) operating in hybrid land-sea environments. This study proposes an intelligent framework that integrates crowdsourced cellular geodata with fuzzy logic inference to optimize mobile network selection and radio technology mapping across maritime zones. Using publicly available OpenCellID data, k-means clustering was applied to stratify cell tower distributions and determine radio-type availability in key land and offshore locations. Then a fuzzy inference system (FIS) was developed to assess the preference of the mobile network operator and predict the optimal radio technologies, GSM, UMTS, or LTE, based on signal strength, tower proximity, sample density, and location priority. Visualization techniques, including geospatial graphs, surface response maps, and signal overlays, were used to validate the FIS outputs. The results show that Smart (MNC 3) is the top choice of operators due to its strong coverage of LTE at ground stations and reliable UMTS access in maritime areas. This radio mapping is an important reference point for creating adaptable data routing strategies, which are crucial to maintaining efficient low-power communications with MAVs that can handle delays. This approach contributes to a scalable, data-driven methodology to enhance wireless communication reliability in maritime Internet of Things (IoT) systems.
#7 Soft-Actor Critic Deep Reinforcement Learning for Reconfigurable Intelligent Surface in Vehicle-to-Everything Network
Reng Yi Kueh (Curtin University, Malaysia); Choo Wee Raymond Chiong (Curtin University Malaysia, Malaysia); Lenin Gopal (University of Southampton Malaysia, Malaysia); Filbert H. Juwono (Xi'an Jiaotong - Liverpool University, China); King Hann Lim (Curtin University Malaysia, Malaysia); Huo-Chong Ling (RMIT University Vietnam, Vietnam)
With the introduction of 5G technologies that promise lower latency, more reliable, and higher network capacity, many researchers have proposed to integrate various 5G technologies such as millimeter waves, multi-radio access technology, and reconfigurable intelligent surfaces (RIS) with vehicle-to-everything (V2X). RIS is first introduced as an arranged array of reflectors that can be independently and dynamically reconfigured to adjust the signal propagation direction in a favorable way. With RIS, the overall performance of the wireless communication system can be improved. However, it is unclear whether RIS can be directly applied in areas of high mobility, specifically within vehicular traffic. This paper presents an RIS-aided network model that can be utilized to enhance the V2X communication networks with multiple vehicles (V2V) and infrastructure (V2I) links. A Soft-Actor Critic Deep Reinforcement Learning (SAC-DRL) approach is proposed to optimize the network model to achieve the stringent quality-of-service (QoS) requirements. Subsequently, numerical simulations were performed to validate the performance of the proposed SAC-DRL method.
#8 Enhanced-ZNN-Based Fixed-Time Model-Free Adaptive Control of Robotic Manipulator
Jin Siang Yap, Ibrahim Bako Abdulhamid and Muhammad Nasiruddin Mahyuddin (Universiti Sains Malaysia, Malaysia); Siang Kok Chia (Sandisk, Malaysia)
This paper presents an enhanced Zeroing Neural Network (ZNN) that is incorporated in the adaptive update law of a fixed-time model-free adaptive control scheme which improves the tracking performance of robotic manipulator. Although there are many research related to precise trajectory tracking control of robot manipulator using conventional model-based methods, it often requires high-fidelity model of the robot system, which is essential for accurately capturing its nonlinear dynamics. However, acquiring such precise models can be challenging due to parameter uncertainties, external disturbances, and unmodeled dynamics, that will lead to potential degradation in control performance or even instability when assumptions are inaccurate. To overcome this limitation, model-free adaptive control (MFAC) is gaining prominence because it directly estimates the unknown robot dynamics from the input and feedback data without the necessity for prior knowledge of robot parameters. The proposed control scheme is an extension of our previous work, which improves the tracking performance compared to the prior approach. A rigorous Lyapunov stability analysis, complemented by extensive simulations, was conducted to validate its effectiveness and demonstrates superiority in achieving precise and stable control.
#9 Fixed-Time Consensus Tracking for Nonlinear Second-Order Multi-Agent Systems with Communication Delays
Ibrahim Bako Abdulhamid, Jin Siang Yap and Muhammad Nasiruddin Mahyuddin (Universiti Sains Malaysia, Malaysia)
This study proposes a novel fixed-time tracking control methodology tailored for second-order multi-agent systems (MASs) operating over directed communication networks and experiencing communication delays. Recent studies have explored fixed-time consensus in MASs with nonlinear dynamics and communication delays. The control design employs a Lyapunov-Krasovskii (L-K) functional framework, integrating Hölder's inequality and a Lipschitz-like condition to ensure system stability. Notably, this approach eliminates the dependence on Linear Matrix Inequalities (LMIs), which are prevalent in traditional control strategies but often lead to increased computational complexity. By circumventing LMIs, the proposed method enhances computational efficiency, making it suitable for large-scale MAS applications. The control protocol guarantees that all agents achieve consensus tracking of the leader's trajectory within a predetermined fixed time, regardless of initial conditions. Theoretical analyses are substantiated through numerical simulations, demonstrating the effectiveness and robustness of the proposed scheme in achieving rapid convergence and maintaining stability in the presence of communication delays.
#10 Learning Alphabets, Seeing Worlds: Enhancing Early Childhood Education Using AR-Powered Handwritten Alphabet Recognition
Sajeena S and Ravanam Durga Venkata Satya Avinash (National Institute of Technology Calicut, India); Chinthakommadinne Vinay Kumar and Konduru Dhanush (NIT Calicut, India); Santosh Kumar Behera (National Institute of Technology Calicut, India); Ajaya Dash (IIIT Bhubaneswar, India)
The purpose of this work is to create an augmented reality (AR) application for supporting primary education that accelerates early literacy through interactive and immersive experiences. Since the early days of AR research, fiducial markers have acted as the backbone of AR applications, on top of which virtual computer-generated 3D objects are superimposed. The popular fiducial markers are printed characters or symbols surrounded by a thick black border in order to make the marker detection and identification task easier. During the testing of an AR alphabet learning kit with printed markers, we found that the learning engagement would have been significantly higher if the printed fiducial markers had been swapped out for the children's handwritten characters on a plane surface. The idea presented in this article is to build a dynamic AR learning space where 3D models corresponding to handwritten alphabet letters are projected, turning conventional learning into a visually engaging experience. Also, to enhance the experience further, when a letter is chosen or touched by a child, the system automatically projects a related 3D object in AR, such as an apple for the letter ‘A' or a ball for ‘B'. This spatial and visual correspondence of letters with actual objects aids in strengthening recognition, memory, and understanding abilities among young learners. To mitigate this, we have trained a custom dataset of handwritten alphabets for object detection and identification. After successful identification of the alphabet, the corresponding 3D objects are superimposed over the alphabet with respect to the pose in an AR setting. The application exploits AR technology through software like Unity and AR development kits to incorporate 3D content into the real world via mobile or tablet devices. The AR application covers a wide range of learning styles and keeps young learners actively engaged through the integration of visual, auditory, and tactile interaction. This work not only fills the chasm between physical interaction and visual learning but also conforms to current pedagogic practices that favor experiential learning. This is a beneficial teaching aid for parents and teachers by providing a fun yet practical means of promoting early literacy skills in children. The end goal is to revolutionize early learning by making education both fun and educational, hence encouraging curiosity and motivation in the minds of children.
#11 Computational and Physiological Perspective of Gender Based Pulmonary Fibrosis Severity
Anjana K and Selvaganesan, N (Indian Institute of Space Science and Technology, India)
Pulmonary fibrosis is a progressive interstitial lung disorder characterized by thickening of alveolar capillary membrane, resulting in reduced gas exchange. Although these disease has been extensively examined, the impact of gender on the severity and course of pulmonary fibrosis remains inadequately investigated. This work offers a computational simulation of alveolar membrane thickening to assess its effect on pulmonary gas exchange, emphasizing gender specific physiological variations. The alveolar compliance was decreased from 30% to 47% by varying alveolar membrane thickness in order to assess the severity of pulmonary fibrosis. In mild pulmonary fibrosis, the results show that gender does not substantially influence gas exchange parameters, such as oxygen saturation levels. However, in the case of severe pulmonary fibrosis the oxygen saturation of the male patient is reduced to below 88% indicating the necessity for oxygen therapy. The simulation results suggest that pulmonary fibrosis results in earlier and more severe gas exchange limits in males compared to females at equal membrane thickening percentages due to anatomical and functional differences.
#12 Computer Vision-Guided Humans and Animals Detection in Aquatic Disaster Zones
Raunit Maurya, Sajeena S and Santosh Kumar Behera (National Institute of Technology Calicut, India); Ajaya Dash (IIIT Bhubaneswar, India)
In geographically diverse countries like India, where floods occur frequently, recurring natural disasters pose a significant challenge to timely and effective disaster response. This work addresses the critical need for surveillance of flood-affected regions, with a focus on detecting the presence of humans and animals in submerged areas. Leveraging advancements in computer vision and artificial intelligence, the proposed system aims to enhance rescue operations through rapid and accurate life detection in aquatic disaster zones. By integrating object detection and semantic segmentation techniques, the solution provides a practical and scalable approach for early identification and improved situational awareness, thereby enabling prompt and effective action by emergency responders. This article at first addresses a dataset creation due to the lack of annotated dataset availability involving animals in waterlogged environments. The dataset is created using real-world videos and AI-generated images simulating various flood conditions. Further, an animal detection system is proposed based on a merged vision pipeline, where the detection is achieved by YOLOv8 supported by semantic segmentation using U-Net. The results of YOLOv8 and U-Net are combined to create the robust context-aware interpretation such as whether the detected object could be mapped to a flood area located in or near submerged regions. The proposed merged pipeline can achieve the purpose with mAP@0.5 as 0.81 and recall as 0.61. The method is useful to reduce irrelevant object detection and provide a meaningful output for use in rescue planning.
#13 Investigation on Q-Band Planar Modified Angled Log-Periodic Meander Slot-Line Slow-Wave Structures
Xinyu Xiang (UESTC, China); Lixia Yang (University of Electronic Science and Technology of China, China); Longfei Dang (Intelligent Policing Key Laboratory of Sichuan Province, Sichuan China); Shaomeng Wang (UESTC, China); Qingying Yi (University of Electronic Science and Technology of China, China); Yubin Gong (UESTC, China)
The traveling wave tube (TWT) has been a competitive candidate for power amplifiers in millimeter wave and terahertz wave communications, for its higher efficiency and broader bandwidth compared to the solid-state power amplifiers. Being the site for energy exchange between a high-speed electron beam and a slow electromagnetic wave, the slow-wave structure (SWS) plays an important role in a traveling wave tube. To improve the power and gain performance of slow-wave structures, a novel planar modified angled log-periodic (MALP) meander slot-Line slow-wave structure (MSL-SWS) is proposed. The high-frequency characteristics of the proposed SWS have been systematically investigated. Particle-in-cell (PIC) simulation results indicate that the proposed aperiodic structure exhibits superior performance to the planar periodic slow-wave structure (MSL-SWS), delivering a 10% enhancement in output power while maintaining improved gain characteristics. The SWS has achieved an output power of over 70W across the frequency range from 39 GHz to 41 GHz. Furthermore, the proposed structure demonstrates remarkable broadband performance, with a 3-dB bandwidth reaching 6 GHz, reaching 15% relative bandwidth.
#14 Optimizing Mobility in Future Wireless Networks with a DDPG-Based UAV-RIS Framework
Yasir Ullah (Multimedia University, Malaysia); Mardeni Roslee (MMU, Malaysia); Mohd Azmi Ismail (TM Research & Development, Malaysia); Farman Ali, Idris Olalekan Adeoye and Irfan Khan (Multimedia University, Cyberjaya, Malaysia)
The substantial growth of beyond 5G (B5G) networks necessitates innovative technologies to support high-speed, reliable, and low-latency communication. Unmanned Aerial Vehicles (UAVs) and Reconfigurable Intelligent Surfaces (RIS) have emerged as promising solutions to overcome the limitations of ground-based networks, particularly in urban environments with severe line-of-sight (LoS) blockages. However, existing UAV-assisted communication strategies still face challenges related to handover failures (HOF) and coverage consistency. This paper proposes a deep deterministic policy gradient (DDPG)-based UAV-RIS framework that jointly optimizes UAV trajectories and RIS phase shifts to address these challenges. The proposed UAV-RIS framework improves throughput and LoS probability while reducing outage probability (OP). Simulation results show that the proposed approach outperforms UAV-only, RIS-only, and without UAV-RIS deployments, offering superior performance in terms of throughput, LoS probability, and reduced OP. The findings highlight its potential to transform urban wireless communication by mitigating LoS blockages and ensuring uninterrupted connectivity in dense urban environments.
#15 Overcoming Data Scarcity in Load Forecasting Using Time Series Transfer Learning
Jayson C Jueco, Wilen Melsedec O. Narvios, Ferdinand F. Batayola, Rafran P de Villa, Cheryll Ann C Villamor, Gilbert Silagpo and Maria Gemel B Palconit (Cebu Technological University, Philippines)
Data scarcity introduces variabilities that compromise forecasting accuracy and reliability. In energy systems, precise load forecasting is vital for grid optimization, resource management, and outage prevention, supporting operational and economic stability. However, most models perform poorly under short time series, non-stationary patterns, and abrupt demand shifts. This paper employs Gramian Angular Field (GAF) to encode load data as images and develops a hybrid Autoencoder-Stacked LSTM network. The Autoencoder extracts latent temporal features from load snapshots, while the Stacked LSTM forecasts day-ahead hourly demand. A ten-year hourly demand dataset from the Philippine grid is aggregated, with missing values zero-filled and outliers preserved. The Autoencoder compresses snapshots into a latent space for feature representation, and transfer learning enables LSTM models trained on data-rich sub-grids to forecast in data-scarce regions. Performance, evaluated via MAE, MSE, MAPE, and RMSE, remains consistent across sub-grids. ANOVA results $(F < 0.002, p > 0.94)$ confirm no significant performance differences, validating the model's robustness and generalizability. Future work focuses on integrating attention, multimodal data, and real-time deployment to enhance forecasting accuracy and grid adaptability.
#16 Multi-Criteria Prioritization and Clustering of Stochastic on-Road Vehicle CO2 Emissions Based on Road Slope, Speed, and Acceleration Using K-Means, PCA, and Fuzzy AHP
Maria Gemel B Palconit and Jayson C Jueco (Cebu Technological University, Philippines)
The carbon emissions from public utility vehicles (PUVs) in the Philippines are projected to contribute up to 80% of vehicle kilometers traveled and become a major source of emissions by 2035 without intervention. Recognizing the stochastic and condition-dependent nature of vehicular emissions, the research aims to identify and prioritize the factors influencing (CO_2) emissions using advanced analytical techniques. Emissions data were clustered using K-Means to identify distinct operational states, while Principal Component Analysis (PCA) reduced dimensionality and revealed key influencing factors. The Fuzzy Analytic Hierarchy Process (FAHP) was then applied to prioritize these factors, considering environmental, technical, and economic implications. Results showed a positive relationship between road slope and (CO_2) emissions (r = 0.3111) and an opposite relationship with respect to speed (r = -0.3078), while acceleration had a minor positive effect (r = 0.1330). FAHP assigned the highest weight to (CO_2) emissions (0.3589), followed by slope (0.3121) and speed (0.2972). Cluster analysis highlighted Cluster 0 as the most emission-intensive operational state, with an average (CO_2) level of 31394.25 g/km, moderate speed, and uphill road conditions. The integration of PCA and FAHP revealed that (CO_2) emissions and slope together accounted for over 67% of the emission profile importance. These insights inform the development of emission control strategies, eco-driving guidelines, and data-driven transport policies.
#17 Evaluating Factors Affecting Headed Reinforcement Performance: a Case for External Beam-Column Joints for Seismic Resistant and Sustainable Structures
Hana Astrid R. Canseco-Tuñacao (Cebu Technological University, Philippines)
Approaches in studying the seismic performance of structural members are through the conduct of rigorous and costly experimental work with highly mechanistic models to explain behaviors. For the case of exterior beam-column joints with headed reinforcement, previous studies have identified factors that contribute to the performance of the member, and these have been included in code provisions. From a design perspective, there is missing information on the prioritization of factors. This study conducted a statistical approach using multiple linear regression, correlation, and dominance analysis to evaluate the significance and contribution of five key factors based on an adopted database. These factors are (1) Axial load applied (Axial), (2) ratio of the joint transverse reinforcement supplied parallel to the applied shear load to the required (ρ), (3) minimum between clear bar spacing between bars in a layer and spacing of bars between layer (c12), (4) clear cover to the bar (c3), and (5) column moment ratio (Mtest/2MNC). Results show that for specimens with good performance (Category I), c12, ρ, and c13 are significant and dominant factors due to the combined action of bonding along the reinforcing bar and confinement. For specimens with poor performance (Category II), c12 and Axial are significant and dominant factors. Axial confinement improves performance through bonding by delaying the development of cracks in the joint core. Addressing the failure of beam-column joints improves their seismic performance, resulting in the structure's overall durability and longevity.
#18 Runtime Optimization of Mahalanobis Distance for Anomalies in Credit Card Fraud Detection Dataset
Selverino A. Magon (De La Salle University, Philippines & University of the Philippines, Philippines); Ronnie Concepcion II, Argel Bandala and Ryan Rhay P. Vicerra (De La Salle University, Philippines)
Efficient anomaly detection is crucial in credit card fraud prevention, where delays in processing can lead to significant financial losses. This study investigates the efficiency of two covariance matrix inversion methods-Normal Inversion and LU Decomposition-using a dataset of over 284,000 transactions. The dataset, transformed through Principal Component Analysis (PCA), exhibits an imbalanced distribution, with fraudulent transactions comprising only 0.172%. Preprocessing steps included feature selection and covariance regularization to ensure numerical stability. Runtime for each method was recorded using Python's time module, and statistical analysis via a paired t-test assessed performance differences. Results show that LU Decomposition reduces runtime by 41.82% compared to Normal Inversion, attributed to its ability to factorize matrices and solve systems efficiently, avoiding computationally expensive matrix inversion. The significance of this improvement was validated with a p-value of 0.0, emphasizing LU Decomposition's computational advantage. Faster MD computation supports real-time fraud detection systems and provides a scalable solution to computational challenges, forming a basis for integrating optimized methods into machine learning frameworks for fraud prevention.
#19 Hybrid Renewable System Optimization with Backup Integration and Sensitivity Analysis
Mohammad Reza Maghami (Asia Pacific University of Technology & Innovation, Malaysia); Jacqueline Lukose and Ka Fei Thang (Asia Pacific University of Technology and Innovation, Malaysia); Arthur G.O. Mutambara (University of Johannesburg, South Africa)
As solar and wind energy gain global traction for their sustainability, their intermittent nature and reliance on weather conditions pose serious reliability challenges. Additionally, the high upfront costs of renewable energy systems hinder their broader deployment, particularly in off-grid and rural areas. Many existing hybrid energy systems (HES) struggle to balance affordability with consistent power delivery, especially under fluctuating load and supply conditions. This underscores the need for optimized HES designs that can reduce power outages while keeping costs manageable. This study introduces an innovative hybrid energy system that combines solar and wind power with battery storage, enhanced by Proton Exchange Membrane Fuel Cells (PEMFCs) and diesel generators as backup sources. To improve energy capture from renewables, a Maximum Power Point (MPP) tracking technique is employed, boosting overall system efficiency. The core advancement lies in integrating backup solutions and power optimization within a unified, multi-objective framework. Four operational scenarios are analyzed: a basic HES, HES with PEMFC backup, HES with diesel generator backup, and HES with MPP tracking. The system is tested in a rural community in South Khorasan Province, Iran. Optimization results show a reduced Net Present Cost (NPC) of $118,800 and a minimal Loss of Power Supply (LPS) of 50 kW. Sensitivity analysis highlights the impact of battery state-of-charge violations, unmet energy demand, and excess generation, reinforcing the importance of adaptive and resilient system design.
#20 PlantEase: an Arduino-Powered Solution for Automating Seed Crop Cultivation and Planting
Joan Katherine N Romasanta (National University, Philippines & NU Dasmarinas, Philippines)
In the Philippines, traditional farming methods are still used. Modern techniques like drip irrigation, controlled-release fertilizers, and digital tools are helping farmers improve efficiency, reduce waste, and increase profits. This research focuses on developing a device to automate crop cultivation and harvesting, offering an alternative to traditional practices and addressing the need for innovative solutions in agriculture like the Philippines' farming context. The system uses an LCD to show the soil moisture levels in real time, two servo motors for automated soil cultivation, and a micro servo for regulated seed dispensing. Furthermore, an ultrasonic sensor tracks the height and development of plants, offering useful information for improving crop management. All servos are synchronized by an infrared (IR) controller, which guarantees precise and efficient planting task execution. The project was evaluated by experts and attain the score of 3.88 which is described as Excellent. This means that the project is a flexible and scalable solution for to decrease human labor, improve planting accuracy, and promote sustainable agricultural methods.
#21 A Weighted Vision Transformer-Based Multi-Task Learning Framework for Predicting ADAS-Cog Scores
Nur Amirah Abd Hamid (Universiti Brunei Darussalam, Brunei Darussalam); Ibrahim Shapiai (Malaysia-Japan International Institute of Technology (MJIIT), Universiti Teknologi Malaysia, Malaysia); Daphne Teck Ching Lai (Universiti Brunei Darussalam, Brunei Darussalam)
Prognostic modeling is essential for forecasting future clinical scores and enabling early detection of Alzheimer's disease (AD). While most existing methods focus on predicting the ADAS-Cog global score, they often overlook the predictive value of its 13 sub-scores, which reflect distinct cognitive domains. Some sub-scores may exert greater influence on determining global scores. Assigning higher loss weights to these clinically meaningful sub-scores can guide the model to focus on more relevant cognitive domains, enhancing both predictive accuracy and interpretability. In this study, we propose a weighted Vision Transformer (ViT)-based multi-task learning (MTL) framework to jointly predict the ADAS-Cog global score using baseline MRI scans and its 13 sub-scores at Month 24. Our framework integrates ViT as a feature extractor and systematically investigates the impact of sub-score-specific loss weighting on model performance. Results show that our proposed weighting strategies are group-dependent: strong weighting improves performance for MCI subjects with more heterogeneous MRI patterns, while moderate weighting is more effective for CN subjects with lower variability. Our findings suggest that uniform weighting underutilizes key sub-scores and limits generalization. The proposed framework offers a flexible, interpretable approach to AD prognosis using end-to-end MRI-based learning. (Github repo link will be provided after review)