Search

Smoking has a serious complicated impact on Cardiovascular Health (CVH), which makes it an easily controlled risk factor for Cardiovascular Disease (CVD). The importance of early diagnosis and treatment is crucial because of CVD, which remains to be the leading cause of morbidity and death across the globe, therefore, smoking behaviour prediction is essential. This research explores the complex relationship between smoking and CVD, and clarifies how smoking negatively impacts CVH. The first step of the analysis is removing the Not-a-Number (NaN) values in the smoking behaviour bio signals data because all NaN values are in the target column ‘smoking’, which could lead to generate biased predictions. Further, robust scaling is applied to standardize the data, where all features are on the same scale. The Random Forest feature selection technique is applied to select the most relevant features for smoking status prediction. The Borderline-Synthetic Minority Oversampling Technique is used to balance the minority class with the majority class, to reduce the risk of overfitting in results. Finally, the cleaned and preprocessed data is passed to the Proposed Blending Model (PBM). In PBM, the Echo State Network and GoogleNet are placed on the base layer, and AlexNet is placed on the meta layer to make a robust smoking status prediction model. In terms of accuracy, precision, recall, and f1-score, the PBM outperforms the baseline techniques with an improvement score of 5%, 6%, 5%, and 7%, respectively. Additionally, the PBM’s results are validated through a 10-fold cross-validation method, which can increase its reliability and durability. Finally, the SHapley Additive exPlanations technique is used to understand the predictions made by the PBM. It can reveal the explainability and interpretability of the PBM. Conclusively, during the thorough simulation process, the results are collected and the ablation study is performed. The simulation results show that the early prediction of hidden smokers can help to reduce the effect of the CVDs. It also makes proactive healthcare strategies possible for better future outcomes and public health.

Recent synergistic deep learning techniques are underexplored in analyzing heterogeneous 2-D and 3-D radiographic data. Despite progress, existing heterogeneous approaches for 3-D volumetric data often rely on image-based methods, requiring manual selection of relevant slices and expert guidance. A prevailing challenge remains in harmonizing the analysis of volumetric radiographic data with variable lengths. To address these challenges, we proposed a unified deep learning-driven computer-aided diagnostic framework that analyzes multimodal 2-D and 3-D radiographic data to enhance diagnostic accuracy and aid clinical decision-making in radiology. The proposed framework primarily leverages the synergistic fusion of multi-level features within a lightweight Vision Transformer by introducing Multilevel-Multilayer Perceptron heads, which exploit and aggregate multilevel spatial features from the given input scan. A recurrent module further exploits 3D structural features, ensuring accurate decisions for volumetric data by dynamically adjusting its computation graphs to varying input lengths of volumetric radiographic data. Subsequently, a contextual map extraction module is designed to generate a well-localized activation map for the input scan, suppressing background activation from patch-level processing in the transformer module. Finally, we applied the proposed model to build a classification-driven radiographic retrieval system to retrieve relevant radiographic scans from the database that closely resemble the input test sample. We empirically validate our method on six publicly accessible radiographic datasets, including both X-ray and CT scans, demonstrating superiority (p-value

The significance of digital data, specifically of digital visual information has grown immensely and driving advancements in fields like medical, entertainment, and scientific research. The domain focused on retrieving such visual content is known as Content-Based Information Retrieval (CBIR). Content-Based Image Retrieval (CBIR) encompasses two primary sub-domains: image searching and reverse image searching. The study presents a significant contribution to the field of reverse image searching. Most current research focuses on reverse searching of individual images, primarily exploring techniques for searching single images in databases. However, composite images or collages are now widely used to represent summarized visual data in fields like medicine, entertainment, and satellite imagery. Even though our prior work for RISiC (baseline method) demonstrated promising results in accuracy, it did not explicitly address the challenge of time constraints. The study addresses the gap by proposing a method that achieves a balance between accuracy and time efficiency. The approach leverages Colored Binary SIFT (CBSIFT) combined with a novel region-balanced feature extraction technique. Additionally, an Inverted Index Structure is used in the online phase, along with a feature pruning method to enhance time efficiency while maintaining high retrieval precision. The work-balanced method for Reverse Image Searching in Collages (B-RISiC) is evaluated on Caltech-101, Corel5k, and Oxford5k datasets. The performance is benchmarked against two approaches: the simple BSIFT with Indexing and the baseline method. The results indicate that B-RISiC outperformed all others, achieving search times of 1.796 s, 7.763 s, and 0.718 s for the Caltech-101, Corel5k, and Oxford5k datasets respectively, with precision values of 0.8217, 0.8575, and 0.8810.

Multilevel inverters (MLIs) have remained a promising contribution to research in the modern era, especially in the area of energy conversion. Their popularity, nominal and optimal operations have merged them into low, medium, and high-power applications. Their use in advanced and wide range of applications have proved them as innovative solutions to many issues. Advanced MLIs structure tackle high-profile total harmonic distortion (THD) issues, maximum switching losses, low fault tolerance, low-quality output voltage, high rating of devices, low switching frequency, and frequency interfacing. Such factors classify them into various categories, which specify their operations in multiple applications. The existing literature lacks proper control strategies, design mechanisms, mathematical and application analysis. The aforementioned strategies help to investigate switching losses, configuration complexity, cost analysis, load management, input and output voltage categorization and many other prominent features of MLIs. Therefore, this paper provides a detailed review study and investigates various MLI topologies based on their structure, synthesis, mathematical explanations, control strategies, design mechanisms and utilization in various applications. The evolution of MLIs their classification, advantages, disadvantages, design mechanisms, control strategies, load vs input voltage interfacing and applications based utilization are tabulated to highlight their importance in various sectors. The proposed review study is compared with other prominent review studies and have covered a wide range of MLI topologies based on their evolutional development, mathematical representations, generic utilization and design methodologies.

Diabetic retinopathy is a leading cause of vision complications and partially sighted which pose considerable diagnostic difficulties because of its diverse and varying symptoms. Some of them include the fact that the disease displays a non-uniform pattern, where patients present different symptoms; the requirement of highly qualified specialists to interpret the images of the fundus; the risk of errors in the interpretation of images or their inconsistency; and the absence of clear morphological signs often makes early diagnosis unlikely. Traditional diagnosis mostly rely on the expert interpretation of retinal images, which can lead to bias and inaccuracy; highlighting the need for improved diagnostic methods. Although traditional Artificial Intelligence (AI) methods enhance the diagnostic capabilities remarkably, their black box nature and information opacity restrict healthcare providers to comprehend the reasoning framework of the AI to build trust and optimize its usage in practice. Explainable AI (XAI) is an emerging approach that addresses the black-box problem by improving the interpretability of models, which allows users to understand the logic behind certain decisions. This research proposed a diagnosis model for detecting diabetic retinopathy using XAI approaches that increases the interpretability of the models to help clinicians understand the reasons behind the decisions. The proposed model is used to enhance diagnostic accuracy, offer comprehensible, and concise insights regarding the diagnostics. The convergence history plots of the proposed model validate the learning process to achieve 94% better diagnostic accuracy than traditional methods while improving interpretability and applicability in healthcare settings, indicating improvement in accuracy and loss reduction.

Species identification is critical for biological studies, ecological monitoring, and conservation efforts. A comprehensive comprehension of the evolutionary mechanisms that lead to biological variety is necessary while species are distinct categories of living organisms; however, naming, identifying, and differentiating between species is more complex than it may seem. Traditional methods, relying on dichotomous keys and manual observation, are time-consuming and error-prone. Precise species identification is crucial for all taxonomic investigations and biological procedures. Numerous experts are currently engaged in the task of identifying a solitary species. To address these challenges, we present a robust artificial intelligence framework for species identification using deep learning techniques, specifically leveraging the ResNet-50 Convolutional Neural Network (CNN). Our approach utilizes a ResNet-50-based CNN to accurately classify 15 species, including humans, plants, and animals, from images taken at unique locations and angles. The dataset was pre-processed and augmented to enhance training, ensuring robustness against variations in lighting, occlusion, and background clutter. Featuring 4 million trainable parameters, our modified ResNet-50 model demonstrated superior computational efficiency and accuracy. The proposed model achieved an overall accuracy of 96.5%, with class-specific accuracies of 98.25% for humans, 97.81% for animals, and 96.90% for plants. These results surpass those of existing models such as GoogleNet, VGG, SegNet, and DeepLab v3+, highlighting the efficacy of our approach. Performance was evaluated using metrics such as sensitivity, specificity, and error rate, further validating its reliability. Our findings suggest that the ResNet-50-based CNN model is highly effective for automatic species identification, offering significant improvements in accuracy and computational efficiency.

The 3-D swin transformer (3DST) and spatial–spectral transformer (SST) each excel in capturing distinct aspects of image information: the 3DST with hierarchical attention and window-based processing, and the SST with self-attention mechanisms for long-range dependencies. However, applying them independently reveals the following limitations: the 3DST struggles with spectral information, while the SST lacks in capturing fine spatial details. In this article, we propose a novel tuning fork fusion approach to overcome these shortcomings, integrating the 3DST and SST to enhance the hyperspectral image (HSI) classification (HSIC). Our method integrates the hierarchical attention mechanism from the 3DST with the long-range dependence modeling of the SST. This combination refines spatial and spectral information representation and merges insights from both transformers at a fine-grained level. By emphasizing the fusion of attention mechanisms from both architectures, our approach significantly enhances the model's ability to capture complex spatial–spectral relationships, resulting in improved HSIC accuracy. In addition, we highlight the importance of disjoint training, validation, and test samples to enhance model generalization. Experimentation on benchmark HSI datasets demonstrates the superiority of our fusion approach over other state-of-the-art methods and standalone transformers. The source code has been developed from scratch and will be made public upon acceptance.

The transformer model encounters challenges with variable-length input sequences, leading to efficiency and scalability concerns. To overcome this, we propose a pyramid-based hierarchical spatial-spectral transformer (PyFormer). This innovative approach organizes input data hierarchically into pyramid segments, each representing distinct abstraction levels, thereby enhancing processing efficiency. At each level, a dedicated transformer encoder is applied, effectively capturing both local and global context. Integration of outputs from different levels culminates in the final input representation. In short, the pyramid excels at capturing spatial features and local patterns, while the transformer effectively models spatial-spectral correlations and long-range dependencies. Experimental results underscore the superiority of the proposed method over state-of-the-art approaches, achieving overall accuracies of 96.28% for the Pavia University dataset and 97.36% for the University of Houston dataset. In addition, the incorporation of disjoint samples augments robustness and reliability, thereby highlighting the potential of PyFormer in advancing hyperspectral image classification (HSIC).

Dynamic topology, high latency, and restricted bandwidth are among the challenges of 3D underwater Acoustic Networks for offshore exploration, pollution monitoring, and oceanographic data collection. The proposed mobility adaptive paradigm addresses these issues for three-dimensional and natural harsh underwater environments without assumptions by integrated energy efficiency, mobility adaptation, and energy-aware forwarder node selection to extend network life. The proposed framework and algorithms dynamically adjust network structure and topology for node placement with their power level adjustments in real time. A modified random walk model for node mobility is incorporated to replicate realistic underwater movements. The energy consumption model considers duty cycle, inactivity, gearbox, reception, and sleep. Relay nodes are selected based on signal quality, residual energy, and proximity of available sink nodes to improve network coverage and load balancing. The MAEARS framework outperforms existing methods, achieving a throughput of 90.82 kbps and a PDR of 90.62% to 96.81%, relaying 3762 to 39392 packets while minimizing packet failures from 423 to 1411.

In the recent era, the practical implementation of Autonomous Vehicular Networks (AVNs) with the vulnerable Vehicle-to-Vehicle (V2V) communication of autonomous vehicles and inadequate intelligent decision-making systems has become a primary concern in smart city mobility. This has led to the traffic congestion concerns such as time wastage, compromised safety, decreased durability and reliability of transportation infrastructure and V2V communication short delay and Roadside Units (RSUs), and reduced traffic flow. To address these issues, secure AVN communication and smart decision-making for autonomous vehicles in smart cities are of utmost importance. It ensures safety on roads, durability of the infrastructure, transparency, reliability, traffic congestion reduction and transportation efficiency. MapReduce is a reliable distributed computing paradigm which is able to analyze and process enormous AVN data in parallel. It contributes to smoother traffic flow by identifying the patterns and providing actionable insights for real-time decision making to decrease congestion. A private blockchain AVN can efficiently solve the problems of data security and reliability by providing tamper-proof record of all the transactions, hence enhancing reliability, and also offering a trusted solution of unauthorized access in real-time V2V communication. Explainable Artificial Intelligence (XAI) which is an efficient way to analyze fairness in traffic data over time providing transparency and availability of intricate traffic patterns, improving real-time traffic management with V2V communication and RSUs and reducing short delays that may occur as well as enabling traffic flow and the development of predictive traffic models that assist in decision making. This research proposed an XAI-based transparent model integrating MapReduce for processing large amounts of data and private blockchain technology for secured and tamper-proof vehicular communication. This proposed model is a promising solution for addressing the AVN data security issues and reliability of the system, mitigating negative effects of traffic congestion, and improving the transparency of decision making on the transport efficiency in smart cities. The proposed model provides a better performance than the previous approaches and gets 96% of the accuracy and 4% of miss rate.

Hyperspectral remote sensing (HSRS) or imaging spectroscopy is a novel method for obtaining a spectrum in every place of a massive array of spatial locations towards many spectral wavelengths that can be employed to make coherent images. HSRS comprises acquisition of digital images in many narrow, contiguous spectral bands through the visible, Near Infrared (NIR), Thermal Infrared (TIR), and Mid-Infrared (MIR) regions of electromagnetic spectrum. For the application of agricultural areas, remote sensing methods can be examined and implemented to the advantage of continuous and quantitative monitoring. In particular, hyperspectral images (HSI) are assumed to be accurate for agriculture as they provide physical and chemical data on vegetation. This study presents a new Agricultural Crop Classification using Coot Bird Optimization with Entropy-based Feature Fusion (ACC-CBOEFF) technique on HSI. The presented ACC-CBOEFF technique exploits the feature fusion and parameter tuning concepts to determine different types of crops on the HSI. Primarily, the ACC-CBOEFF technique exploits Gabor filtering as a preprocessing step. The entropy-based feature fusion process comprises ShuffleNet, EfficientNet, and DenseNet models for feature extraction. Moreover, Deep Variational Autoencoder (DVAE) system was utilized for crop type classification process. Furthermore, the CBO technique can be employed for the optimal hyper-parameter tuning of the DVAE approach. The simulation outcomes of the ACC-CBOEFF technique are tested on benchmark dataset, and the extensive outcomes highlighted the performance of the ACC-CBOEFF system over other recent state of art approaches.

In computer vision and image processing, vehicle detection and tracking in complicated aerial images have become important subjects. The need for automated systems that can precisely detect and track vehicles in aerial image data is growing due to the abundance of data coming from numerous sources, including drones and satellites. This study introduces a new method for lane extraction that relies on centroid tracking and template matching, followed by co-registration and geo-referencing. Our approach offers robust vehicle detection and tracking over a range of sizes and positions in complicated backgrounds, while also efficiently segmenting the region of interest. Our suggested method, which makes use of machine learning and feature extraction techniques, shows excellent precision and effectiveness when it comes to detecting and tracking vehicles in complex aerial images. This finding has important implications for traffic management and urban planning, going beyond computer vision and image processing. Our technology has the potential to transform traffic management procedures by making it simpler to detect traffic bottlenecks and monitor traffic flow. Furthermore, our method can help identify damaged vehicles in disaster response scenarios, which will help prioritize rescue and recovery activities. All things considered, our suggested approach is a significant addition to the domains of computer vision and image processing, with a broad range of uses in traffic control, urban planning, and disaster management.

Driver Drowsiness Detection (D3) is a challenging task as it requires analysis based on various behavioral and physiological signs such as health issues, mental stress, and exhaustion. Data analytics reveals that driver drowsiness is the reason for one-fifth of all traffic accidents worldwide. Thus, safety devices are valuable for alerting sleepy drivers regarding more danger that may occur. Constant real-time drowsiness detection in complex conditions and denoting is still an open issue. However, facing these challenges, this article proposed a technique called Driver Drowsiness Detection using Custom Deep Learning Model (D3-CDLM). This approach contains four different modules: In the given procedure the investigation and exploration include, 1) feature extraction and selection; 2) machine learning and ensemble methods; 3) deep learning; and 4) the combination of the two, Hybrid. The first operation computes HOG, which stands for Histogram of Oriented Gradient, which is rotation and illumination invariant and resistant to the information in the local areas. Then Principal Component Analysis or PCA is used to obtain the best or top HOG features that are used as inputs to machine learning and ensemble methods-based modules. For hard to learn facial features, transfer learning is also carried out, and a new 30-layer CNNs structure is proposed called CDLM. Finally, the hybrid module’s top features are investigated using the PCA control of the architecture in coordination with the proposed CDLM for detecting drowsiness. Empirical analysis that encompasses all districts were applied on Yawning Detection Dataset. The results reveal that the developed and designed deep learning and hybrid modules acquire better accuracy than the proposed and utilized machine learning-based module along with the compared existing approaches in the pertinent literature.

Wireless Body Area Networks (WBANs) have significantly transformed human life, particularly in sectors such as healthcare, fitness, entertainment, and sports. The strategic placement of sink nodes in WBANs plays a pivotal role in influencing network connectivity, power efficiency, and overall network performance. In the context of designing WBAN sink placement, the primary challenges revolve around ensuring energy efficiency and robust connectivity. For monitoring a patient’s vital signs, sensor nodes are implanted at various locations within the patient’s body, which transmit physiological data to a central hub known as a sink node. The selection of the optimal position for the sink node is crucial in minimizing node energy consumption during data transmission. Addressing the complexity of the problem, this paper introduces four approaches; (i) Weighted geometric-based sequential least square programming (WG-SLSQP). WG-SLSQP incorporates three key approaches: a) Geometric median b) Weighted average technique and c)Sequential least square programming technique, (ii) Humpback Whale Optimization Algorithm (HWOA), (iii) Distance-based random mean shift (D-RMS), and (iv) Voronoi-based Positioning (VP) for sink node placement. WG-SLSQP demonstrates greater stability and lower localization errors as compared to D-RMS, HWOA and VP. The residual energy of WG-SLSQP is 93.6%, D-RMS is measured at 92%, while VP and HWOA exhibit values of 90.4% and 90% respectively. Moreover, the Average Localization Error (ALE) for WG-SLSQP is 0.500 m, D-RMS is 0.568 m, whereas VP and HWOA have ALE values of 0.60 m and 0.619 m, respectively. The results indicate that the suggested WG-SLSQP approach outperforms its predecessors.

The aim of this study is to explore innovative non-linear programming (NLP) models utilizing the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method, tailored for resolving decision-making challenges within the framework of cubic $p,q-$ quasirung orthopair fuzzy sets $({\mathrm {C}_{\mathrm {(p,q)}}}\mathrm {QOFSs})$ . In prior research, data pertaining to an element has typically been gathered in the form of either interval-valued $p,q-$ quasirung orthopair fuzzy sets $({\mathrm {IV}_{\mathrm {(p,q)}}}\mathrm {QOFSs})$ or $p,q-$ quasi orthopair fuzzy sets ( $p,q-$ QOFSs) information. As an alternative approach, ${\mathrm {C}_{\mathrm {(p,q)}}}\mathrm {QOFSs}$ emerges as an extension of these sets, wherein information is compiled by simultaneously considering both ${\mathrm {IV}_{\mathrm {(p,q)}}}\mathrm {QOFSs}$ and $p,q-$ QOFSs. Driven by this motivation, we constructed the NLP models incorporating interval weights and integrating the relative closeness coefficient concept alongside weighted distance measures. Furthermore, we scrutinized some notable characteristics (RCC) of these models. Additionally, we introduce an innovative multicriteria decision-making (MCDM) technique and illustrate its application with a real-world case study related to green supplier selection. A comparative analysis is also performed to validate the effectiveness and rationality of the method.

Artificial Intelligence (AI) has garnered considerable attention from educators, policymakers, and scholars within the realms of education and social sciences, achieving notable historical significance. Employing bibliometric analysis, this paper aims to provide a comprehensive analysis of the most noticeable trend of using AI in education. The corpus comprises a dataset of publications sourced from Web of Science (WoS) and Scopus-indexed journals. It concentrates on contemporary scientific research and innovative methodologies within the education sector, particularly relevant to the intersection of education and technology in the AI era. The results reveal a significant rise in AI-related technological research beginning in 2018, with citations reaching their zenith in 2019. Furthermore, collaborative metrics indicate that the United States and China are leading in publication volume. In addition, the study highlights a rising interest in AI applications within education and healthcare, particularly in the context of the COVID-19 pandemic.

Short-term load forecasting (STLF) is vital in effectively managing the reserve requirement in modern power grids. Subsequently, it supports the grid operator in making effective and economical decisions during the power balancing operation. Therefore, this study comprehensively reviews STLF methods, including time series analysis, regression-based frameworks, artificial neural networks (ANNs), and hybrid models that employ different forecasting approaches. Detailed mathematical and graphical analyses and a comparative evaluation of these methods are provided, highlighting their advantages and disadvantages. Further, the study proposes a hybrid CNN-LSTM model comprised of Convolutional neural networks (CNN) for feature extraction of high dimensional data and Long short-term memory (LSTM) networks to boost the model’s efficiency for temporal sequence prediction. This study assessed the model using a comprehensive dataset from Pakistan’s NTDC national grid. The analysis revealed superior performance in short-term load prediction, achieving enhanced accuracy. For single-step forecasting, the model yielded an RMSE of 538.71, MAE of 371.97, and MAPE of 2.72. In 24-hour forecasting, it achieved an RMSE of 951.94, MAE of 656.35, and MAPE of 4.72 on the NTDC dataset. Moreover, the model has outperformed previous models in comparison using the AEP dataset, demonstrating its superiority in enhancing reserve management and balancing supply and demand in modern electricity networks.

Histopathology images are visual representations of tissue samples that have been processed and examined under a microscope in order to establish diagnoses for various disorders. These images are categorized by deep transfer learning due to the absence of big annotated datasets. There are some classifiers such as softmax and Support Vector Machine (SVM) used to perform multiple and binary classification respectively. Feature reduction for high dimensional images, is an emerging technique which can meet two basic criteria’s of classification i.e. it deals with over-fitting issue and it can also incredibly increase the classification accuracy. As disease diagnosis requires accurate histopathological image classification, so the proposed Dual Featured Auto-encoder (DFAE) based transfer learning is introduced with Triple Layered Convolutional Architecture. The Histological CIMA dataset is used after pre-processing by PHAT, a mathematical and computational framework to get spatial features as well as spectral features. In order to achieve the two objectives, the proposed integrated methodology uses reduced informative features from DFAE and fed them to Triple Layered Convolutional Architecture (TLCA). The conventional Convolutional Neural Network (CNN), ResNet50, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) are also tested against reduced dimensional image data but we found moderate or even low accuracies i.e. 25% for DFAE-ResNet50, 66% for DFAE-LSTM, 33% for DFAE-GRU and 67% for DFAE-CNN. While the accuracy of our proposed architecture Dual Featured Auto-encoder with TLCA (DFAE-TLCA) is better i.e. 96.07%. The proposed methodology has the potential to revolutionize the medical research.

Recommender systems are crucial in today’s digital world, by enhancing user engagement experience in digital ecosystems. Internet of things (IoT) have huge potential to generate dynamic and real time data. The data generated through IoT are being utilized to extract dynamic context of the user. Integrating recommender systems with context-aware (IoT data) and cross-domain (Knowledge Transfer) capabilities have the capacity to further enhance the accuracy and relevance of recommendation systems. However, recommender systems struggle with the cold start problem, where non-availability of data hinders to make effective recommendations for new users. Therefore, IoT-enabled Context-Aware Cross-Domain Recommender Systems may employ latent user profiling to provide personalized and exceedingly relevant recommendations across domains. The proposed system, named Knowledge Transfer Cross-Domain User Latent Factors (KT-CDULF), creates a user profile that spans multiple data domains, capturing a wide range of user behavior on all domains. The KT-CDULF captures the combined knowledge across domains to make recommendations even with limited user data, i.e. cold start problem. Domain-independent factors, and context can be used across domains to make relevant recommendations. The effectiveness of a recommender system depends on the density of the ratings in the data. To address this, KT-CDULF used two benchmark datasets to create user profiles and an item-rating matrix with additional context extracted from IoT generated dataset. KT-CDULF is evaluated and compared it with state-of-the-art models for recommender systems and achieves an accuracy of 98%, demonstrating the benefits of transferring knowledge containing context across data domains in recommender systems.

Stuttering is a speech disorder characterised by the repetition, prolongation, or blocking of sounds, syllables, or words, which can cause significant social and emotional difficulties for those who experience it. To help find and diagnose stuttering early, it is important for clinicians and researchers to accurately separate stuttering from normal speech. This helps them understand the disorder, find possible causes, and come up with effective interventions and treatments. The UCLASS dataset was used in this study, and 40 MFCC features were taken out to see how well different machine learning classifiers could indicate the difference between normal and stuttering speech. The major problem is the UCLASS imbalanced dataset. The authors address it with the synthetic minority oversampling technique. After machine learning experimentation’s, we propose a novel hybrid model that performs better than individual machine learning. In hybrid, the lightweight SutterNet model takes the best features from the data and then make prediction. The results indicate that the evaluated classifiers showed varying levels of performance. Overall, the results suggest that hybrid classifiers have the potential to accurately classify normal and stuttering speech, which could have important implications for the early identification and diagnosis of stuttering, as well as the development of assistive technologies and effective interventions and treatments.

Unmanned aerial vehicles (UAVs) have revolutionized various industries by providing efficient and automated flight capabilities. However, reliance on GPS and traditional navigation systems poses challenges in scenarios where signal interference or failures occur. In this research, we present a novel computer-vision-based method to enhance UAV navigation, enabling accurate height and location estimation. Our approach utilizes a sophisticated network that leverages a pair of images to estimate UAV height. The pyramid stereo-matching network is employed to extract robust image features and generate a disparity map. Subsequently, a custom network processes and convolves these data, employing diverse computer vision techniques to achieve precise height estimation. To evaluate the effectiveness of our proposed method, we collected a comprehensive dataset by conducting flights with a Phantom 4 Pro drone over the NUST Main campus, H-12 Islamabad. The dataset encompasses images captured at 10 different heights, spanning from 100 to 280 m, with flights evenly spaced 20 m apart. In rigorous evaluations, our approach demonstrates promising results compared to existing methods. By liberating UAVs from reliance on GPS, this vision-based 3-D localization technique holds immense potential to ensure successful flights even in challenging environments.

Computed Tomography (CT) is a non-invasive imaging modality used to detect abnormalities in the human body with high precision. However, the electromagnetic radiation emitted during CT scans poses health risks, potentially leading to the development of metabolic abnormalities and genetic disorders, which increase the risk of cancer. The Low-Dose CT (LDCT) scanning technique was developed to address these hazards, but it has several limitations, including noise, artifacts, reduced contrast, and structural changes. These drawbacks significantly reduce the diagnostic capabilities of Computer-Aided Diagnosis (CAD) systems. Eliminating these noises and artifacts while preserving critical features poses a significant challenge. Traditional CT denoising algorithms struggle with edge blurring and high computational costs, often generating artifacts in flat regions as noise levels increase. Consequently, deep learning-based methods have emerged as a promising solution for LDCT image denoising. In this study, a comprehensive Systematic Literature Review (SLR) following PRISMA guidelines was conducted to explore the latest advancements in deep learning algorithms for LDCT image denoising. This SLR spans LDCT image-denoising research from 2018 to 2024, providing a detailed summary of methodologies, benefits, limitations, parameters, and trends. This study delves into the acquisition process of CT scans, investigating radiation absorption across various anatomical regions, as well as identifying sources of noise and its distribution within the LDCT images. Additionally, it enhances our understanding of LDCT image denoising trends and provides valuable insights for future research, thus making a substantial contribution to ongoing efforts to enhance the quality and reliability of LDCT images.

Image dehazing is a crucial task in the early stages of real-time image processing pipelines used in applications such as surveillance and advance driver assistance systems. Dehazing algorithms mitigate the spatially selective degradation of image details caused by natural phenomenon such as fog and sandstorms. The problem is exacerbated in the presence of color cast which can affect the color-sensitive processing in the downstream tasks. In order to correct these two aberrations, we have proposed a light-weight algorithm which is not only quantitively more effective than the state-of-the-art works but also uses minimal computational resources. Specifically, it has been proposed to tackle both aforementioned problems by processing luminance and chrominance channels separately through a custom resource-efficient colorspace transform. Moreover, it has been proposed to employ online calculation of the relevant video stream statistics to estimate the degradation model over several temporally adjacent frames. This approach not only reduces the hardware resource utilization when implemented inside the video processing pipeline but also reduces the flicker effect observed when frames are processed individually. It has been demonstrated through quantitative analysis on standard datasets that the proposed approach either works at par or better than the reference works in terms of image quality metrics. Furthermore, the proposed framework has been developed as a real-time video processing system on an FPGA platform. The synthesis results of this implementation suggest that the proposed framework achieves this performance using minimal logic resources.

Systems must be capable of detecting and tracking autonomous vehicles for intelligent management and control of transportation. Even though several methods are used to create intelligent systems for traffic monitoring, this article explains how to detect and track vehicles using pixel labeling and particle filter algorithms. We suggested a novel technique that segments the image using image segmentation to retrieve the foreground objects. We have divided our proposed model into the following steps: at first, geo-referencing is used to find the exact location; secondly, the images are denoised by using pre-processing; image segmentation is used to separate the background from the foreground; multiple objection detection is performed using the random forest to classify different objects; vehicles are detected through a method called template matching; after this, the vehicles are counted using histogram of oriented gradients (HOG); and after counting, the tracking of vehicles is obtained using particle filter; lastly, the trajectories are predicted by comparing the rectangular centroid of each car against the frame number and using it as a time stamp reference, the last match that the tracking algorithm obtained for each vehicle was recorded and used to estimate the trajectories. Our model outperforms current traffic monitoring approaches in terms of detection and tracking accuracy by 0.87 and 0.92, respectively using the Aerial Car dataset and 0.84 and 0.88, respectively, using the AU-AIR datasets. Vehicle recognition, traffic density detection, traffic flow analysis, and pedestrian route generation are all possible uses for the proposed system. It can transform traffic management and improve overall road safety due to its powerful algorithms and cutting-edge technologies. The system’s adaptability makes it a significant asset in modern transportation, from optimizing signal timing to improving pedestrian navigation.

Intrusion Prevention Systems (IPS), capable of preventing the organizational network from a cyber-attack in addition to detecting it, are widely adopted by organizations to protect their networks from unauthorized access, attacks, and malicious activities. Similarly, Snort an open-source IPS is extensively used for effective network security monitoring and analysis. When functioning as an IPS, Snort can be deployed in inline mode within an organizational network, so that all the organizational network traffic travels through it, hence actively blocking or preventing malicious traffic in real-time. This requires Snort to process the network traffic fast enough to match the network traffic line rate. But the Snort IPS default data acquisition module i.e. advanced packet filtering (AF_PACKET) cannot process network traffic at the line rate that causes packet loss and network services disturbance. This research work discusses the technologies available to make Snort IPS process network traffic at line rate. Packet filtering framework (PF_RING) and data plane development kit (DPDK) are the most effective and widely used software technologies, whereas the Napatech smart network interface card (smartNIC) is a very efficient hardware technology for achieving line rate traffic processing. A throughput comparison shows that PF_RING and DPDK achieve a throughput close to 1G with 100% CPU utilization whereas Napatech smartNIC achieves full 1G throughput with CPU utilization of less than 5%. Furthermore, the integration of Snort IPS with the security information and event management (SIEM) system has been discussed for better attack detection in an organizational network.

In the era of digital data proliferation, agriculture stands on the cusp of a transformative revolution driven by Machine Learning (ML). This study delves into the intricate interplay between Information and Communications Technology (ICT) and conventional agriculture, emphasizing the role of ML in reshaping farming practices. With the ongoing data tsunami impacting data-driven businesses, the fusion of smart farming and precision agriculture emerges as a beacon of innovation. ML algorithms, analyzing historical and real-time environmental data, soil conditioning, predicts suitable crop for maximum yields, detect diseases, and optimize irrigation in smart farming, facilitating informed decision-making. Precision agriculture benefits from autonomous vehicles and drones, driven by ML, ensuring precision in planting, harvesting, and crop monitoring. Resource efficiency increases as ML optimizes energy consumption, manages fertilizer application, and promotes climate-resilient practices. This comprehensive assessment underscores ML’s pivotal role in maximizing productivity, minimizing environmental impact, and navigating the complexities of modern agriculture.

Researchers are turning to nanofluids in PV/T hybrid systems for enhanced efficiency due to nanoparticle dispersion, improving thermal and optical properties over conventional fluids. Three different concentrations of formulated soybean oil based MXene nanofluids are considered 0.025, 0.075 and 0.125 wt.%. Maximum specific heat capacity nanofluids ( $c_{pNF}$ ) augmentation is 24.49% at 0.125 wt.% loading of Ti3C2 in the base oil. The calculation of the $c_{pNF}$ based on the temperature and nanoflakes concentration is very expensive and time-consuming as it should be calculated via the practical test investigation. This study employs a long short-term memory (LSTM) as an efficient machine learning method to extract the surrogate model for calculating the $c_{pNF}$ based on the temperature and nanoflakes concentration. In addition, a couple of other machines learning methods, including support vector regression (SVR), group method of data handling (GMDH), and multi-layer perceptron (MLP), are developed to prove the higher efficiency of the recently proposed LSTM model in the calculation of the $c_{pNF}$ . In addition, the Bayesian optimization technique is employed to calculate the optimal hyperparameters of the developed SVR, GMDH, MLP and LSTM to reach the highest efficiency of the system in predicting the $c_{pNF}$ based on temperature and nanoflakes concentration. Notably, 95% of the recorded data via differential scanning calorimetry (DSC) is used for training machine learning techniques. In comparison, 5% is used for testing and validation purposes of the developed algorithm. The newly proposed optimized SVR, GMDH, MLP, and LSTM are modelled in MATLAB software. The results show that the newly proposed optimized LSTM model can reduce the mean square error in calculating the $c_{pNF}$ by 99%, 99% and 91% compared with SVM, GMDH and MLP, respectively. The proposed methodology can be used to calculate other thermophysical properties of nanofluids.

The traditional Control Shed Poultry Farm (CSPF) systems are operated manually or in a semi-automatic culture. The existing CSPF system is still not free from human intervention and needs 24/7 monitoring. The existing approach is laborious, prone to caretakers bias and human error, and repeatability is difficult. The Intelligent CSPF system is based on the Internet of Things (IoT) and Machine Learning (ML). The proposed approach comprises the learning and mastering phase. In the learning phase, the system will observe or learn from user behaviour patterns regarding poultry activity operations for a specific time duration. Upon completion of the learning phase, the mastering phase will eventually automate the system and control the poultry farm’s environmental parameters, such as temperature, humidity, water level, light, and hazardous gases based on the historical data. The proposed system will replicate the operator’s past business control behavior. For this learning task, we have used Supervised ML techniques to analyze the performance of ML algorithms such as Random Forest (RF), Decision Tree (DT), K-nearest Neighbors (k-NN), Support Vector Machine (SVM), and Naïve Bayes (NB). DT is the successful ML classifier with the highest accuracy performed in the Intelligent CSPF System; this leads to the development of a smart CSPF culture where successful business experiences, i.e. models, will be exchanged among the community to get a collective benefit in terms of revenue.

With an increase in the number of processing cores or systems, the high-performance edge-computing system’s power consumption along with its computational speed will increase, essentially. However, this comes at the expense of high-energy utilization. One notable solution to reduce the energy consumption of these systems is to execute these systems at the slowest feasible speed so that the job’s deadline times are met. Unfortunately, this method is at the expense of more response time and performance loss. To resolve this issue, in this paper, we propose a scheduling approach that associates the genetic algorithm (GA) with the first feasible speed (FiFeS) technique i.e. GA-FiFeS algorithm. This does not jeopardize real-time tasks’ deadlines. The GA-FiFeS algorithm proposes an energy-efficient schedule while still ensuring high response times. The results of the proposed approach, using plausible assumptions and experimental parameters, are compared with currently in-practice approaches, i.e. FiFeS and LeFeS (least feasible speed) approaches. Using numerical simulations and plausible assumptions, our investigation suggests that the proposed GA-FiFeS technique outperforms the FiFeS technique in terms of energy consumption (~18.56%) and response times (~2.78%). Furthermore, the GA-FiFeS has comparable outcomes with the LeFeS method while taking the expected time of execution as an assessment feature for analysis.

Real-time audio/video transmission through Internet media is an important part of communication. Due to bandwidth limitations and a noisy environment, delivery of multimedia content to a remote location is not 100% guaranteed. These limitations are the basic cause of missing packets which affect the Quality of Service (QoS). A protocol for the recovery of lost packets is described in [Maxemchuk, Nicholas F., K. Padmanabhan, and S. Lo. “A cooperative packet recovery protocol for multicast video.” Network Protocols, 1997. Proc. of 1997 International Conference on. IEEE]. This protocol claims significant improvement in QoS. We formally specify the protocol in a network of timed automata. By model-checking (A mathematical technique), we find that packet recovery is not always there. In this article, We report such scenarios of malfunctioning in the protocol when the size of multimedia contents is known (e.g., live video/audio broadcasting) and middle-level servers have different rates of data sending and receiving. We formulate the effect of inter-packet delay and transmission speed difference on a buffer.

Industrial Internet of Things (IIoT) applications consist of resource constrained interconnected devices that make them vulnerable to data leak and integrity violation challenges. The mobility, dynamism, and complex structure of the network further make this issue more challenging. To control the information flow in such environments, access control is critical to make collaboration and communication safe. To deal with these challenges, recent studies employ attribute-based access control on top of blockchain technology. However, the attribute-based access control frameworks suffer due to high computational overhead. In this paper, we propose an improved role-based access control framework using hyperledger blockchain to deal with IIoT requirements with less computational overhead making the information control process more efficient and real-time. The proposed framework leverages a layered architecture of chaincodes to implement the improved access control framework that handles the permission delegation and conflict management to deal with the dynamism of the IIoT network. The system uses a Policy Contract, Device Contract, and Access Contract to manage the workflow of the whole access control process. Each chaincode in the proposed framework is isolated in terms of its responsibilities to make the design low coupled. The integration of improved access control with blockchain enables the proposed framework to provide a highly scalable solution, tamper-proof, and flexible to manage conflicting scenarios. The proposed system outperforms the recent studies significantly in computational overhead in extensive simulation results. To verify the scalability and efficiency, the proposed is evaluated against a large number of concurrent virtual clients in simulation and statistical analysis proves that the proposed system is promising for further research in this domain.

GaN high-electron-mobility transistor (HEMT) has superior features of wide band gap, high electron mobility and very high electric field strength due to its material advantages. By using the GaN HEMT, switching frequency can be enhanced up to megahertz with extremely high efficiency. Unfortunately, GaN HEMTs accomplished by undesirable switching oscillations and voltage spikes due to extremely fast switching frequencies with very high ${dv}/{dt}$ , ${di}/{dt}$ and parasitic parameters. In this paper, RLC equivalent circuit models are developed for turn on and turn off conditions, including all parasitic components. In addition, the relative effect of each parasitic parameter is analyzed and estimated. Moreover, simple mathematical model is developed for theoretical analysis of switching oscillation phenomenon and, for guidance of snubber or damping circuit design. To validate these simple equivalent circuit models, both circuit simulation and experimental measurements are employed.

Health-related data has a decisive role in disease diagnosis. Collecting relevant information from health-related data in medical records has been facilitated by evaluating the features of the data. Relevant research has shown that outcomes are significantly impacted by the use of feature selection (FS) in different medical domain data. FS provides an analysis of the most significant features to improve classification accuracy. The FS technique aims at minimizing the number of input variables and computational overload to maximize classification performance results. However, identifying the optimal features poses issues due to the high dimensionality of large features and the small sample size of health-related data. The metaheuristics optimization algorithm (MOA) plays an important role in generating the best subset features with exploration and exploitation phases. This study experiments with well-known MOAs and supervised learning from the UC Irvine Machine Learning Repository, PhysioNet, Kent Ridge Bio-Medical Dataset, and MIMIC-III v1.4 Repository with varying feature dimensions. To increase the quality of health-related data, this study proposes missing data imputation based on a deep learning approach, an autoencoder (AE). With AE imputation, the performance results obtain 0.0167 mean squared error (MSE) and 0.129 root mean squared error (RMSE). As a result, MOA shows its excellence in achieving minimal features, but still outstanding performance in low- and high-dimensional data. MOA is successfully applied to varying diverse health-related datasets with low- and high-dimensional data.

In wind turbine engineering, stability and control rely on precision. A new approach for discrete-time systems is presented in this study, which makes use of constrained Gramians and frequency weights. Wind turbines with a double-fed induction generator and dynamic rotational speeds can have their model order reduced using the suggested method, which makes use of sophisticated state-space representations. A novel balanced realization method, along with frequency-weighted and limited Gramians, successfully lowers the dimensionality of large state models. Minimizing approximation errors and ensuring stability are both achieved by the resulting lower-order system. This paper makes a significant contribution by offering an a priori formula for error boundaries, which allows for more efficient and faster computations. A paradigm shift in improving the accuracy of modeling techniques is marked by this groundbreaking method, which applies frequency-weighted and limited Gramians to real-time systems like wind turbines.

Cryptocurrencies, recognized by their extreme volatility due to dependency on multiple direct and indirect factors, offer a significant challenge regarding precise price forecasting. This uncertainty has led to investment hesitation within the digital currency market. Previous research attempts have presented methodologies for price forecasting and trend prediction in cryptocurrencies. However, these forecasts have typically suffered from increased error rates, leaving the opportunity for improvement in this field. Furthermore, the influence of sentiment-based factors could compromise the reliability of price predictions. In this research, we have proposed a machine learning-driven framework that provides precise cryptocurrency price projections and adds an alert mechanism to guide investors. Our fundamental analyzer, Bi-LSTM and GRU hybrid model use historical data of digital currencies to train and reliably anticipate future values. Complementing this, a sentiment analyzer, utilizing a BERT and VADER hybrid model, analyzes sentiments to assess the forecast price as trustworthy or uncertain. Besides assisting investor decision-making, this technique also helps risk management in the dynamic realm of cryptocurrency. Our suggested approach delivers highly precise price predictions with dramatically decreased error rates compared to prior competitive studies. The proposed Bi-LSTM-GRU-BERT-VADER (BLGBV) model is tested for three cryptocurrencies, namely BTC, ETH, and Dogecoin and reports an average root mean square error (RMSE) of 0.0241%, 0.0645%, and 0.0978%, respectively.

Node localization is one of the most essential features of wireless sensor networks (WSNs). Vavarious localization algorithms exist for densely deployed 3-D wireless sensor networks. However, for a sparse 3-D network, range-based localization is still a challenging task because it is difficult to find sufficient anchor nodes and distance information among nodes in a sparse 3-D network. To mitigate the sparseness issues in 3-D sensor networks, we present a component-based localization method in this paper in which we split the entire network into small overlapping sub-networks called components and assign local coordinates to each component. Then, we merge these small components to make a globally coordinated system. With a meager anchor ratio, we localize the whole network. We define merging conditions according to the number of common nodes, actual measured distances among nodes, and the calculated distance based on the local coordinates of the nodes. We assess how well our proposed algorithm performs by conducting extensive simulations. The outcomes confirm that the proposed algorithm works comparatively better in a sparse 3-D sensor network than in a densely deployed 3-D sensor network. Our algorithm localizes more than 83% of nodes at a node degree of 10 having 5% anchor ratio; however, other algorithms localize only 18%-79% in the same scenario.

Drug Side Effects (DSE) are inconvenient and inadvertent retorts of the drugs. DSEs impact on public health and healthcare can prove costly. These DSEs can be an important factor in the failure/acceptance of drugs. Every approved drug should be either free from DSEs or these should be minor and reported properly. The drug discovery process should be capable of predicting and preventing these effects in advance. Previously, proposed studies for the prediction/prevention of DSEs utilized the features of 1D drug chemical structures or Natural Language Processing (NLP). Both these techniques required a complex transformation process. In this research authors have proposed a deep learning model, specifically using a transfer learning approach to predict DSEs directly from 2D chemical structure images, eliminating the need for the hefty transformation process of the NLP domain. For this study, a unique dataset is prepared that associates each image (taken from PubChem) with its specific side effects (SIDER). The results are evaluated using Accuracy, Precision, Recall and F-measure. The proposed model showed its dominance with an Accuracy of 73%, Precision of 83%, Recall of 73%, and an F1 score of 75%. The achieved results of the proposed model are compared against established transfer learning models like VGG16, DenseNet121 and some previously used traditional machine learning models like SVM and KNN. The collected results indicate a significant advancement in predicting drug side effects and offer a promising avenue for streamlining the drug development process.

The deployment of 5G has exposed capacity constraints in realizing the key vision of the Internet of Everything (IoE). Therefore, the researchers are exploring potentials of Digital Twin Network (DTN) in wireless networks. DTN is a novel technology to create virtual replicas of physical environment for testing, optimizing, and managing wireless networks. The integration of Artificial Intelligence (AI) and DTN appears to be a promising approach to address communication systems by providing an efficient environment for testing and improving AI models before deployment in real networks for effective network management, optimal resource allocation, and precise behavior prediction. Therefore, AI-enabled DTN in 6G represents a compelling avenue to address multifaceted challenges faced by wireless networks. In this comprehensive work, we offer a holistic survey that delves into the state-of-the-art approaches for AI-enabled DTNs in 6G. Firstly, we discuss the evolution of wireless networks and concept of AI-enabled DTN in 6G. Secondly, we discuss the role of AI-enabled DTN in 6G and driving advancements in fundamental components of 6G including resource allocation, caching, data offloading, and data security. Thirdly, we conduct a detailed discussion on key enabling technologies for realizing the capabilities of AI-enabled DTN in 6G. Fourthly, several applications of AI-enabled DTN in 6G are discussed for the practical relevance and significance in various industries such as smart cities, healthcare, and transportation etc. Finally, we provide lessons learned and highlight existing challenges and research directions to embark on further research efforts in the realm of AI-enabled DTN in 6G.

Smart grid power networks are essential for addressing the global energy crisis and combating climate change. In the past few decades, information and communication infrastructure have greatly improved. As a result, studying the characteristics of smart grids has become important. To accurately represent the connectivity of different components in power networks, we need precise models. In this study, we introduce a new growth model called PowerX. This model is designed to capture the characteristics of real-world power networks. PowerX is a growth model that is designed to capture the characteristics of real-world power networks by incorporating both random and ordered elements. Specifically, it is designed to accurately capture power networks’ degree distribution and clustering coefficient. To assess the effectiveness of PowerX, we compared it with existing growth models such as Watts Strogatz Small World model, Henneberg’s model, and Modified Henneberg’s model, using the US Western States Power Grid dataset consisting of 4789 nodes and 5571 edges. Our results show that PowerX precisely captures the degree distribution of the real dataset, and its clustering coefficient is close to the actual dataset, outperforming the other comparable models. In addition, we used Gephi to demonstrate the features of the Western States power grid, including identifying the most important node of the network, community structure, and the strongest and weakest nodes. This research provides valuable insights into the characteristics of power networks and demonstrates the effectiveness of PowerX in accurately modeling them. The datasets and codes are publicly available for further research at: github.com/irfan2inform/powerX.

Artificial intelligence (AI) related to intelligent control in healthcare denotes using AI techniques to enhance the management and control of healthcare processes and systems. Damage to the inner and middle ear caused by accidents and diseases even causes hearing impairment in the ear that has been harmed or injured. Traditional otoscopy devices were utilized to check the tympanic membrane (TM) to identify OM in medical practice, and a conclusion is drawn depending on the outcomes of the examination. While developing a computer-aided method to support the OM diagnosis, it is possible to focus on methods like feature extraction, image pre-processing, classification, and image segmentation. The existing methodology of detecting the ear infection experiences a reduction of accuracy due to the influence of the noise in the input ear image. This presence of noise affects the feature extraction process, directly influences the accuracy in detection process. To overcome this issue, in this manuscript, a Deep learning (DL) is utilized to find biomedical ear infections by examining images of the eardrum and ear canal. The process includes training a DL method with a large dataset of ear images, where the images were labeled as either not infected or infected. With this motivation, this article emphasizes the design of Bayesian optimization with a deep learning-based automated ear infection detection and classification (BODL-AEIDC) model. The BODL-AEIDC technique exploits the DL model with a metaheuristic optimization algorithm for the ear infection classification process. The BODL-AEIDC technique employs a Wiener filtering (WF) based noise removal process to eliminate the noise data. In addition, the BODL-AEIDC technique exploits W-Net-based segmentation and the EfficientNet model for feature extraction purposes. Moreover, the BODL-AEIDC technique employs a fuzzy Restricted Boltzmann machine (FRBM) model for ear infection detection. Furthermore, the BO algorithm is utilized to adjust the FRBM technique’s hyperparameter values effectively. The BODL-AEIDC technique’s experimental outcomes occur using the medical dataset. The comprehensive comparative study stated the enhanced performance of the BODL-AEIDC approach over other existing methods.

Globally, pests and plant diseases severely threaten forestry and agriculture. Plant protection could be substantially enhanced by using noncontact, extremely effective, and reasonably priced techniques for identifying and tracking pests and plant diseases across large geographic areas. Precision agriculture is the study of using other technologies, such as hyperspectral remote sensing, to increase cultivation instead of traditional agricultural methods with less negative environmental effects. In this article, we proposed a novel deep-learning architecture and optimization algorithm for crop leaf disease recognition. In the initial step, a multilevel contrast enhancement technique is proposed for a better visual of the disease on the leaves of cotton and wheat. After that, we proposed three novel residual block and self-attention mechanisms, named 3-residual block-deep convolutional neural network (RBNet) Self, 5-RBNet Self, and 9-RBNet Self. After that, the proposed models are trained on enhanced images and later extracted deep features from the self-attention layer. The 5-RBNET Self and 9-RBNET Self performed well in terms of accuracy and precision rate; therefore, we did not consider the 3-RBNET Self for the next process. The dragonfly optimization algorithm is proposed for the best feature selection and applied to the self-attention features of 5-RBNET Self and 9-RBNET Self models to improve the classification performance further and reduce the computational cost. The proposed method is evaluated on two publically available crop disease images, such as the cotton, wheat, and EuroSAT datasets. For both crops, the proposed method obtained a maximum accuracy of 98.60% and 93.90%, respectively, whereas for the EuroSAT, the proposed method obtained an accuracy of 83.10%. Compared to the results with recent techniques, the proposed method shows improved accuracy and precision rate.

As the world’s population rises, there will be a greater need for food, which will have repercussions on the environment and on crop yields. Increased production, efficient resource allocation, climate change adaptation, and diminished food waste are the four cornerstones of Agriculture 4.0’s vision for the future of farming. Agriculture 4.0 makes use of cutting-edge data systems and Internet technology to acquire, analyze, and organize massive amounts of farming facts such as weather reports, soil conditions, market demands, and land usage to better guide farmers’ decisions and boost their bottom lines. As a result, research on agricultural decision support systems for Agriculture 4.0 has gained significant momentum. Crop monitoring and yield forecasting are two applications where remote sensing has proven useful, and these two areas are intrinsically linked to variations in soil, weather, and biophysical and biochemical factors. Multi- and hyper-spectral data, radar, and lidar imaging are just some of the remote tools that could be employed for crop monitoring and yield forecasting. This paper’s goal is to examine some of the difficulties that can arise in the future while using agricultural decision-support platforms in the context of Agriculture 4.0. Addressing these identified obstacles may help future researchers create better decision-assistance systems. This research examines the possibilities, benefits, and drawbacks of each method, as well as how well they work in various agricultural settings. Furthermore, these methods are demonstrated in a variety of strategies that can be effectively employed. In this research, we take a look at some remote sensing techniques developed to increase farm profits while minimizing their impact on the natural world. This research shows how remote sensing information can be used to predict crop yields, evaluate plant nutrient needs and soil nutrient levels, calculate plant moisture levels, and manage weed populations, among other applications.

Smart contracts are becoming increasingly popular for managing transactions or activities in fog computing environments. However, the use of smart contracts for registration and resource access granting is vulnerable to various types of attacks that can compromise their security. Detecting these attacks can be challenging, as attackers can use sophisticated techniques to evade detection. This research uses a machine learning-based approach for detecting different attacks on smart contracts used for registration and resource access granting in fog computing. Data is collected from online Ethereum’s official site “etherscan.io”. Different feature extraction methods and machine learning models are tested. Using accuracy, precision, recall, F1 score, cross-validation, and computational time, the performance of models is evaluated. Results indicate that extreme gradient boosting (XGB) and random forest (RF) provide the highest accuracy of 80% using the term frequency-inverse document frequency (TF-IDF) approach. The light gradient boost classifier provides the highest accuracy of 81% with the Bag of Word (BoW) approach. Similarly, the extra tree provides the highest accuracy of 83% using the N-gram technique. Furthermore, performance using TF-IDF is slightly poorer than BoW and N-gram, however, it has less computational complexity.

Advanced assistive devices developed for activities of daily living use machine learning (ML) for motion intention detection using wearable sensors. Trunk assistive devices provide safety, balance, and independence for wheelchair users individuals who spend prolonged hours in sitting positions. We used ML for trunk movement intention detection with a trunk orthosis. Sensor fusion technique with four electromyography (EMG) and one inertial measurement unit (IMU) sensor signals are used to develop a three-level classification system. Forty participants engaged in seated trunk movement trials wearing the orthosis. The trials comprised 30 movements involving trunk flexion/extension, lateral bending, and axial rotation. The wrapper method was used to reduce essential EMG features. Ensemble (ES), k-nearest neighbors (KNN), and support vector machine ML classifiers were used. Twenty-six features (five EMG for each of four muscles and six for IMU) were used to develop ten individual ML models, resulting in an average accuracy of 95.44%. Eight models achieved the highest accuracy with the ES and two with KNN. The models were then cascaded to form a trunk motion detection system that achieved a test accuracy of 87.0%. The promising result of this study can be implemented for trunk motion recognition with active trunk orthosis.

Due to exponential population growth, climate change, and an increasing demand for food, there is an unprecedented need for a timely, precise, and dependable assessment of crop yield on a large scale. Wheat, a staple crop worldwide, requires accurate and prompt prediction of its output for global food security. Traditionally, the development of empirical models for crop yield forecasting has relied on climate data, satellite data, or a combination of both. Despite the enhanced performance achieved by integrating satellite and climate data, the contributions from various sources (Climate, Soil, Socioeconomic, and Remote sensing) remain unclear. The lack of well-defined comparisons between the performance of regression-based approaches and different Machine Learning (ML) methods in yield prediction necessitates further investigation. This study addresses the gaps by combining data from multiple sources to forecast wheat yield in the Multan region in the Punjab province of Pakistan. The findings are compared to the benchmark provided by Crop Report Services (CRS) Punjab, with three widely used ML techniques (support vector machine (SVM), Random Forest (RF), and Least Absolute Shrinkage and Selection Operator (LASSO)) by integrating publicly available data within the GEE (Google Earth Engine) platform, including climate, satellite, soil properties, and spatial information data to develop alternative empirical models for yield prediction using data from 2017 to 2022, selecting the best attribute subset related to crop output. The district-level simulated yield data set was analyzed with three ML models (SVM, RF, and LASSO) as a function of seasonal weather, satellite, and soil. The results indicate that combining all datasets using three ML algorithms achieves better yield prediction performance ( $R^{2}$ : 0.74-0.88). Incorporating spatial information and other properties into benchmark models can improve the prediction from 0.08 to 0.12. Random forest outperformed the competitor models with a Root Mean Square Error (RMSE) of 0.05 q/ha and $R^{2}$ of 0.88. Comparative analysis shows that random forest with 97% and SVM with 93% yielded better results in the study area.

One of the challenges in image processing involves single image deraining (SID). Existing methods do not exploit images at multi-scales and often overlook spatial and channel information. These methods fail to account for different rain conditions, leading to challenges in effectively removing a wide range of rain streak patterns, such as diverse directional or dense streaks. Moreover, they often result in the loss of fine texture details or a blurred background in the process of eliminating rain streaks from images captured in heavy rain. Mostly state-of-the-art (SOTA) deraining models achieve higher performance in removing rain from rainy images but at the expense of a high number of parameters, which results in computational complexity and memory requirements. Furthermore, they also do not consider high-level visioned evaluation metrics to perform deep evaluations of the proposed models. In this paper, a simple lightweight network with few parameters and relatively shallow depth is proposed by fusing the traditional Gaussian-Laplacian pyramid technique with the attention module. We propose a novel attention-based lightweight pyramid network for image deraining and detection (LPN-IDD) to achieve better deraining performance. The proposed model includes a dual attention module integrated with the Gaussian-Laplacian pyramids network. In LPN-IDD, different levels of Laplacian pyramids can extract multi-scale features to adapt to different shapes and types of rain streaks. Residual and recursive blocks are used in each subnetwork along with dual attention blocks to resist the occlusion or texture feature while suppressing unnecessary features. Extensive experimentation performed on the SID synthetic and real-world datasets demonstrate the effectiveness of the proposed model in image deraining tasks. For deep evaluation of the proposed methodology, object detection models including faster-RCNN, YOLO-V3, YOLO-V7 are used along with full reference evaluation metrics.

In recent eras, the complexity and fluctuations of the global crude oil prices have affected the economic progress of society. It is therefore, the oil price prediction has hauled the attention of scholars and policymakers. Driven by this critical concern for forecasting of crude oil prices, we introduces a novel hybrid model keeping in mind the primary objective of enhancing prediction accuracy while considering the specific characteristics as inherent in the data. To achieve this achievement, the trend is eliminated, allowing the scrutiny of whether the residual component validates the assurance of a series ran by stochastic trends. Following the removal of the trend, the residual component undergoes rigorous evaluation through autoregressive model following the decomposition model. Then we got support from the support vector machine, autoregressive integrated moving average and long-short term memory. The predictions accuracy can be evaluated by using the various performance metrics. The proposed hybrid model’s robustness and forecasting performance are rigorously evaluated through Diebold-Mariano test in comparison to competing models. Furthermore, the forecasting ability is evaluated via directional forecast. Ultimately, the empirical findings explicitly determine the superior predictive capabilities of the proposed hybrid model over alternative approaches.

Cybersecurity is important in the realization of various smart grid technologies. Several studies have been conducted to discuss different types of cyberattacks and provide their countermeasures. The false command injection attack (FCIA) is considered one of the most critical attacks that have been studied. Various techniques have been proposed in the literature to detect FCIAs on different components of smart grids. The predominant focus of current surveys lies on FCIAs and detection techniques for such attacks. This article presents a survey of existing works on FCIAs and classifies FCIAs in smart grids according to the targeted component. The impacts of FCIAs on smart grids are also discussed. Subsequently, this article provides an extensive review of detection studies, categorizing them based on the type of detection technique employed.

3-D CNNs have demonstrated their capability to capture intricate nonlinear relationships within hyperspectral images (HSIs). However, the computational complexity of 3-D CNNs often leads to slower processing speeds, limited generalization, and susceptibility to overfitting. In response to these challenges, this study introduces the concept of depthwise separable convolutions using (2+1)-D convolutions as an alternative to traditional 3-D convolutions for hyperspectral image classification (HSIC). The study observes that (2+1)-D convolutions can effectively approximate the complex relationships represented by 3-D convolutions while requiring fewer convolutional operations, thereby reducing the computational overhead associated with classification. Experimental results obtained from benchmark HSI datasets, including Indian Pines, Botswana, Pavia University, and Salinas, demonstrate that the proposed model yields results that are comparable to those achieved by various state-of-the-art models in the existing literature.

The cyber realm is overwhelmed with dynamic malware that promptly penetrates all defense mechanisms, operates unapprehended to the user, and covertly causes damage to sensitive data. The current generation of cyber users is being victimized by the interpolation of malware each day due to the pervasive progression of Internet connectivity. Malware is dispersed to infiltrate the security, privacy, and integrity of the system. Conventional malware detection systems do not have the potential to detect novel malware without the accessibility of their signatures, which gives rise to a high False Negative Rate (FNR). Previously, there were numerous attempts to address the issue of malware detection, but none of them effectively combined the capabilities of signature-based and machine learning-based detection engines. To address this issue, we have developed an integrated Anti-Malware System (AMS) architecture that incorporates both conventional signature-based detection and AI-based detection modules. Our approach employs a Generative Adversarial Network (GAN) based Malware Classifier Optimizer (MCOGAN) framework, which can optimize a malware classifier. This framework utilizes GANs to generate fabricated benign files that can be used to train external discriminators for optimization purposes. We describe our proposed framework and anti-malware system in detail to provide a better understanding of how a malware detection system works. We evaluate our approach using the Figshare dataset and state-of-the-art models as discriminators. Our results showcase enhanced malware detection performance, yielding a 10% performance boost, thus affirming the efficacy of our approach compared to existing models.