Bulletin of Electrical Engineering and Informatics
Not a member yet
2885 research outputs found
Sort by
Enhancing skin cancer detection using transfer learning and AdaBoost: a deep learning approach
Get PDFSkin cancer is one of the most prevalent types of cancer worldwide, with early detection playing a critical role in improving patient outcomes. In this study, we propose a deep learning model based on LeNet-7 combined with adaptive boosting (AdaBoost) to classify skin lesions as either benign or malignant using the International Skin Imaging Collaboration (ISIC) dataset. We evaluate the proposed model alongside other well-established deep learning architectures, such as residual network (ResNet), VGGNet, and the traditional LeNet model, through various performance metrics including precision, recall, F1-score, specificity, Matthew’s correlation coefficient (MCC), area under the receiver operating characteristic curve (AUC-ROC), and testing accuracy. Our results demonstrate that the proposed model (LeNet-7+AdaBoost) significantly outperforms the other models, achieving a testing accuracy of 91.3%, precision of 0.92, recall of 0.91, and AUC-ROC of 0.93. The model successfully addresses issues of overfitting and generalization, providing a robust solution for skin cancer classification. However, some misclassifications of visually similar benign and malignant lesions highlight areas for future improvement. The proposed model shows promise in real-world medical applications and paves the way for further research into optimizing deep learning models for skin cancer detection
Modeling the process of magma rising in the bowels of the Earth and its eruption to the surface
Get PDFThis paper presents a numerical model for simulating magma ascent in the Earth’s interior and its eruption to the surface, aimed at improving earthquake prediction. Magmatic flows are modeled as highly viscous fluids with a low Reynolds number (Re) using simplified Navier–Stokes equations. The approach incorporates hydrodynamic instability arising from density differences between magmatic and asthenospheric layers. Initial and boundary conditions were formulated for magma outflow from a narrow crack, and a dimensionless Euler–Reynolds (ER) parameter was introduced to characterize flow behavior. Numerical experiments for different ER values revealed that at low ER, magma spreads slowly, forming stable layers, while higher ER values accelerate vertical rise, increase pressure gradients, and enhance instability. The model identifies zones of stress accumulation that may precede seismic events. An additional method—monitoring fluid levels in deep wells—showed correlation with seismic fluctuations, supporting its potential for early warning. The results confirm the reliability of the proposed approach, demonstrating good agreement with seismological data. The developed methodology can be applied to enhance early warning systems and reduce risks in seismically active regions
Artificial neural network maximum power point tracking for mitigation photovoltaic harmonic distortion
Get PDFThis study introduces a novel methodology aimed at minimising total harmonic distortion (THD) in grid-connected photovoltaic (PV) systems (GCPVs) through the implementation of a maximum power point tracking (MPPT) approach based on artificial neural networks (ANN). High THD levels in PV systems can lead to inefficiencies, power quality issues, and potential damage to the grid infrastructure. Although traditional MPPT methods effectively optimise the power output, they often fail to address harmonics. The proposed ANN-based MPPT algorithm improves PV power harvesting while actively minimising the harmonic distortions. The ANN was trained using a comprehensive dataset that included various environmental conditions, ensuring robust performance in diverse operational scenarios. Simulation results demonstrate that the ANN-based MPPT approach significantly reduces THD to below 1% across various irradiance levels, in contrast to the 1.18% to 2.72% observed with conventional methods such as perturb and observe (PO), while simultaneously preserving optimal power output. Reducing harmonic distortion improves the power quality, system efficiency, and lifespan of grid-connected components. This study highlights ANN-based control strategies for addressing the challenge of maximising energy harvesting and maintaining power quality in modern PV systems, offering a solution for the sustainable integration of solar energy into the grid
Evaluating digital competency among statistical educators: a comparative analysis of input-oriented DEA models
Get PDFAs the educational landscape shifts towards online learning, assessing educators' digital competencies has become crucial. This study aims to evaluate the digital competencies of university educators using data envelopment analysis (DEA), specifically comparing the banker, charnes, and cooper (BCC) input-oriented models (super efficiency and Bi-O multi-criteria data envelopment analysis (MCDEA) super efficiency BCC models). The research was conducted in three phases. Initially, the BCC model assessed educators' digital competencies. Subsequently, the Bi-O MCDEA model evaluated these competencies within an online learning context. Finally, the effectiveness of the two models was compared. Data was collected through a survey administered to 30 educators from Universiti Teknologi MARA, with a response rate of 75%. Results showed that while the BCC model identified 23 out of 30 educators as efficient, the Bi-O MCDEA model recognized only two as efficient. This discrepancy highlights the different stringencies of the models and their impact on assessing digital competencies. The super efficiency (SE) model was then used to rank the efficient educators to determine the most proficient. The study underscores the need for precise assessment tools in online education to enhance digital competencies effectively. It suggests that integrating advanced DEA models can significantly improve the identification and training of educators, thereby enriching the educational outcomes in digital environments
Advancements in machine learning techniques for precise detection and classification of lung cancer
Get PDFLung cancer remains one of the most prevalent and lethal malignancies worldwide, necessitating early detection and accurate classification for effective treatment. In this work, we present a unique machine learning (ML) model that uses medical imaging data to detect and classify lung cancer. Utilizing a dataset of 613 images which obtained from Kaggle, our model combines sophisticated feature extraction methods with three essential algorithms: AdaBoost, stochastic gradient descent (SGD), and random forest (RF). Orange3 data mining software was used to classify the model after it was preprocessed and features were extracted using MATLAB. Nonetheless, the model showed good performance in identifying lung cancer lesions in four different categories: squamous cell carcinoma, big cell carcinoma, adenocarcinoma, and normal. With an accuracy of 0.998 and an AUC range of 1.000, AdaBoost notably produced the best results. Overall, ensemble ML techniques demonstrated notable benefits over single classifiers, indicating its potential to aid in the creation of accurate instruments for the diagnosis of lung cancer in its early stages
Advances in artificial intelligence-driven 3D model generation: a review of GAN and VAE methodologies
Get PDFThis paper offers a comprehensive review of current developments in artificial intelligence (AI)-based 3D model creation, with an emphasis on techniques utilizing variational autoencoders (VAEs) and generative adversarial networks (GANs). 3DGAN, paired 3D model generation with GAN, conditional GAN, FaceVAE, voxel-based 3D object reconstruction, and 3D-VAE-SDFRaGAN are the six main techniques that are studied in this work. Each method is discussed, highlighting its architectural framework, data representation, and specific approach to generating 3D models. First, the paper introduces basic terms and classical 3D modeling techniques and provides a comparative analysis of them based on their workflow, purpose and field of application. In subsequent chapters, methods for generating 3D models based on the use of GANs and VAEs are reviewed, describing its methodology, experimentation technique, results, and comparison with other methods. The review outlines the strengths and limitations of each approach and their applications in object reconstruction, shape generation, and maintaining model consistency. It concludes by emphasizing how AI-driven methods can advance 3D modeling, underscoring the need for further research to enhance quality, control, and training reliability. The findings show AI’s significant impact on automating complex modeling tasks and enabling new creative opportunities in 3D content development
Integrating RPA, BPM, and DT in the context of Industry 4.0 and 5.0: a strategic approach for modern enterprises
Get PDFThis paper examines the interaction between robotic process automation (RPA), business process management (BPM), and digital transformation (DT)-three critical components in improving operational efficiency and driving business modernization. RPA automates repetitive tasks, reduces errors, accelerates processing, and optimizes resource use. When combined with artificial intelligence (AI) and machine learning (ML), it further enhances data analysis and decision-making. BPM focuses on analyzing, designing, and optimizing business processes to ensure organizational agility. DT provides a technological foundation for broader innovation in processes and structures. The paper contributes a comprehensive and updated perspective on how RPA, BPM, and DT interrelate—not only functioning independently but also reinforcing one another to create greater business value. It emphasizes that their integration is a strategic approach to improving performance, responsiveness, and continuous innovation. Importantly, the research is relevant to both Industry 4.0 and Industry 5.0. While Industry 4.0 (I4.0) prioritizes automation and data-driven systems, Industry 5.0 (I5.0) highlights human–technology collaboration for more adaptive and human-centric organizations. This study enriches theoretical insights and offers practical guidance for building effective and sustainable DT strategies
Brain tumor classification using PCA-NGIST features with an enhanced RELM classifier
Get PDFBrain tumours may cause severe health risks because of abnormal cell growth, which may result in organ malfunctions and death in adulthood. As precise identification of the tumour type is required for effective treatment. Magnetic resonance imaging (MRI) has recently been provided as an effective method for brain tumour diagnosis by computer-based based systems. To categorize brain tumours from MRI images, the paper offered a fusion model integrating an enhanced regularized extreme learning machine (RELM) classifier with principal component analysis (PCA) and normalized GIST (NGIST) feature extraction. While NGIST extracts strong spatial and texture features essential for modelling the tumour, PCA reduces the dimension of the input features without sacrificing significant data patterns. The improved RELM efficiently categorizes brain tumours into three categories: pituitary, meningioma, and glioma. It is optimized to improve learning capacity and generalization. The novelty of this study lies in the integration of NGIST descriptors with PCA-driven dimensionality reduction and an enhanced RELM classifier in a single lightweight framework. Unlike conventional methods that trade accuracy for computational cost, the proposed model ensures high precision and recall while remaining computationally efficient. This unique fusion demonstrates significant improvements in both diagnostic accuracy of 96% and clinical applicability, offering a balanced solution for real-time brain tumor classification
Handwritten Kaganga script classification using deep learning and image fusion
Get PDFClassification of traditional handwriting script and to preserve many cultures have been developed in some parts of the world, including image classification of handwriting Kaganga script. This study aims to propose a new combination model by implementing top-hat transform (THT) and contrast-limited adaptive histogram equalization (CLAHE) with discrete wavelet transform (DWT) to support the performance of the convolutional neural network (CNN) in Kaganga script classification. The top-hat transform and contrast-limited adaptive histogram equalization with discrete wavelet transform Fusion L2 convolutional neural network (DWT-THCL L2 CNN) models get the best accuracy from the CNN with L1 regularization, CNN with dropout regularization, CNN with L2 regularization and CNN with L2 regularization and CLAHE models. Based on the experimental results, the DWT-THCL L2 CNN model successfully increased training accuracy by 7.76%, validation accuracy by 5.11%, and testing accuracy by 3.73% from the CNN L1 model. The DWT-THCL L2 CNN model received a training accuracy of 99.87%, validation accuracy of 82.61%, and testing accuracy of 82.61%, while the CNN model with L1 regularization (L1 CNN) only received a training accuracy of 92.11%, validation accuracy of 77.50%, and testing accuracy of 78.88%
Linear and nonlinear control design for a quadrotor
Get PDFIn the current study, the quadrotor's nonlinear dynamic model is developed using the Newton-Euler approach. Following that, several nonlinear and linear control strategies for tracking the quadrotor's trajectory are applied. First, by employing distinct controllers for each output variable, direct application of the linear proportional integral derivative (PID) controller to the nonlinear system is realized. This system may also be linearized about an operational point to generate linear controllers, according to the linear quadratic regulator (LQR) demonstration. Nevertheless, in practice, the system dynamics may not always be accurately reflected by this linear approximation and may even be relatively wasteful. Nonlinear regulators, including the feedback linearization (FBL) controller, sliding mode controller (SMC), and modified sliding mode controller (MSMC), perform better in such situations. The trajectory tracking capabilities, dynamic performance, and potential disruption impact of both methods are evaluated and compared. The FBL with LQR was the best controller among them all. The SMC and the MSMC were also very good in tracking the trajectory