This article explores the potential for addressing issues related to improving the sustainability and safety of a city's transportation system. The study demonstrates the need for a comprehensive approach that considers not only the direct but also the indirect effects of potential solutions, selecting those that enhance the overall sustainability and safety of the system. The article presents a conceptual model of such a system and illustrates implementation methods using a large city as an example. Furthermore, it demonstrates how simulation modeling can address urban transportation network issues and quantifies the impact of implementing the proposed solutions.
With the advancement and widespread use of digital devices, capturing images has become effortless in any location. Images serve as evidence, making the authenticity of digital images increasingly critical. Some individuals alter images by adding or removing elements, rendering the images unreliable. Consequently, detecting image forgery has become essential. The evolution of image editing software has intensified this issue within the realm of computer vision. Recently, a variety of algorithms have been developed to identify image forgery. However, with the progress in digital technology, the ease of image manipulation has led to a surge in forgery cases, presenting significant obstacles when verifying their authenticity. Thus, there is a pressing requirement for effective forgery detection methods. This paper introduces a novel copy-move forgery detection approach based on the fusion of a handcrafted forgery detection system utiliszing the scale-invariant feature transform and the support vector machine with a deep-learning-based forgery detection system using VGG16. The aim is to address the challenges of subtle forgery detection and blending and seamless Integration. We conduct the experiments on the MICC_F2000 image manipulation dataset and assess the efficacy of the proposed approach, achieving $\mathbf{9 6 . 7 5 \%}$ accuracy, $\mathbf{1 0 0 \%}$ precision and $95.5 \%$ F1-score. This research demonstrates superior performance compared to state-of-the-art methods.
MobileNet is a lightweight convolutional neural network optimized for resource-limited environments. However, its use of depthwise separable convolutions to minimize parameters and computation can reduce accuracy due to oversimplified channel interactions. Principal Component Analysis (PCA) can address this issue by reducing the dimensionality of weight matrices while preserving key features. Applying PCA can help maintain accuracy and compress the model simultaneously. Based on this, we propose a novel copy-move forgery detection approach based on MobilNetV2 and PCA for feature extraction, and a random forest for classification. This allows us to enhance MobileNet accuracy while keeping its model size compact. The results of our experiments conducted on the MICC-F2000 dataset reveal that the proposed hybrid lightweight model outperforms the individual transfer learning structures and the existing literature, achieving 96.37% accuracy.
Mental Health (MH) is a fundamental and indispensable component of general well-being that permits an individual to perform their activities to the fullest, in harmony with themselves and their social and physical surroundings. It enables an individual to manage life’s challenges (literacy, attitudes toward disorders, and cognitive abilities) effectively. MH comorbidity commonly denotes the coexistence of several different MH problems, a matter of considerable importance in the realms of clinical care and the overall well-being of society. However, the precise prediction of comorbidity in mental health (MH) enables the implementation of earlier treatment strategies and improves overall treatment outcomes. In this study, we used Linear Regression (LR), Support Vector Machine (SVM), Random Forest (RF), Multi-layer Perceptron (MLP), K-Nearest Neighbors (KNN), Decision trees (DT), and AdaBoost with DT for the modeling of MH comorbidities. The aforementioned models were evaluated for accuracy using these metrics: Root Mean Squared Error (RMSE), Mean Squared Error (MSE), Rsquared ($\mathrm{R}^{\mathbf{2}}$), and Mean Absolute Error (MAE). The study found that KNN outperformed other models by achieving the highest scores of $0.0187,0.0109,0.00$, and 1.00 for RMSE, MAE, MSE, and $R^{2}$ respectively. In conclusion, ML can enhance the prediction of MH comorbidities. Therefore, with accurate and earlier predictions, patient satisfaction and effective medical therapies will be achieved.
This study explores the classification of date fruit varieties in the Kingdom of Saudi Arabia using advanced image processing techniques and genotype analysis. Utilizing a unified model with 30 epochs and 640x640 pixel resolution, we achieved a classification accuracy of $99.382 \%$ with both median and Gaussian filters. Notably, misclassifications were observed between Sugaey and Sokari varieties. The training loss exhibited a sharp initial decrease, stabilizing at a low value, while the validation loss showed fluctuations, particularly around the 20th epoch. Top-1 accuracy increased rapidly, reaching near $100 \%$ by epoch 10, with Top- 5 accuracy consistently higher. The low pass and high pass filters demonstrated similar trends, with Top-1 accuracy stabilizing near $90 \%$ and misclassifications predominantly involving the Sugaey variety. These findings underscore the potential of refined image processing techniques to enhance fruit classification, improve crop management, and support genetic enhancement efforts in the date industry, ultimately contributing to economic growth and product quality in Saudi Arabia. From the analysis the following results were obtained. In the median filter, there is a notable misclassification between Sugaey and Sokari. The validation loss decreases but shows some fluctuations, especially around the 20th epoch
the rapid spread of fake news online, coupled with the difficulty of distinguishing it from real news, has become a serious issue, especially with social media being a primary news source for many people. The fake news can have damaging effects on individuals, communities, and political and economic systems. Detecting Arabic fake news presents additional challenges due to the limited availability of relevant datasets and research. In this study, Perturbation Adversarial attacks was applied as a regularization technique for fake news detection, generating adversarial examples by modifying the model’s word embedding matrix. The AraBERTv2 model is used for preprocessing, and the lack of Arabic data was overcome by utilizing a translated English fake news dataset. A Recurrent Neural Network (RNN) model was trained on clean data, testing it with both clean and adversarial examples to evaluate its generalization capability. The results demonstrate that the RNN model performs effectively, achieving strong accuracy in Arabic fake news detection.
Cryptography plays a crucial role in Industry 4.0, safeguarding the security and privacy of data and communications within highly interconnected and automated industrial systems. As vast amounts of data are generated and transmitted across devices, systems, and cloud platforms, cryptographic methods are essential for ensuring data confidentiality and restricting access to authorized entities. This paper introduces two novel hybrid encryption algorithms: modified HIGHT-Simon and modified HIGHT-Trivium, which integrate the lightweight HIGHT cipher with Simon and Trivium ciphers, respectively. Both algorithms are evaluated across three configurations with different rounds 8,10, and 12 to assess their encryption and decryption performance. The results reveal that the modified HIGHT-Simon algorithm consistently achieves faster encryption and decryption times, making it ideal for real-time applications in Industry 4.0, such as industrial IoT and smart manufacturing systems, where both speed and security are critical. In contrast, the modified HIGHT-Trivium algorithm, while providing a more balanced encryption-decryption process, exhibits slower performance, making it better suited for applications that prioritize security over speed. Overall, the modified HIGHT-Simon algorithm emerges as the more efficient solution for ensuring both privacy and rapid data communication in the fast-paced, data-intensive environments of Industry 4.0.
Prostate-specific antigen (PSA) detection is very vital for the early, accurate diagnosis and management of prostate cancer. Enzyme-linked immunosorbent assay (ELISA) is a promising PSA detection method due to its high sensitivity, and researchers continuously explore its potential. However, selecting the best ELISA method from available options can be challenging. As a result, we focused on using the Preference Ranking Organization Method for Enrichment Evaluation (PROMETHEE) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) to assess the optimal technique for prostate cancer detection. In this study, the PROMETHEE and TOPSIS were used to evaluate different types of ELISA methods such as Direct (D-ELISA), Indirect ELISA (I-ELISA), Sandwich ELISA (S-ELISA), and Competitive ELISA (C-ELISA) based on Detection range (DR), Incubation time (IT), Ease of use (EOU), Cost-effectiveness (CE), Specificity (Sp%), Sensitivity (Se%), Linearity (Li), Precision (Prec), Accuracy (Acc), Limit of detection (LOD) $(\mathrm{ng} / \mathrm{mL})$, and Coefficient of variation (CV%) as criteria. The results showed that C-ELISA ranked first with a net flow value (Phi) of 0.0030 in PROMETHEE and second with a performance rating (Pi) of ${0. 5 8 6 4}$ in TOPSIS. Conversely, D-ELISA ranked second with a net flow value of 0.0000 in PROMETHEE and first with a Pi value of 0.5944 in TOPSIS. In both analyses, I-ELISA and S-ELISA ranked third and fourth, respectively, among the selected ELISA methods. This study highlights the importance of PROMETHEE and TOPSIS in selecting the most appropriate ELISA method for accurate prostate cancer detection using PSA.
The lungs regulate breathing and supply oxygen to every cell in the body. Simultaneously, they act as air filters, blocking potentially harmful particles and microbes from entering the respiratory system. Humans have built-in defences that keep their lungs safe. However, these cannot guarantee complete protection against lung diseases. The lungs are vulnerable to infection, inflammation, and malignant tumors. The aim of this research is to investigate the applications of three models, specifically an ANN, ANFIS, and a classical linear regression MLR for the prediction of lung cancer. To evaluate the models, we have considered four performance parameters: $\mathbf{R}^{2}$, MSE, RMSE, and R. To make the research robust, the study also deployed an MCDM tool called the fuzzy PROMETHEE to evaluate, compare, and rank the performance of the deployed models. The performance analysis revealed that ANFIS is the most effective model; hence, it is the central proposition of this study.
This paper focuses on the signal classifier component of a tele-rehabilitation framework that uses wearable surface electromyography (sEMG) devices to identify hand grasps and deliver real-time, adaptive therapy. The prediction accuracy and robustness of a stacking ensemble classifier significantly — with the combination of Support Vector Machine (SVM), Random Forest (RF), and a Logistic Regression meta—learner-improves performance, enabling continuous and personalized monitoring. Through the use of machine learning techniques, this study enables patients to conduct rehabilitation exercises remotely and to track their progress in real-time with therapists.
This study includes detailed steps of design and implementation of a wide-range grid voltage controller for application with DC and AC loads. The proposed controller offers a transformer-less function, which reduces the size of the controller. It is implemented using SCR switches controlled by a flexible pulse generator. The pulse generator is capable of producing trigger pulses of controllable widths for successfully switching on the controller’s SCRs. Within the pulse generator, a cascade of operational amplifier stages in various configurations is used to generate the firing pulses. Furthermore, an alternating voltage of 230 V/50 Hz is used from the grid to supply the controller. Thus, a wide range of AC load voltage from 0 V to 230 V is generated. This voltage is rectified in the case of a DC load. The system is first simulated, then practically implemented to validate the design of the controller. The test results reflect the capability of selecting suitable duration of firing pulses to guarantee SCR toggling. The presented solution is highly promising for varying AC and/or DC load voltages.
Distance learning is gaining prominence due to lower educational expenses and more autonomy in the learning process. Covid-19 pandemic also highlighted the significance of distance learning. This study recommends various pedagogical strategies using computational design in distance learning for architectural engineering courses. Herein, the question is how computational design using mixed reality wearables can develop new pedagogical strategies and optimize architectural engineering distance learning courses. A survey questionnaire was conducted on 45 participants to answer the question. The participants included students and instructors of architectural engineering courses and IT professionals with good experience in mixed reality wearables. By considering this population frame, the sample size of the study was set as 45. The researchers have selected an equal representation of 15 instructors, 15 students, and 15 IT professionals. The study found that the proposed technology of computational design using mixed reality wearables is highly relevant for architectural engineering courses for making the intellectual discourse interactive, engaging, and accessible. The use of computational design can develop better comprehension of map digitization and GIS. Future studies should expand on findings and highlight implementations for other educational domains based on a computational design.
Internet of Things applications are on the rise within different areas of health and medical service. With this rapid rise of these applications, threat actors become more interested in targeting such devices. Within the health and medical context, there are particular challenges in privacy and security. In this paper, we present an explainable machine learning model designed to detect multi-protocol network-based attacks on Internet-of-Medical-Things with high accuracy. The proposed system was trained and tested using CICIoMT-2024 dataset. The proposed system delivered an accuracy exceeding $99.9 \%$, with an $F_{1}$ score exceeding 0.99. To increase trust in the obtained results, the proposed system was explained using SHAP values to provide insights into the most impactful features, and the nature of their impact on the system’s decisions.
Osteoporosis, a prevalent skeletal disorder characterized by weakened bone strength and integrity, poses a significant health risk, particularly for older adults and postmenopausal women. Early detection is critical to mitigate fracture risks and improve patient outcomes. This research investigates the potential of pretrained convolutional neural networks for automated osteoporosis detection in knee X-ray images and highlighting the impact of image preprocessing techniques on model performance. We evaluate four pretrained models (ResNet-50, ResNet-101, VGG16, and VGG19) for feature extraction, coupled with a Random Forest classifier optimized using Bayesian Optimization. Our framework explores the effectiveness of different preprocessing methods, including Bilateral Filtering and Contrast Limited Adaptive Histogram Equalization, to enhance feature quality and improve classification accuracy. Evaluation on a dataset of knee X-ray images collected from the University of Sharjah Hospital reveals that VGG architectures, particularly VGG16, demonstrate superior performance in detecting knee osteoporosis. VGG16, preprocessed using Bilateral Filtering and CLAHE, achieved a $\mathbf{7 8 . 3 9 \%}$ accuracy, $\mathbf{8 0 . 0 0 \%}$ precision, and $\mathbf{7 6 . 9 2 \%}$ recall. This study underscores the importance of selecting both model architecture and preprocessing techniques for optimal performance. Further optimization and more data could potentially lead to a more robust and accurate model.
This research explores the potential for reducing the cost of advanced metering infrastructure (AMI) by leveraging raw images captured from the screens of digital energy instruments. The study focuses on the extraction of text and the recognition of 7 -segment numbers through optical character recognition (OCR) techniques. The proposed OCR-based dataset holds promise for facilitating fully automated electricity billing processes. Additionally, the research highlights the impact of utilizing high-resolution smart metre data in enhancing the efficiency, reliability, and resilience of distribution power grids. The “digital metre” dataset, comprised of images of digital energy metres, serves as a valuable resource for advancing the field. The study introduces a methodology for automatically reading dial-metre digits through the utilization of a deep learning model based on the YOLOv5 architecture. Evaluation metrics such as precision, recall, and mean Average Precision (mAP) are employed to assess the model's effectiveness. This research underscores the efficacy of the suggested network model in executing object-detection tasks, showcasing superior recall, mAP, and precision in the context of smart metre reading.
Non-destructive identification of cosmetics is essential to confirm their identity and purity. Near-infrared (NIR) spectroscopy provides a fast and non-destructive method for cosmetic identification. For this study, 99 raw materials and 30 products were measured using a palm sized NIR spectrometer. Spectral pre-treatment and analysis involved techniques such as multiplicative scatter correction-first derivative (MSC-D1), correlation in wavelength space (CWS), and Principal Component Analysis (PCA) respectively. Among the raw materials, 66 exhibited strong NIR activity, 25 showed medium NIR activity, and eight had weak NIR activity, while all 30 samples were found to be NIR active. The CWS method revealed a high frequency of Type II errors, with 81 out of 99 raw materials mismatching. All 30 of the products also resulted in mismatches. PCA proved to be the more accurate of the data analysis techniques with creams $\mathbf{9 7 \%}$ of variance was accounted for in PC1, and PC scores demonstrating strong differentiation among seven of the eight cream products. In contrast, the perfume products displayed $\mathbf{8 4 \%}$ variance in PC1 and $13 \%$ in PC2, with two of the seven products showing significant discrimination from the others. This study confirmed that NIR spectroscopy is well-suited for this application as it provides a quick analysis time, no sample preparation and allows for sample preservation.
The COVID-19 pandemic, declared by the World Health Organization in March 2020, originated in Wuhan, China and quickly spread worldwide. In response to the global health crisis, scientific cooperation across the globe has intensified, with a critical focus on leveraging machine learning and deep learning for faster and non-invasive medical diagnostics. This study aims to contribute to these efforts by utilizing cough audio signals to detect COVID-19. It proposes a comprehensive study that starts with gathering and preprocessing cough audio data. The research further enhances the performance and generalization of deep learning models through transfer learning, employing VGG19 to adapt pretrained neural networks for this specific task. Alongside evaluating the models’ performance, the study also explores their interpretability and explainability, which are crucial for practical implementation. The outcome of this research is expected to provide a reliable, non-invasive, cost-effective, and scalable method for early detection of COVID-19, potentially easing the heavy reliance on traditional RT-PCR testing. Through the comparative analysis of machine learning and deep learning models in this context, the study also aims to provide deeper insights into the effectiveness of these computational approaches in tackling global health challenges.
Synthetic aperture radar (SAR) image change detection (CD) involves identifying changes between images captured at different times over the same geographical region. SAR provides unique advantages for remote sensing applications, such as disaster monitoring, due to its capability to penetrate clouds and operate under all-weather conditions. However, the presence of speckle noise remains a significant challenge, hindering accurate change detection. To address this issue, this paper introduces a novel SEBlock and Multi-Head Self-Attention-Enhanced Bi-dimensional Aggregation Module (SEBAM) for robust feature extraction and noise suppression. SEBAM combines the Squeeze-and-Excitation (SE) block and Multi-Head Self-Attention (MHSA) mechanism to adaptively emphasise critical channel-wise and spatial dependencies, thereby enhancing the network’s ability to differentiate subtle changes in complex SAR images. Extensive experiments on three SAR datasets demonstrate that the proposed method significantly outperforms state-of-the-art techniques, achieving higher accuracy and robustness in change detection tasks.
Using machine learning (ML) techniques to determine factors affecting radiation dose for patients undergoing enhanced abdomen and pelvis computed tomography (CT). Factors such as patient characteristics, CT scan acquisition parameters, and patient misalignment were considered as input variables. CT dose indices namely dose-length product (DLP), volumetric CT dose index (CTDIvol), size-specific dose estimate (SSDE), size-specific estimate based on water equivalent diameter (SSDEWED), and the effective dose (effD) were considered output variables. Herein, 5 different ML algorithms were used, the Logistic Regression (LR), Random forests (RF), Neural Network (NN), k-Nearest Neighbors Algorithms (kNN), and Support Vector Machine (SVM). Demonstration of the most accurate ML technique to predict the most effective and influencing input variables will be included. Own to our data classification setup we were able to determine the most effective gender-based input variables. Although input variables for males and females were the same, their influence and ranking were different.
This paper presents a thorough review of recent advancements in screening heart and lung diseases via auscultation using artificial intelligence (AI) methods. Auscultation has historically been fundamental in diagnosing cardiopulmonary conditions; however, conventional techniques depend significantly on clinician proficiency, rendering diagnosis vulnerable to human error. Recent advancements in digital stethoscopes and AI-based sound analysis algorithms have transformed the conventional analysis, facilitating more precise, real-time identification of anomalies such as murmurs, arrhythmias, wheezes, and crackles. This paper delineates the principal methodologies employed in sound acquisition, feature extraction, and disease classification, while assessing the reliability of diverse models of machine learning and deep learning. Moreover, the paper addresses the obstacles in implementing these technologies in clinical practice, including data standardization, computational constraints, and integration with current healthcare systems. The results indicate that AI-augmented stethoscope systems have significant potential to enhance early diagnosis and patient outcomes for cardiac and pulmonary conditions.
