Conferences Links
|
|
Paper Review Claims |
| # |
Paper ID |
Title |
Abstract |
Congress |
Claim |
| 1 |
.Cnf-421 |
#14-Smart Transformer Integration in Jordan: A Technological Leap in Power Distribution |
The Solid-State Transformer (SST) represents a transformative innovation in electrical power systems, offering substantial advantages over conventional transformers in terms of efficiency, compactness, and seamless integration with modern smart grids. Leveraging advanced power electronic converters, SSTs enable precise voltage regulation, bidirectional power flow, and enhanced interfacing with renewable energy sources, electric vehicles, and microgrids. These attributes contribute to improved energy efficiency, better power quality, and superior fault tolerance, making SSTs an ideal candidate for modernizing power distribution, especially in regions such as Jordan, where renewable integration and energy reform are rapidly advancing. This paper explores the technical principles behind SSTs, their operational benefits, and their compatibility with Jordan’s energy goals. A key contribution lies in reviewing control of SST , which eliminate the need for traditional mechanical tap changers. The study also discusses implementation challenges, expected benefits, and future directions for integrating SSTs into Jordan’s grid. |
|
Claim |
| 2 |
.Cnf-432 |
#15-Empowering Spiral Antennas with Metamaterials: A High-Efficiency Solution for C-Band Systems |
This study introduces a novel spiral antenna performance improvement technique using metamaterials for use within systems in the C-band frequency range. The authors designed a unit cell of four copper spirals supported by a doubly negative Rogers RT/duroid 5880 substrate for an operating frequency range of 6-9.8 GHz. The unit cell was incorporated into a slotted circular patch to create a fully formed antenna structure. The spiral metamaterial antenna maintains improved characteristics, including a minimum reflection of -21.18 dB at a frequency of 6.353 GHz, impedance match, a gain of 5.05 dB, and a maximum efficiency of 86% at a resonant frequency of 6.535 GHz. The proposed design is a high-efficiency, compact design for use within the communication equipment of the C- band that eliminates the conventional shortcomings of spiral antennas, such as large physical size, impedance mismatch, moderate gain, and uncontrollable radiation patterns. |
|
Claim |
| 3 |
.Cnf-433 |
#5-Graphene: An Atom Away from Success – A Comprehensive Survey |
Graphene is often described as a wonder material with the potential to revolutionize science, technology, and environmental solutions. However, more than a decade after its discovery, its transformative promises remain largely unrealized due to persistent challenges in production, scalability, and integration. This paper traces graphene’s journey from scientific breakthrough to stalled innovation, analyzes the gap between its potential and current applications, and examines how emerging designs—such as a graphene-based solar absorber—might offer a more sustainable and realistic path forward. |
|
Claim |
| 4 |
.Cnf-454 |
#16-Machine Learning-Based Forecasting of Fog Events Using Big Meteorological Data: The Case of Amman Civil Airport |
Fog events pose serious challenges to air traffic control, road safety, and daily urban activities, particularly in cities like Amman, where sudden weather shifts are common. This study proposes a two-stage machine learning pipeline to forecast fog events using hourly meteorological data recorded at Amman Civil Airport over a ten-year period (2015–2025). In the first stage, a classification model identifies fog presence using key indicators such as temperature, dew point, humidity, wind speed, and hour of the day. Among several classifiers, the XGBoost model achieved the highest recall score, making it the most effective at detecting fog events. In the second stage, regression models were used to predict visibility distances during fog hours. While the models were evaluated using conventional metrics such as R² and RMSE, we found that using operational tolerance thresholds (±100m and ±500 m) to measure the difference between actual and predicted visibility provided a more practical and informative assessment. The Random Forest regressor consistently outperformed other models in this regard, showing superior accuracy and greater stability across threshold evaluations. Overall, the proposed framework demonstrates strong potential as a decision-support tool for transportation and aviation authorities in Jordan and similar environments. |
|
Claim |
| 5 |
.Cnf-456 |
#17-Evaluation of Machine Learning-Assisted Dental Age Prediction Using Demirjian Method in Jordanian Children |
Dental age (DA) is a reliable indicator of developmental maturity. Demirjian’s method of DA estimation is widely used and easy to apply, but many studies have shown that it overestimates many populations. Machine learning (ML), a subset of artificial intelligence, provides more accurate and efficient DA prediction compared to traditional methods when used for different populations. Due to the lack of studies in Jordan, this study aims to predict the DA of Jordanian children, aged 3 to 16 years, by utilizing gender and the developmental stages of the seven lower left permanent teeth, as described by the Demirjian technique, using different ML algorithms. Methods: A total of 1,038 orthopantomograms from Jordanian children were analyzed. After applying inclusion and exclusion criteria, a final dataset of 1,000 Jordanian children (511 females and 489 males), aged 3 to 15.99 years, was established. The dataset variables included the children's ages and the seven Demirjian developmental stages. DA was calculated using the Demirjian method, and DA estimation using ML-assisted methods was performed with six models (Ridge Regression, Decision Tree Regressor, Gradient Boosting Regressor, Support Vector Regressor, Random Forest Regressor, and K-Neighbors Regressor). The results were evaluated based on accuracy and compared with chronological age (CA) and DA predicted by the conventional Demirjian method. Results—ML-based methods significantly outperformed the conventional Demirjian method in DA prediction accuracy, with no notable difference between the predicted DA and chronological age. Conclusion—ML enhances DA prediction based on the Demirjian developmental stages, surpassing the accuracy of traditional approaches. |
|
Claim |
| 6 |
.Cnf-470 |
#18-Dual-Band Textile Microstrip Antenna with Dual Hexagonal Slot for WLAN and X-Band Wearable Applications |
This paper presents a dual-band textile microstrip patch antenna with a central dual hexagonal slot for compact, wearable wireless communication systems. The antenna is fabricated using ShieldIt Super electro-textile as the conductive material and felt as the dielectric substrate. The substrate has a relative permittivity of εr = 1.44 and a thickness of 3 mm. The ShieldIt Super layer, with a thickness of 0.17 mm, is provided by LessEMF Inc. The antenna features a full ground plane to minimize electromagnetic coupling with the human body. The overall size is 30 × 30 × 0.17 mm3, with a patch dimension of 20 × 20 × mm2. The proposed design operates at two resonant frequencies, 5.9 GHz, and 8.2 GHz, covering WLAN (IEEE 802.11a/ac), intelligent transport systems (ITS), and X-band satellite communication applications. The antenna achieves operating impedance bandwidths of approximately 520 MHz (5.51–6.03 GHz) and 280 MHz (7.98–8.26 GHz). Reflection coefficients (S₁₁) of –27 dB and –13 dB at the respective bands confirm good impedance matching. Peak realized gain of 5.5 dB and 3dB, along with total efficiencies of approximately 60% and 30%, validate the antenna’s suitability for dual-band, body-centric wireless applications. |
|
Claim |
| 7 |
.Cnf-481 |
#20-Pneumonia Prediction Using Machine Learning Techniques |
This Study highlights a recent breakthrough in using Machine Learning and Deep Learning to identify pneumonia cases through Chest X-Ray (CXR) images. Similarly, machine learning has the potential to offer more accurate and consistent diagnosis than conventional methods, since differences in expertise, fatigue, and human error may affect medical professionals. Machine learning carries this weight in particular. 'Pneumonia' and ’Normal’ CXR images are included in the 5,863 samples subjected to this test. However, we employ small groups of 100 images to evaluate the model's generality and robustness for testing and training purposes.... This method enables us to assess the effectiveness of the models in tight data environments and provides an edge in identifying their strengths and weaknesses. |
|
Claim |
| 8 |
.Cnf-498 |
#22-Comprehensive Analysis Regarding Imbalanced Class Distribution Problem: Best Classification Model and Best Feature Selection Technique |
Classification is one of the main tasks in machine learning that aim to classify data into pre-defined groups called classes. To classify data, we need to use classifier and determine the importance of features by using feature selection techniques, but the performance of techniques always varies depending on the type of data. Therefore, we will classify the data, specifically imbalanced data, to determine the best feature selection technique and the best classifier for this type of data in this paper. To achieve these goals, six datasets with different imbalanced class distribution ratio were used, and seven feature selection techniques and six classifiers were applied. According to the results of three evaluation metrics which are F1-score, MCC, the area under the ROC curve; it became clear that the best feature selection technique is Information Gain Ratio and the best classifier is Random Forest. |
|
Claim |
| 9 |
.Cnf-499 |
#23-Skin Cancer Detection Using Hybrid of Deep Learning and Machine Learning |
Cancer in most of its types is a malignant and deadly disease, and skin cancer, or what is known as Melanoma is one of the most widespread types in the world, as the rates of infection and death from it are constantly increasing. The earlier the disease is detected, the greater the possibility of recovery. Therefore, in this research, work was done on three main goals that help give a role to machine learning and deep learning in diagnosing Melanoma. The first is to discover the best image embedder to extract features that classify each image as if it indicates Melanoma or determines whether the tumor is malignant or benign from among six image embedders. The second goal is to determine the best feature selection technique out of seven well-known techniques, and the final goal is to determine the best classifier out of nine popular ones according to two image datasets with respect to several evaluation metrics such as Accuracy, Precision, and Recall. The results revealed that Inception is the best embedder, Relief is the best feature selection technique, and Neural Network is the best classification model.
|
|
Claim |
| 10 |
.Cnf-527 |
#24-Cybersecurity Attacks Detection Using Machine Learning: Ensemble Learning |
With the scientific and technological development, the importance and sensitivity of data and the danger of its leakage have increased, which prompts amateurs or fraud experts to seek to obtain this data or destroy the entire system, either for personal or competitive purposes or even just for fun, through cybersecurity attacks. Over the years, these attacks have become more complex, forcing us to strive to develop defensive methods. The most important step in defending and confronting cybersecurity attacks is to detect them. Therefore, in this research, we sought to use machine learning models, specifical ensemble learning, to determine its capabilities to detect malicious data movement in the network. The main goal of this paper is to determine the best ensemble learning algorithm to detect cybersecurity attacks or determine their type according to three evaluation metrics: Accuracy, F1- score, and MCC. According to the methodology followed and the obtained results, it has been concluded that boosting is better than stacking, and the best boosting algorithm is scikit-learn. |
|
Claim |
| 11 |
.Cnf-531 |
#25-A Sensitive Analysis for Teacher Burnout Identification using Machine Learning Method: Jordan Case Study |
Accurate burnout prediction is vital for research and decision-making, especially as burnout—caused by prolonged emotional, physical, and mental stress—is common among teachers facing challenges like large class sizes, frequent classes, and salary concerns. This study surveyed 500 Jordanian teachers using the Maslach Burnout Inventory and applied a hybrid machine-learning model to predict burnout levels. The model’s performance was evaluated using Accuracy, R2, MAE, and RMSE metrics via Python. The findings highlight the importance of prediction in reducing burnout’s negative impact, enabling data-driven interventions to improve working conditions and reduce turnover. Key teacher factors—Academic Qualification, Marital Status, Age, Years of Experience, and Salary Scale—combined with Maslach Inventory questions, proved essential for accurately predicting burnout levels. |
|
Claim |
| 12 |
.Cnf-550 |
#26-Explainable Deep Transfer Learning for Multi-Class Knee Osteoporosis Diagnosis Using Grad-CAM and DenseNet121 |
Osteoporosis is a chronic bone disease characterized by decreased bone mineral density and is frequently unrecognized until patients suffer debilitating fractures. Early detection—differentiating between normal, osteopenia, and osteoporosis—is crucial for minimizing clinical risk. This study introduces an interpretable deep learning model for classifying knee osteoporosis based on X-ray images. Leveraging transfer learning, this research fine-tuned several pre-trained Convolutional Neural Network (CNN) architectures, with DenseNet121 achieving the highest classification accuracy of 92.50% on the Multi-Class Knee Osteoporosis X-Ray dataset. To enhance interpretability, the proposed work employed Gradient-weighted Class Activation Mapping (Grad-CAM) to visualize the salient regions in the Xray images that contributed most to the model’s predictions. Our approach not only surpasses state-of-the-art benchmarks in classification performance but also enhances model transparency, facilitating its potential deployment in clinical settings. These findings underscore the promise of explainable AI in medical imaging and support the use of our DenseNet121-based model with Grad-CAM as an effective tool for early and interpretable osteoporosis assessment. |
|
Claim |
| 13 |
.Cnf-554 |
#27-From Data to Net Zero : How SkyBreath is Driving Jordan’s Green Economy |
In the face of the accelerating climate crisis, the ability to accurately measure and manage carbon emissions has become fundamental to achieving sustainable economic development. SkyBreath, a digital platform, offers an innovative solution that integrates Artificial Intelligence, Internet of Things, big data analytics, and cybersecurity to transform complex environmental data into precise and real-time predictive insights. The platform enables organizations to track and reduce their carbon footprint with transparency and efficiency by combining live data streams, machine learning, blockchain technology, and automated reporting tools. SkyBreath serves as a practical and scalable model supporting Jordan’s green economy and strengthening its leadership role in the region’s green technology revolution, providing a vital roadmap toward a sustainable and clean future.Keywords— Carbon Footprint, Artificial Intelligence (AI), Internet of Things (IoT) , Green Economy , Net Zero , SkyBreath |
|
Claim |
| 14 |
.Cnf-558 |
#66-EvalUI: An AI-Driven Framework for Structural Assessment of UI Layouts |
While modern interfaces change quickly, traditional accessibility evaluation is still limited by its focus on markup-based analysis. This restricts effective assessments to completed visual implementations. EvalUI is a new computer vision framework which allows thorough UI layout evaluation through visual analysis. The approach combines several new techniques: adaptive thresholding for reliable component detection in different visual conditions, perceptual modeling for WCAG 2.1 compliant contrast assessment, and graph-based spatial analysis that captures both overall layout structures and individual component relationships. The system is gradio-based and uses visual saliency simulation to provide measurable insights into viewing patterns and the effectiveness of visual hierarchy. A thorough evaluation of two different mobile interfaces, a minimalist contact management screen and the informationpacked GOV.UK homepage shows consistent performance across design types. The framework delivers detailed visual diagnostics, such as accessibility overlays and saliency heatmaps, along with standard metrics for contrast compliance (AA/AAA), spatial balance, and hierarchy effectiveness. Importantly, EvalUI works without needing source code, special file formats, or training data. This makes it particularly useful during early design stages when accessibility feedback is most needed but often unavailable. The results highlight the system’s ability to reveal subtle yet significant accessibility problems while offering practical suggestions for improving layout, thus connecting aesthetic design with an inclusive user experience and layout design assessment. |
|
Claim |
| 15 |
.Cnf-563 |
#30-Feasibility and Design of an Off-Grid Solar-Powered EV Charging Station at Al-Huson College, Jordan |
This paper presents a comprehensive feasibility study and preliminary design of an off-grid, solar-powered DC fast charging (DCFC) station at Al-Huson College in Jordan. The project addresses the growing demand for sustainable transportation and leverages Jordan’s high solar energy potential to design a self-sufficient electric vehicle (EV) charging infrastructure. The study integrates a multi-phase methodology involving solar irradiance assessment, load demand estimation, component selection, and system simulation using industry-standard tools. A financial viability assessment and environmental impact analysis are also included. Key outcomes demonstrate the potential of such infrastructure to reduce reliance on fossil fuels, cut greenhouse gas emissions, and provide economic benefits for the institution. The proposed system design utilizes fixed-tilt photovoltaic panels, battery energy storage, and five 160 kW DCFC units, ensuring high efficiency and reliability. This research contributes a scalable model for similar campuses and off-grid facilities aiming to transition to green mobility. |
|
Claim |
| 16 |
.Cnf-565 |
#31-Exploring Embedded Systems Toolchains, Language Safety and Efficient Programming Techniques
|
This paper explores embedded systems toolchains, and safety-critical programming. It covers the role of compilers, cross-compilation, and debugging tools in embedded development. Additionally, the paper examines safety standards such as MISRA C, and common bugs and optimization techniques for memory and performance in embedded systems .Additionally, the paper includes a case study on the application of these concepts in the development of infusion pumps. |
|
Claim |
| 17 |
.Cnf-567 |
#32-Enhancing Hate Speech Detection on Social Media Using Machine Learning Algorithms |
The propagation of hate speech on social networking websites has been a growing concern in recent years, affecting communities and people globally. In this study, we try to recognize and classify hate speech based on machine learning algorithms. We employed two different datasets, which were pre-processed by text cleaning, Down Sampling and feature engineering, and trained a set of classifiers, including Logistic Regression, Naive Bayes, Random Forest, k-Nearest Neighbors (KNN), and Decision Trees. The preprocessing phase involved equalization of the data, removal of duplicate entries, and conversion of text data into numerical forms through methods such as Bag of Words and lemmatization. Following thorough testing with Orange, Logistic Regression performed best, which achieved 0.961 AUC on one dataset and 0.960 on the other. This research demonstrates that machine learning algorithms, and more specifically Logistic Regression, are valuable instruments in hate speech detection, thus playing a role in enhancing safety and inclusivity online. |
|
Claim |
| 18 |
.Cnf-568 |
#33-Prediction of Asteroid Diameter Using Optimization Techniques |
This research explores a robust predictive framework for estimating asteroid diameters using a combination of machine learning models and optimization algorithms. Given the importance of accurately determining asteroid size for planetary defense strategies, we integrate advanced regression algorithms with Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) to enhance performance. The dataset, sourced from NASA's JPL repository, includes multiple orbital and physical parameters. Preprocessing steps included handling missing data, removing outliers, and constructing novel interaction features. Among the models tested, XGBoost optimized using PSO achieved the best results with an R² score of 0.9669. This study demonstrates the efficacy of hybrid AI-optimization techniques in real-world space applications and paves the way for further interdisciplinary research in astroinformatics. |
|
Claim |
| 19 |
.Cnf-582 |
#34-Early Prediction of Epileptic Seizures Using Artificial Intelligence Techniques |
Such sudden and unexpected neurological events have a significant impact on the safety and well-being of patients experiencing epileptic seizures. In order to analyze electroencephalogram (EEG) signals and detect seizures early, we propose a deep learning-based approach in this paper. The CHB-MIT Scalp EEG dataset was used to train the Convolutional Neural Network (CNN) model that we developed and assessed. After being divided into pre-ictal and inter-ictal periods, the EEG signals were normalized and noise-removed. The model outperformed traditional machine learning techniques with 99.19% accuracy, 98.70% sensitivity, and 99.48% specificity. The results show the potential of AI-based methods for real-time seizure prediction and point out important issues for clinical application in the future, including system integration, power consumption, and interpatient variability, The results emphasize the clinical potential of real-time artificial intelligence-driven seizure prediction systems for enhancing clinical outcomes and patient independence. |
|
Claim |
| 20 |
.Cnf-584 |
#35-CardioRisk: A Machine Learning-Based Approach to Heart Disease Risk Prediction and Diagnosis |
This research aims to address the issue of cardiovascular diseases in an AI manner using machine learning models for the early prediction of heart disease to both decrease it’s mortality rate and reduce the economical weight it puts on patients and hospitals. At first we will collected the data and used preprocessing techniques to properly set up the data to be used in machine learning models using techniques like (Random Forest Imputer, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), One-Hot Encoder, Min-Max Scalar) then we used multiple machine learning models like (Gradient Boosting, Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), Decision Tree, Random Forest, K-nearest neighbor (KNN), Linear Regression, ) after that we used the most suitable evaluation metrics like (Accuracy, Percision, Recall, F-1 score, Receiver Operating Characteristic(ROC) and Area Under the Curve(AUC) score) to identify the model with the best results. Our best-performing model, XGBoost, achieved an F1-score of 0.960, demonstrating improved predictive performance over similar models reported in previous studies, which achieved lower precision and recall due to less rigorous preprocessing. This demonstrates the effectiveness of our preprocessing pipeline and model selection in enhancing predictive performance. We hope that by the end of the research we would have contributed to the early detection of heart disease increasing patient outcomes through early interventions and decreasing the economical toll on patients and health sectors by decreasing the need for severe medical care for patients. |
|
Claim |
| 21 |
.Cnf-593 |
#36-AraGrader Automated Arabic Short Answer Grading System using NLP and AraBERT Transformer |
The AraGrader project describes an automated Arabic short-answer grading model that uses Natural Language Processing (NLP) and the AraBERT transformer concept. This model is intended to solve the complexity of Arabic’s diverse morphology, syntax, and dialectal variances, improving efficiency, consistency, and fairness in grading student responses. |
|
Claim |
| 22 |
.Cnf-594 |
#37-Control System Design for Magnetic Levitation Using Classical PID Techniques |
This paper presents the modeling, simulation, and real-time implementation of a magnetic levitation (Maglev) system using a PID controller. The system utilizes a Hall effect sensor for position feedback and an electromagnet driven by an IRF520N MOSFET module, all integrated through an Arduino Uno platform. A nonlinear dynamic model was developed and linearized around an operating point to facilitate controller design. The PID controller was tuned using the Ziegler–Nichols method and implemented in real time via MATLAB/Simulink. Experimental results demonstrated the effectiveness of the control strategy in maintaining stable levitation with minimal steadystate error and fast settling time, confirming the viability of the low-cost embedded control approach. |
|
Claim |
| 23 |
.Cnf-622 |
#38-Intracranial Hemorrhage Detection In CT Scans Using Deep Learning |
Intracranial hemorrhage detection (ICH) is a serious medical condition that necessitates a prompt and exhaustive medical diagnosis. This paper addresses the multilabel classification task of ICH based on the RSNA 2019 brain
computed tomography (CT) dataset. This paper presents a hybrid deep learning approach leveraging a ResNeXt-101 CNN for spatial feature extraction with a bidirectional LSTM network to capture temporal dependencies across slice sequences. The model achieved a log loss of 0.04604, demonstrating strong generalization to unseen data. To enhance model interpretability and mitigate user resistance, XAI techniques were implemented, providing visual explanations of the model’s predictions. Additionally, a webbased application was developed to facilitate seamless clinical adoption and real-time interaction with the AI system. |
|
Claim |
| 24 |
.Cnf-635 |
#39-Sensorless Induction Motor Control-Comparative Study |
Sensorless control of induction motors without the use of encoders has gained significant attention due to its lower cost, improved system robustness, and reduced maintenance requirements. In this study, we investigate and compare two modulation techniques—conventional Pulse Width Modulation (PWM) and advanced Space Vector PWM (SVPWM)—within a sensorless induction motor control framework. Rotor speed and position are estimated using a Proportional Resonant (PR) observer, eliminating the need for any mechanical sensors As explained in [1]. A detailed simulation model is developed in MATLAB/Simulink to analyze motor behavior under identical operating conditions. The results demonstrate that while both methods can achieve acceptable control performance in an encoderless configuration, SVPWM offers superior dynamic response, reduced torque ripple, lower harmonic distortion, and improved voltage utilization compared to standard PWM. The findings highlight the effectiveness of SVPWM when combined with PR-based sensorless estimation in achieving highperformance
control of induction motors without relying on physical feedback devices |
|
Claim |
| 25 |
.Cnf-641 |
#40-Impact of Fault Location on Transient Stability in Power Systems |
Transient stability is a key aspect of power system performance under large disturbances such as faults. The location of a fault on a transmission line significantly influences the system’s ability to maintain synchronism.[1] This paper investigates the effect of fault location on transient stability by analyzing the system behavior during pre-fault, faulted, and post-fault periods[2]. A Single-Machine Infinite Bus (SMIB) system is used to simulate three-phase faults at different positions along the transmission line. The resulting swing curves reveal that faults closer to the generator reduce the critical clearing time and increase the risk of instability |
|
Claim |
| 26 |
.Cnf-669 |
#41-Proposed Approach to Analyzing Quantum Noise using Operant Learning to Detect Invisible Threats Inside Industrial Control Systems |
The security of Industrial Control Systems (ICS) is a cornerstone of our collective efforts to protect cyberspace. In this research, an innovative approach to enhance the security of the industrial control system from modern and invisible cyber threats that traditional approach fail to detect. The proposed approach is based on the process of analyzing quantum noise resulting from industrial control systems while applying active learning algorithms to continuously improve the accuracy of the detection process. This research aim to develop a framework that enhances safety in critical infrastructure systems such as power plants and water networks. Experiments demonstrate the superiority of this approach over a range of conventional approaches in terms of robustness of accuracy and efficiency in detection, contributing and aiding the process of enhancing cyber resilience of industrial systems. |
|
Claim |
| 27 |
.Cnf-889 |
#42-Design and Implementation of a Digitally Controlled AC-DC Synchronous Buck Converter |
This paper presents the design, simulation, and implementation of a digitally controlled synchronous buck-type AC-DC converter using the STM32G491RE microcontroller. The converter steps down an AC input of 110–285 V into a regulated DC output of up to 16.8 V, optimized for charging applications. Digital voltage mode control was implemented using PI and Type-III compensators, modeled in SmartCtrl and verified in PSIM. A custom PCB integrating voltage/current sensors, EMI filtering, and gate drivers was fabricated. The converter maintained less than 3% output ripple under variable conditions, proving the stability and efficiency of the digital control. |
|
Claim |
| 28 |
.Cnf-892 |
#43-Artificial Intelligence in Space Exploration: Enhancing Autonomous Robotics for Navigation and Landing in Extreme Space Environments |
Autonomous planetary landing is one of the toughest challenges in space exploration, especially in unfamiliar and rough terrains where quick, smart decisions are crucial. This research proposes an integrated system that uses Artificial Intelligence AI, terrain sensing, and vision analysis to improve landing navigation accuracy and safely performance. It combines LIDAR-based elevation maps and CCD images, then it is been analyzed by Convolutional Neural Networks (CNNs) to identify hazards and choose safe landing spots. A landing guidance system using an Extended Kalman Filter (EKF) and PID control adjust the landing path in real time and saves fuel. As a result, simulations show a significant improvement in landing error, it dropped from 150m in previous studies to nearly 20m. In addition, fuel efficiency rose from 70% to 85%, and hazard avoidance improved from 60% to 92%. These results show the ability of the proposed system to make smart decisions on its own, reduce the need for human control, and handle complex surfaces. The framework’s flexible design, real-time adaptability and trajectory stability make it a strong candidate for future missions to Mars, asteroids, and beyond. |
|
Claim |
| 29 |
.Cnf-896 |
#44-Analyzing Recurring Patterns in IPv6 Vulnerabilities to Identify Architecturally Weak Points |
This research focuses on analyzing security vulnerabilities associated with the Internet Protocol version 6 (IPv6) protocol, as documented in security databases such as the National Vulnerability Database (NVD) between 2023 and 2025, with the objective of identifying common patterns among them. The study involved collecting and documenting real-world vulnerabilities identified as Common Vulnerabilities and Exposures (CVEs), followed by classification based on the type of vulnerability, their Common Vulnerability Scoring System (CVSS) severity scores, the affected protocol components, and recurring weaknesses. Subsequently, the behavior of these vulnerabilities was examined to uncover repetitive patterns within the IPv6 architecture that are prone to attacks, aiming to highlight these areas for future remediation efforts. |
|
Claim |
| 30 |
.Cnf-905 |
#45-Feature Selection Using EVORA: An Elite Vector-based Optimization Algorithm with Reweighted Adaptation |
The increasing availability of high-dimensional datasets in modern applications poses significant challenges related to computational cost, overfitting, and model interpretability. Feature Selection (FS) is critical in reducing these issues by identifying the most relevant subset of features. Given the combinatorial nature of FS, metaheuristic algorithms have emerged as effective solutions. This study proposes a novel binary population-based metaheuristic, Elite Vector-based Optimization with Reweighted Adaptation (EVORA), explicitly designed for FS tasks. EVORA employs a probability vector to guide the generation of candidate solutions. Iteratively updates this vector by learning from elite-performing individuals, whose contributions are reweighted based on their normalized fitness. Before introducing EVORA’s methodology and theoretical formulation, a comparative review of twelve widely used metaheuristic algorithms for feature selection is conducted. Moreover, EVORA is evaluated on five diverse benchmark datasets: BCW, Sonar, Madelon, RELATHE, and BASEHOCK, using a K-Nearest Neighbors (KNN) classifier as the wrapper model. Experimental results demonstrate that EVORA achieves the highest or secondhighest classification accuracy across all datasets, with standout performance on Madelon (88.77%), BASEHOCK (92.15%), and Sonar (97.94%). Furthermore, EVORA consistently selects
significantly fewer features, achieving up to 86% reduction on high-dimensional datasets while exhibiting strong stability with the lowest standard deviation values. These findings underscore EVORA’s robustness, efficiency, and effectiveness as an FS method, making it a compelling tool for high-dimensional machine learning problems. |
|
Claim |
| 31 |
.Cnf-906 |
#46-Empowering Emotion Recognition with Generative AI: Classifying VAE-Augmented EEG Spectrograms Using Deep Learning |
This work presents a deep learning framework for EEG (Electroencephalogram)-based emotion recognition using spectrograms and generative augmentation. Raw EEG signals from the SEED-V dataset are converted into 2D spectrograms, then augmented using a Variational Autoencoder (VAE) with a 128-dimensional latent space. In addition, we have added a Graphical User Interface (GUI) that helps users adjust VAE parameters such as noise level and number of generations. Augmented data is used to fine-tune a pretrained ResNet50 model, achieving 90% accuracy. Results show that combining VAE-based augmentation with transfer learning significantly improves emotion classification performance from EEG data. |
|
Claim |
| 32 |
.Cnf-914 |
#47-HomoRoboNet: Deep Learning for Recognition of Humans vs Robot Face Images |
This paper presents HomoRoboNet, a deep learning-based system for distinguishing between real human faces and AI-generated (robotic) faces. With the rapid advancement of Generative Adversarial Networks (GANs), artificial images are becoming increasingly realistic, posing significant ethical and security challenges. To overcome such challenges, our approach uses Convolutional Neural Networks (CNNs) to detect subtle distortions and artifacts in AI-generated images that are usually imperceptible to the human eye. The model was trained on two datasets: the CelebA dataset of real human faces and the Thispersondoesnotexist.10k dataset of artificial faces. The experimental results show high classification accuracy, with a test accuracy of 93%. This work contributes to the ongoing efforts to fight digital fraud and misinformation by providing a robust tool for detecting AI-generated conten. |
|
Claim |
| 33 |
.Cnf-938 |
#48-Race Strategy Optimization in Formula One Using Machine Learning and MATLAB Simulation |
Formula One (F1) racing is a mix of athletic precision and technological sophistication. With the rapid advancement of telemetry systems and real-time data communication, machine learning (ML) has become an essential element in optimizing race strategies. This paper investigates the integration of ML techniques with MATLAB-based simulation tools to enhance decision-making during competitive racing scenarios. Specifically, it focuses on predicting tire degradation—a dominant force in pit stop planning—through linear regression models trained on historical race data. The simulation results, tested with performance metrics including Mean Absolute Error (MAE) and R-squared (R²), show that even slight machine learning (ML) models are able to produce amazing tactical benefit by adding on point data and appropriate preprocessing. This approach bridges the gap between theoretical algorithmic models and real motorsport application, and hints at some new race intelligence engineering directions. |
|
Claim |
| 34 |
.Cnf-947 |
#49-Utilization of Sludge for Clean Energy Production and Biodegradable Plastics |
This research paper investigates sustainable methods for converting biological sludge from wastewater treatment plants into renewable energy and biodegradable materials. Separating the sludge and determining which organic components are appropriate for treatment is the first step. The process of anaerobic digestion, in which microorganisms break down organic matter without oxygen and produce methane gas, is being investigated. In order to help the plant become more energy independent and less reliant on fossil fuels, this methane is subsequently used to power fuel cells. Following digestion, the leftover sludge goes through microbial synthesis, where organic matter is transformed into biodegradable plastics, an environmentally friendly substitute for traditional plastics that aids in the fight against plastic pollution worldwide. Composting is also used to improve soil fertility by turning any leftover organic material into high-quality compost. Utilizing sludge from the first and second categories—which contain organic materials that decompose naturally—is the main goal of the study. Energy recovery, waste recycling, and the creation of biodegradable products are all combined in this integrated approach to provide a thorough sludge management plan. Overall, the study emphasizes how this system supports the circular economy, lessens its impact on the environment, and encourages sustainability in wastewater treatment. |
|
Claim |
| 35 |
.Cnf-985 |
#50-Toward Smarter Spacecraft: A Review of AI-Driven Autonomous Landing Systems |
The integration of autonomous systems and onboard artificial intelligence is transforming the landscape of extraterrestrial missions by enhancing landing accuracy and reducing operational risks. This paper presents a comprehensive review of how technologies such as vision-based hazard detection, sensor fusion, edge computing, and onboard AI are addressing the complexities of planetary exploration. Through the evaluation of real-time autonomous decision-making systems—like ALHAT, Φ-Sat, and IPEX—this study highlights the role of autonomy in enabling spacecraft to operate independently in unpredictable environments. Key benefits include improved data efficiency, minimized latency, and higher mission resilience. The paper also discusses the challenges of power constraints, data standardization, and fault detection integration, proposing pathways for future development in autonomous spacecraft architecture. These advancements signify a critical step toward sustainable and intelligent space exploration. |
|
Claim |
| 36 |
.Cnf-1005 |
#51-Real-Time Football Match Analysis: Leveraging YOLO for Enhanced Object Detection and Possession Tracking |
This paper presents a novel, real-time computer vision system for comprehensive football match analysis, focusing on the accurate detection and tracking of players, referees, and the ball. Leveraging advanced YOLO-based models, this research conducted a rigorous comparative evaluation of YOLOv8x and YOLOv12x models, marking the first documented application and performance comparison of YOLOv12x in football analytics. Our findings show YOLOv8x outperforms YOLOv12x in detection accuracy (F1-score), despite YOLOv12x’s higher efficiency in terms of fewer parameters and lower inference time. To enhance ball detection, we integrated a Kalman filter, significantly boosting its F1-score. The system also precisely calculates ball possession based on player-to-ball proximity. Tested on both local Jordanian and diverse global football matches, the proposed system offers substantial potential for improving sports broadcasting and advanced analytics, providing deeper insights into game dynamics and player performance. |
|
Claim |
| 37 |
.Cnf-1024 |
#53-User Engagement Prediction from Large-Scale Web Logs Using Efficient Feature Engineering and Gradient Boosting |
Our study explores a practical approach to predicting user engagement from large-scale web interaction data, using logs provided by the RecSys 2025 Challenge. Faced with the challenge of processing over 150 million records on limited hardware, we developed a memory-efficient, batch-wise processing framework that enabled the extraction of 39 meaningful behavioral features. These features capture key aspects of user activity, including frequency, timing, and variability. Using data from more than 850,000 users, we trained and evaluated Random Forest, XGBoost, and LightGBM models. XGBoost achieved the best performance, with an F1 score of 0.6978 and an accuracy of 0.7423, followed closely by LightGBM. Our findings highlight the effectiveness of carefully engineered temporal features and tree-based models in user engagement prediction, even in resource-constrained environments. This work offers a scalable and accessible solution for large-scale behavioral modeling and provides a strong foundation for incorporating deep learning methods in future research. |
|
Claim |
| 38 |
.Cnf-1027 |
#54-Design of a Probe-Fed UWB Antenna with Reconfigurable Notch Band |
This paper presents the design and analysis of a compact probe-fed ultra-wideband (UWB) monopole antenna with reconfigurable band notch capability. The antenna utilized a hexagonal monopole geometry fabricated on a low-cost FR4 substrate with dimensions of 35 × 35 × 1.6 mm³. To address the impedance-matching challenges associated with thick substrates and probe feeding, the design incorporated a strategically positioned probe feed at the vertex of the hexagon. Performance enhancement is achieved through the integration of rectangular, defective ground structures (DGS), and parasitic elements, which improve the bandwidth and radiation characteristics. The proposed antenna achieves an ultra-wideband response covering 3.3 GHz to 10.74 GHz, representing approximately 100% coverage of the FCC-defined UWB spectrum (3.1–10.6 GHz). A reconfigurable band-notch feature is implemented using a PIN diode-controlled parasitic element, enabling dynamic suppression of interference in the 4.34–4.5 GHz band to mitigate potential conflicts with aeronautical and WLAN IEEE 802.11a systems. The antenna demonstrates stable radiation patterns and efficient spectrum utilization, making it suitable for various UWB applications including wireless communications, radar sensing, and integrated sensing and communication (ISAC) systems. |
|
Claim |
| 39 |
38.Cnf-1210 |
Supervised Neural Network Based Intrusion Detection Systems |
The magnificent problem for today’s computer networks is the internet attacks which is growing rapidly. Therefore, applying security methods to prevent such attacks on computer network is very important. Network attacks are challenging because they are continuously changing their patterns and techniques. Machine learning approaches play vital roles in detecting, and preventing attacks in different types of computer networks. The security is a critical issue whereas the security mechanism may affect the way of detecting, analysing, and preventing attacks. Therefore, security methods must modify their techniques to deal with. We have proposed Supervised Neural Network Based Intrusion Detection System (SNN-based IDS) to address a critical and timely issue in computer networking - the escalating threat of internet attacks. As these attacks continue to evolve, finding effective methods to detect and prevent them is of paramount importance. this paper leverages advanced technology to tackle the complex task of intrusion detection. This demonstrates the practical application of cutting-edge methods in cybersecurity. The paper suggests that the proposed model has the potential to predict and mitigate new types of attacks, even those not present in the training dataset. This feature is crucial in a rapidly evolving threat landscape. Our SNN is trained by KDDCUP’99 dataset which include 14 attack types in the test data, with an overall number of 24 training attack types. Experiments results show that the proposed system is capable to identifying the attacks and classifying them with high accuracy and reliability, regardless of the dataset's nonlinearity, size, or incompleteness. |
GC-Technology 2026 |
Claim |
| 40 |
.Cnf-1029 |
#55-Design and Analysis of a Compact Reconfigurable Hexagonal Ultra-Wideband Antenna with Electronically Controllable Dual Notch Bands |
This paper presents a novel reconfigurable hexagonal ultra-wideband (UWB) antenna with switchable dual-band notch functionality. The proposed antenna was fabricated on an FR4 substrate with dimensions of 35×35×1.6 mm³ and fed through a 50-ohm microstrip line. The antenna incorporates two PIN diodes, which enable dynamic frequency–response control. When both diodes are activated (ON state), the antenna operates across the entire UWB spectrum (3.1–10.6 GHz). Individual diode control creates selective notch bands: the first diode generates a notch at 3.2–3.55 GHz to mitigate interference with WiMAX applications, while the second diode produces a notch at 6.7–7.5 GHz to avoid interference with ITU satellite downlink communications. When both diodes were deactivated (OFF state), dual notch bands were achieved simultaneously. The antenna design utilizes an S-shaped slot in the radiating patch and a U-shaped notch in the ground plane to realize a frequency-notching mechanism. The compact size, reconfigurable nature, and effective interference mitigation make this antenna suitable for modern wireless communication systems that require the coexistence of multiple services. |
|
Claim |
| 41 |
.Cnf-1023 |
#52-TechArt: Bridging Engineering and Visual Storytelling |
This paper outlines TechArt, an project aimed at developing interactive installations that merge engineering principles with visual arts to tell stories about how technology changed. The project seeks to create hands-on activities that highlight how technology affects people, fostering a deeper public understanding of engineering concepts through artistic expression. |
|
Claim |
| 42 |
.Cnf-1065 |
#59-IOTA-Based Dynamic DNS |
Recent research on DNS usability and security has identified ongoing challenges, many of which have been addressed through decentralized approaches. This paper presents a leaderless, peer-to-peer DNS system that replaces traditional domain names with user-generated public keys. Although prior studies have explored the feasibility of peer-to peer DNS in Internet of Things environments, our work offers a practical implementation tailored to the constraints of such systems. Using IOTA’s Fast Probabilistic Consensus algorithm, we address key challenges related to scalability and low transaction throughput. Additionally, our use of the Proof-of-Connection protocol enables a trustless architecture. Our results indicate that a decentralized DNS can operate securely without central authority or complex consensus mech anisms, both of which are often unsuitable for IoT. Simulations show that our Proof-of-Connection protocol achieves IP valida tion in a quick and lightweight fashion. |
|
Claim |
| 43 |
.Cnf-1080 |
#60-Unmanned Traffic Management System (UTMs) for Jordan : A Federated Architecture for Middle Eastern Drones Airspace Integration |
The rapid expansion of Unmanned Aircraft Systems (UAS) across various sectors, including inspection, surveying, monitoring, and delivery, has necessitated the development of a UAS Traffic Management (UTM) system. This paper explores the challenges and opportunities associated with UTM, focusing on its global adoption, key services, architecture, and the integration of communication, hardware, and software components. We highlight the case for Jordan as an example of how UTM can enhance airspace safety, economic growth, and innovation. |
|
Claim |
| 44 |
.Cnf-1081 |
#61-Linguistic Fingerprints: Fusing Stylometric Analysis and BERT for Robust Fake News Detection |
Fake news is a major online problem. While computer models like BERT are good at understanding text meaning, they often ignore writing style, which can be a key clue. This study tested if combining writing style analysis with BERT could improve fake news detection. We created 10 style features, like sentence length and punctuation use, and combined them with BERT's features from a large Kaggle dataset. Our hybrid LightGBM model was very effective, reaching 87.0% accuracy. Most importantly, it correctly identified 89% of fake news articles, showing a high recall for this critical task. In contrast, a BERT-only model performed poorly, which proves our approach works well. This research shows that an author's writing style is a powerful signal. Combining it with text meaning helps create better and more robust tools to fight digital misinformation. |
|
Claim |
| 45 |
.Cnf-1085 |
#62-Multi-Objective Chrnobly Disaster Optimizer |
One of the most challenging optimization problems in the real world is multi-objective optimization. Addressing these problems requires optimization techniques that can effectively handle the trade-off between objectives while ensuring both convergence and diversity in the solution set. Therefore, effective multi-objective algorithms should be developed to solve such problems. The Chornobyl disaster optimizer (CDO) is a new meta-heuristic optimization algorithm that was recently introduced to solve single-objective optimization problems. The algorithm was inspired by the Chernobyl disaster. The algorithm simulates the movement of the harmful particles that are released after the nuclear reactor core explosion in Chornobyl. In comparison with other algorithms, CDO has proven its ability to significantly achieve good outcomes. This motivated us to conduct a study aimed at optimizing the CDO to handle multi-objective optimization problems (MOOP) and achieve better outcomes for these kinds of problems. Specifically, this study introduces an enhanced version of the CDO for addressing multi-objective optimization problems. In the proposed algorithm, the crowding distance was employed to obtain non-dominated solutions in the population, which will maintain the diversity between the obtained solutions. Additionally, to improve the algorithm's exploration ability, an archive has been used for storing non-dominated solutions. To increase the efficiency of the algorithm, a polynomial mutation was applied to maintain the diversity of solutions. Moreover, the sine-cosine algorithm was integrated with the proposed algorithm to enhance the exploration and exploitation strategy. The performance of the suggested method was assessed by applying it on three sets of unconstrained multi-objective benchmark problems, that include ZDT, UF, and DTLZ benchmark problems. The outcomes were compared with four well-known multi-objective algorithms, which are Speed-constrained Multi-Objective Particle Swarm Optimization (SMPSO), Generalized Differential Evolution 3 (GDE3), Non-dominated Sorting Genetic Algorithm II (NSGA-II), and A Multi-objective Evolutionary Algorithm Based on Decomposition (MOEA/D). We utilized the GD, IGD, HV performance metrics to compare the proposed algorithm with other algorithms. The experimental results show that the multi-objective Chornobyl disaster optimizer (MOCDO) has a superior performance compared to other algorithms. |
|
Claim |
| 46 |
.Cnf-1090 |
#63-Astronomy Access for Visually Impaired Individuals |
Astronomy is a field that could be challenging for visually impaired people in both education and professional fields. That is because it relies heavily on visual data and the need to analyze graphs and images. However, technological advancements in artificial intelligence, 3-D printing, and sonification have led to improvements in accessibility. This paper examines existing efforts by organization, and highlights important technologies and methods to provide more accessibility. |
|
Claim |
| 47 |
.Cnf-1054 |
#58-Engineering a Low-Cost Rotary Collector for Enhanced Electrospun Nanofiber Orientation in Biomedical Applications |
The current study demonstrates effective low-cost rotary drum collector system for receiving the electrospun product in a typical biomedical electrospinning process. The main goal in this work was to design a low-cost functional device that could produce oriented nanofibers with potential usage for biomedical purposes. The collector system was made of CFR-PLA and had a drum diameter of 50.7mm and surface area of 185 cm². A 12V DC motor driven drum achieves rotation speeds of up to 12,000 RPM, while precise speed control is easily programmed using an adjustable controller and infrared sensor. The assembly was fixed on a 3D-printed base encouraging accessibility and modularity.
A 2% (w/v) polylactic acid (PLA) solution was prepared in chloroform and electrospun with a homemade syringe pump and a low-cost high voltage power supply (HVPS) under specific flow and voltage parameters. Aluminum foil-assisted conductive surface of the rotating drum facilitated efficient gathering of nanofiber under electric field. A scanning electron microscopy (SEM) examination of the nanofibers prepared under optimal conditions showed that relatively uniform morphology and an average fiber diameter of ~ 587 nm were achieved, which is competitive with the performance of existing systems. This design succeeds to being a financially viable solution while being able to maintain the integrity wetting functionality of electrospinning technology in lab and research settings. |
|
Claim |
| 48 |
.Cnf-1140 |
#64-Cyclodextrin-Based Nanosponges in Drug Delivery: Overcoming Poor Solubility and Bioavailability Challenges |
Nanosponges are innovative, porous nanoparticles emerging as a revolutionary tool in drug delivery. Their unique three-dimensional structure, often built from polymers like cyclodextrin, allows them to encapsulate various substances. A primary advantage of this technology is achieving significant solubility enhancement for poorly soluble drugs, a major challenge that often limits bioavailability. By forming inclusion complexes and increasing the drug's effective surface area, nanosponges can improve dissolution rates, protect therapeutic agents from degradation, and enable controlled release. This paper provides an overview of nanosponge technology, focuses on the mechanisms for solubility enhancement, and discusses the key challenges limiting their application. |
|
Claim |
| 49 |
.Cnf-96 |
#67-Multi-Sensor Fusion for Victim Detection: Integrating Thermal Imaging, UWB Radar, and GPR in Search and Rescue Operation |
This paper investigates the practical applications of thermal imaging, UWB radar, and GPR in the context of SAR operations. It presents an overview of each technology, explores their synergistic potential through sensor fusion, and identifies the technical and operational challenges associated with their deployment. By examining case studies and recent research, this work aims to provide insights into how integrated sensing systems can transform the effectiveness of search and rescue missions in complex and hazardous environments. |
|
Claim |
| 50 |
68.Cnf-1206 |
Design of Array Inverted F Antenna for IOT
|
This work analyzes the performance and designs a rectangular Microstrip patch F antenna. 5.9 GHz is the antenna's resonance frequency range, making it suitable for Internet of Things (IOT) applications. Simulation software for this work was Computer Simulation Technology (CST) software. A rectangular Inverted F coplanar antenna array structure was used in the antenna's design. The bandwidth, gain, and return on loss of these antennas were evaluated to determine their respective performances. The main findings of this study shown that, in comparison to a conventional antenna, the optimized array-shaped antenna increased bandwidth, gain, and return on loss. Furthermore, the improved antenna attained an operating frequency of 5.9 GHz, making it appropriate for Internet of Things applications. |
GC-ElecEng 2026 |
Claim |
| 51 |
68.Cnf-1208 |
A Blockchain Framework for Enhancing Selection of Optimal Parameters at NB-IOT |
By 2030, wireless communications will have connected over thirty billion devices. Narrowband Internet of Things (NB-IOT) technology has grown in popularity in response to the rapid growth of the internet of things (IOT) sector. The main aim of this study is to supply overall survey of the design modifications transported in the NB-IoT standardization along with comprehensive study evolutions according to popular companies in many countries such as: Telia, Elisa, Orange, Telecom Italia, Telstra, Vodafone, On the other hand, because there is a lot of work in NB-IOT on optimizing parameters or improving optimization methods, Consequently, this work presents a Blockchain architecture that may be used to enhance security, authentication, and efficient data access while maintaining data integrity. We describe the optimization parameters for physical channel and signal transmission and reception. Using the Physical Downlink Control Channel, we devise an adaptation scheme for 200 KHz bandwidth in NB-IOT networks (PDCCH). Finally, we want to locate the following: optimal parameters: Number of frames, Doppler frequency and Diversity performance. |
GC-ElecEng 2026 |
Claim |
| 52 |
.Cnf-100 |
#68-Polarization Reconfigurable Microstrip Patch Antenna for Wireless Communications |
A polarization reconfigurable patch Antenna is proposed in this paper, the design consists of two rectangular patch antennas fed by a four-way power divider, the polarization diversity was achieved using PIN diodes, a diode was placed at the end terminal of each power divider connecting the antenna, the switching states can be achieved by changing the biasing of the external DC circuit, four different polarization states were achieved (RHCP, LHCP with two different phase shifts for each polarization), various characteristics of the designed antenna were shown such as Reflection coefficient, Axial Ratio, Gain and Antenna efficiency the antenna is designed to operate at a frequency of 4.7 GHz |
|
Claim |
| 53 |
.Cnf-102 |
#69-Machine Learning-Based Classification of Plasma Cell Neoplasms from SEER Dataset |
This study investigates the classification of plasma cell neoplasms using machine learning methods on SEER registry data. Focusing on the three most frequent subtypes—plasmacytoma, multiple myeloma, and extramedullary plasmacytoma—the research builds predictive models using Random Forest, XGBoost, Support Vector Machine, and Logistic Regression. The highest performance was achieved by Random Forest and XGBoost, each attaining over 97 percent accuracy. Important features such as tumor location, diagnostic year, surgical history, and sex played a significant role in classification. SHAP analysis further revealed the impact of each variable on model output, highlighting the potential of machine learning in clinical decision support. |
|
Claim |
| 54 |
.Cnf-110 |
#70-Adaptive Neural Control of a Two degrees of Freedom Robotic Arm |
This paper proposes a hybrid online adaptive Radial basis functions network (RBFNN) to control the joint angles of a two degrees of freedom robotic arm. The method is also compared with the base line proportional integral derivative (PID) controller. The results show that the adaptive neural controller achieves better tracking than the PID controller with less overshoot and steady-state error. |
|
Claim |
| 55 |
.Cnf-115 |
#71-A Compact Metamaterial-Inspired Unit Cell for Enhanced C-Band Antenna Response |
In this paper, we present a new microstrip patch antenna design, which utilizes double negative (DNG) metamaterial unit cells to improve their performance. The proposed antenna consists of a rectangular patch on an FR-4 substrate with dimensions of 30 mm × 30 mm × 1.6 mm. Four identical unit cells with negative permittivity and permeability at 5.88 GHz are added to the antenna structure to improve performance. The effective permittivity and permeability of the metamaterial were extracted to confirm its double negative behavior. Simulation results show improved performance in terms of reflection, impedance matching, radiation efficiency, and directional gain due to the inclusion of the metamaterial. This antenna operates in the C-band, which makes it suitable for applications such as 5G wireless communication and satellite systems that require compact size and high efficiency at microwave frequencies. |
|
Claim |
| 56 |
.Cnf-117 |
#72-Real-Time Violence Detection for In-Vehicle Safety |
Public transportation faces tremendous safety challenges, including harassment and conflict between passengers and drivers. These problems put lives in danger and discourage people from using public transportation. Several studies and pilot projects have explored real-time surveillance and behavior analysis to mitigate these issues. The proposed approach aims to minimize the consequences caused by these actions, by integrating an intelligent monitoring system that predicts such events in real-time. Utilizing the revolution of Machine Learning (ML) and Deep Learning (DL) models to detect, predict, and respond to potential fights between people inside vehicles. Once a critical occurrence is detected, the system can issue immediate alerts to relevant parties such as law enforcement, while also storing event data for documentation and model improvements. The initial testing of the designed system has shown an overall accuracy of approximately 97% in correctly identifying various types of violence actions compared to other existing works. This solution improves safety in shared transportation environments. |
|
Claim |
| 57 |
.Cnf-120 |
#73-Crises and Consequences on Water Resources from the Environmental and Social Impacts of Traditional and Renewable Power Plants: Consumption And Pollution |
This paper examines the impacts of energy production on water resources, whether from conventional or renewable sources, with a focus on water consumption and pollution. The study highlights the adoption of sustainable energy practices and the improvement in water management, considering societal participation `and impacts. |
|
Claim |
| 58 |
.Cnf-122 |
#74-Design and Development of Al-Rhim Sat A 1U CubeSat Jordanian Platform for Wildlife Tracking Using LoRa-Based IoT Technology |
Several native animals from several species are risking extinction in Jordan due to hunting, trade and collection of wildlife. The Royal Society for the Conservation of Nature (RSCN) encounters challenges in monitoring and regulating conservatories. This work presents the design of a 1U CubeSat mission supporting wildlife conservation in Jordan through space-based LoRa communication. The satellite enables longrange, low-power data collection from LoRa-enabled ground sensor nodes deployed in remote reserves where terrestrial networks are unavailable. We detail the end-to-end system architecture, including the CubeSat platform and custom LoRa ground stations designed for local assembly by students. The mission aims to validate LoRa performance in low Earth orbit and provide a scalable, low-cost solution for wildlife monitoring |
|
Claim |
| 59 |
.Cnf-123 |
#75-Optimized design of 32-bit Pipelined MIPS based on Verilog implementation |
This report presents the design, implementation, and verification of a five-stage, 32-bit pipelined processor conforming to the MIPS instruction set architecture (ISA). By dividing instruction execution into distinct stages—Instruction Fetch (IF), Instruction Decode (ID), Execute (EX), Memory Access (MEM), and Write-Back (WB)—we achieve instruction-level parallelism that significantly increases throughput. Key features include a hazard detection unit that stalls the pipeline on load-use hazards, a forwarding unit resolving most data dependencies, and support for control-flow instructions (branches and jumps), including the jr instruction. Simulation in ModelSim confirms correct functional behavior across arithmetic, logical, memory, and control instructions. Performance analysis indicates an average CPI of approximately 1.2 on benchmark programs, compared to CPI≈1.0 for an ideal pipeline and CPI≈1.8 for a non-pipelined design. |
|
Claim |
| 60 |
.Cnf-128 |
#76-On-Road Real-Time Animal Detection System |
Animal-vehicle collisions (AVCs) pose serious safety and ecological risks. This paper introduces a cost-effective, real-time animal detection system utilizing YOLOv5 on Raspberry Pi 5. It immediately alerts drivers locally and transmits detection data to AWS IoT Core to inform nearby vehicles within a 1 km radius. Field tests demonstrated an 83% mean Average Precision (mAP) and latency consistently below 1.2 seconds, validating the system’s practicality. |
|
Claim |
| 61 |
.Cnf-320 |
#1-Steganography as an Anti-Forensics Technique |
Steganography is a powerful anti-forensics technique that involves hiding information within digital media, such as images, in a way that conceals the existence of the data itself. This paper explores a range of image-based steganographic methods, from basic techniques like Least Significant Bit (LSB) insertion to more advanced approaches including Discrete Cosine Transform (DCT)-based embedding, adaptive steganography, and deep learning-based models. The study highlights how these methods function, their levels of detectability, and the tools used to implement them. It also discusses the forensic detection methods used to uncover such hidden data, emphasizing the ongoing battle between steganography developers and digital forensic analysts. By examining both the implementation and detection of steganography, this paper provides a comprehensive view of its role in digital antiforensics and its implications for cybersecurity and forensic investigation. |
|
Claim |
| 62 |
.Cnf-322 |
#65-Engineering Sentient UI: Bridging Emotion, Behavior, and Context for Dynamic UI Adaptation |
Despite increasing interactivity in digital systems, most user interfaces remain fundamentally static, unable to reflect the emotional and behavioral richness of human interaction. This disconnect impairs usability, limits accessibility, and fails to support truly personalized experiences. While research in affective computing and adaptive systems has outlined promising theoretical models, few practical implementations have bridged emotion, behavior, and context in real-time interfaces. This paper presents Sentient UI, a modular adaptive UI framework engineered in Flutter that dynamically adjusts interface components based on emotion recognition, behavioral patterns, and environmental context. It operates entirely offline, preserving user privacy while leveraging clean architectural design to
ensure extensibility and developer control through configurable adaptation layers. Sentient UI is being developed as part of a software engineering graduation project and is scheduled for open-source release in February 2026. By turning complex multi-modal sensing into accessible APIs, it offers a scalable foundation for building emotion-aware, real-time adaptive interfaces across everyday applications. |
|
Claim |
| 63 |
.Cnf-324 |
#3-Hyperparameter optimization (HPO) using Evolutionary Computation Algorithms (ECAs) |
Due to the numerous applications of artificial intelligence and Machine Learning (ML) algorithms in different applications, it becomes very necessary to introduce an optimization technique for these algorithms. ML models execution time, performance metrics and storage consumption are the main goals of these optimization techniques. Hyperparameter Optimization (HPO) algorithms are considered a promising performance enhancement process. This paper introduced two applications of ML models (Trajectory prediction model, Website fingerprinting attack generation and Intrusion Detection System), implementing Genetic algorithm (GA) as a HPO algorithm to enhance the performance of the introduced models. Two different datasets were evaluated (CitySim dataset and IP Network Traffic Flows dataset). A decision tree model provided the best accuracy in the trajectory prediction application, with 98.55% and an execution time reduction of 53% due to the impact of GA. For the Website fingerprinting attack generation model, a random forest model performs the best in terms of accuracy (97.94%) with an execution time reduction of 80.8% due to the impact of GA. Finally, a QDA model outperforms the other ML models in the IDS application with 99.53% accuracy and 73% execution time reduction due to the impact of GA. |
|
Claim |
| 64 |
.Cnf-331 |
#8-Attention Deficit Hyperactivity Disorder Detection Using Machine Learning Techniques On Reduced Datasets |
Attention Deficit Hyperactivity Disorder (ADHD) is commonly identified during childhood and frequently persists into adulthood. Children with ADHD may find it challenging to focus, control impulsive actions (such as acting without thinking about the outcomes), or may exhibit excessive activity. There are three main types of ADHD: Inattentive presentation, hyperactive/impulsive presentation, and combined presentation. in this paper, two main objectives are considered. The first is to identify the best classifier that can detect ADHD among twenty-one different classifiers that represent six learning strategies. The second objective is to determine the most effective feature selection method for datasets related to ADHD. The analysis highlighted the effectiveness of SMO, Multilayer Perceptron, in detecting ADHD. Additionally, SymmetricalUncertAttributeEval and Principal Component Analysis methods demonstrated the best performance among the ten considered feature selection techniques evaluated in this paper. |
|
Claim |
| 65 |
.Cnf-332 |
#29-Design and implementation of a real-time full digital substation using RTDS simulator and protection relay |
Real-time simulation technologies are playing an increasing role in power system studies, driving the adoption of Hardware-in-the-Loop (HIL) methodologies for protection testing and validation. In addition, the new generation of the relays (Intelligent Electronic Devices - IEDs) support communication through the IEC61850 protocol, the state-of-the-art digital substation protocol. This research aims to explore the potential of using the IEDs for advanced research by connecting the IEDs to the Real-time digital simulator (RTDS), and to investigate the importance of this scheme for academic research, and testing the protection and control techniques before applying them to real-life applications, for optimization and risk minimization. In this research the RTDS device was connected to the ABB REX640 IED, using IEC61850. The research illustrates the steps and the requirements of connecting the IED to the simulator, in a Full digital substation scheme, where the IED receives the voltage and current values data and the breaker position from the RTDS. Also, the RTDS receives the trip signal from the IED via GOOSE communication. The results demonstrate an accurate, valid and stable connection between the RTDS and the IED. The simulation was built to measure the communication time using instantaneous over current relay function in the REX640 IED. Where the communication time for detection and isolating a fault was less than 6 milliseconds from the point when the fault occurred until the breaker opened to isolate the fault. |
|
Claim |
| 66 |
.Cnf-343 |
#9-A Study on Neural Network-Based Cell Balancing for Lithium-Ion Batteries |
In multi-cell battery systems, charge imbalances can arise naturally from differences in manufacturing, aging, and usage conditions. To address these challenges, efficient cell balancing is essential, as it helps equalize the charge and discharge rates across all cells. This prevents instances of overcharging or deep discharging in individual cells, ultimately improving performance, boosting reliability, and extending the battery's lifespan. This research investigates the application of machine learning (ML) models to enhance the balancing in batteries. This study explores how machine learning (ML) models can be utilized to improve charge balancing in battery systems. Conventional algorithms often face limitations in terms of time and efficiency.
This research tackles these challenges by training machine learning (ML) models to optimize the selection of balancing resistors, factoring in the degree of imbalance to enhance balancing time. Feedforward Neural Networks (FFNN) algorithm are employed to develop models that mimic and improve upon conventional methods on three-cells configuration. The model’s performance was evaluated using the F1 score, yielding results of approximately 0.99 for both the training and testing datasets across all targets. Thereafter, the model was implemented in MATLAB Simulink. The results showcased the model's capability to utilize its learned knowledge efficiently, achieving notable improvements in balancing time compared to traditional methods. |
|
Claim |
| 67 |
.Cnf-361 |
#10-Enhancing Virtual Screening for Cancer Therapeutics: A Machine Learning Approach for HDAC1 Inhibition Prediction |
Accurate prediction of compound bioactivity is a key component in accelerating drug discovery, particularly for high-value targets such as Histone Deacetylase 1 (HDAC1). As a critical epigenetic
regulator, HDAC1 is implicated in the progression of several cancers, making it a promising therapeutic target. However, traditional compound screening is often costly and time-intensive. This study presents a machine learning–based virtual screening framework designed to predict the inhibitory activity of compounds against HDAC1. A dataset of approximately 9,000 compounds was retrieved from the ChEMBL database, with 8,804 molecular descriptors generated using PaDEL-Descriptor. Bioactivity data, represented by IC50 values, were discretized into three classes:
active, intermediate, and inactive. Using the PyCaret framework, we developed and evaluated several supervised learning models, including Light Gradient Boosting Machine (LightGBM), XGBoost, AdaBoost, Decision Tree, and Random Forest. Among them, the Gradient Boosting Classifier achieved the highest performance, with an accuracy of 93.13% and a Cohen’s kappa score of 0.8654. These results highlight the ability of ensemble models to capture complex structure–activity relationships. The proposed pipeline is scalable, reproducible, and efficient, offering a practical solution for early-stage virtual screening and supporting the rapid identification of HDAC1-targeted drug candidates. |
|
Claim |
| 68 |
.Cnf-400 |
#11-Enhancing DNS Security in Public Wi-Fi Using Modern DNS Protocols |
The Domain Name System (DNS) serves as a core element of internet infrastructure, enabling the translation of human-readable domain names into machine-readable IP addresses. Despite its
critical role, DNS was not originally designed with security in mind, making it susceptible to various threats, especially in untrusted environments such as public Wi-Fi networks. To address these vulnerabilities, several modern security protocols have been developed to enhance DNS privacy and integrity. These include Domain Name System Security Extensions (DNSSEC), which ensure data authenticity through digital signatures; DNS over HTTPS (DoH) and DNS over TLS (DoT), which encrypt DNS queries to prevent interception; DNSCrypt, which secures communication between clients and resolvers; and Response Policy Zones (RPZ), which apply custom filtering policies to block malicious domains. The integration of these protocols forms the basis of this study, which aims to evaluate their individual and combined effectiveness in mitigating common DNS-based attacks such as spoofing, content manipulation, and eavesdropping. The study is motivated by the growing need to secure user privacy and data integrity in open network environments and seeks to provide a conceptual framework for deploying these protocols efficiently while offering recommendations for broader implementation and future policy development. |
|
Claim |
| 69 |
.Cnf-410 |
#12-Double Negative Reconfigurable Multilayered Imaging Metamaterial Antenna |
This paper presents a new, multilevel, dynamic feed-direction-controllable, double-polarized, C-band operation metamaterial antenna, i.e., to mention one example here, for bone fracture diagnostic application. It is constructed using PIN diodes to ensure dynamic feed directionality, copper walls, and levels of image gain enhancement and accuracy control. The novelty encompasses the hexagonal resonator type, square spiral shape metamaterial unit cell, utilized in the rotated 45-degree 2⨉2 lattice to achieve independent polarization operation. Simulation studies ensured double-negative behavior, polarization controllability, and proper enhancement of the gain. It is found that the whole antennal system presents a promising low-cost, portable, and emission-free approach relative to conventional approaches to bone fracture diagnostics, with the potential of enhanced patient access and outcome, to be utilized in medical imaging applications. |
|
Claim |
| 70 |
.Cnf-411 |
#13-Impact of Electric Vehicle Charging Penetration on Power Grid Stability Using PSS®E Simulation |
This project investigates the impact of increasing electric vehicle (EV) charging loads on the stability of a simplified power grid using PSS®E software. Four EV penetration scenarios (0%, 30%, 60%, and 100%) were simulated on a 9-bus system to evaluate voltage deviation and system behavior. The results demonstrate a clear decline in bus voltages as EV loads increase, especially at buses directly connected to EV chargers. The findings highlight the importance of load management and infrastructure reinforcement to maintain voltage stability under high EV demand. |
|
Claim |
| 71 |
46.Cnf-1207 |
Design and analysis of off -grid PV/diesel system for small scale factory located in Wadi Rum, Jordan using Homer software |
The fuel prices are increasing nowadays, causing a burden on the power system. From this, diesel generators that are still in use, must be hybridized with renewable energy (RE) to levelized the overall projects costs. This study aims to schedule an operation of a hybrid system photovoltaic (PV)/ diesel in a remote area which located in Wadi Rum district in Jordan to serve factory load in addition to obtain the best economical system combination that can serve the needs of a continuous electrical load without any interruption. Based on the results of the analysis using HOMER software, the configuration in the system will be a combination between PV modules, and diesel generator. The hybrid system has met the target for electricity production which is documented in upcoming sections. |
GC-ElecEng 2026 |
Claim |
|
|
