Introduction:Early and accurate detection of pumpkin leaf diseases is essential for precision agriculture; however, manual inspection remains slow, subjective, and difficult to scale in real field environments. To address these limitations, this study proposes a robust deep-learning framework for automated pumpkin leaf disease classification. Methods:This study introduces DualFusion-CBAM-Stochastic, a hybrid deep-learning architecture that integrates two complementary convolutional backbones: DenseNet121 for fine-grained texture representation through dense connectivity and EfficientNetB3 for multi-scale contextual feature extraction using compound scaling. Input images are preprocessed through resizing to 224 × 224 pixels, ImageNet-based normalization, and controlled data augmentation, including horizontal and vertical flips, rotation, and zoom. Feature refinement is achieved using the Convolutional Block Attention Module (CBAM), which applies sequential channel and spatial attention, while stochastic-depth regularization improves generalization by randomly bypassing deep layers during training. Results:The proposed model was trained on a balanced dataset of 2,000 images across five pumpkin leaf disease categories. Experimental evaluation using ablation studies and comparative analysis against state-of-the-art models demonstrates that the proposed architecture achieves 96% classification accuracy, outperforming existing CNN-based approaches. Discussion:The results confirm that the synergistic integration of dual-backbone fusion, attention-guided refinement, and stochastic-depth regularization significantly enhances classification performance, feature interpretability, and model stability under diverse visual conditions. These findings advance automated pumpkin leaf disease diagnosis and provide a strong methodological foundation for future research in agricultural image analysis.

Food safety risk assessment is a complex multi-criteria decision-making (MCDM) issue characterized by high uncertainty in both data and expert opinions. Traditional MCDM methods struggle to effectively manage this uncertainty and subjectivity. This paper extends the multi-attributive ideal-real comparative analysis (MAIRCA) method to an uncertain decision-making environment by embedding it within a cubic Pythagorean fuzzy (CuPyF) framework, integrating a variation coefficient similarity measure (VCSM) and the rank-sum (RS) method. Cubic Pythagorean fuzzy sets (CuPyFSs) are used to represent both precise and interval-valued information, enabling better uncertainty modeling. The proposed VCSM objectively determines criteria weights, while the RS method provides subjective weights, leading to balanced comprehensive weighting. The extended CuPyF-based MAIRCA method is then applied to rank alternatives and select the optimal solution. A food safety case study validates the model, demonstrating that it delivers stable, discriminative, and interpretable results, outperforming traditional MCDM models and offering policymakers a reliable and scientific tool for food safety risk management.

The Da Vinci Surgical Systems, a leading platform for minimally invasive robotic surgeries that relies on real-time data fusion from several sensors to increase precision, security, and efficiency. Threats including tampering, delay, and unauthorized access could jeopardize the integrity, security, and dependability of this crucial data, which could have a negative impact on patient outcomes in terms of treatment. This paper proposes a novel framework to integrate Hyperledger Sawtooth, a blockchain modular infrastructure propose by Linux. It collaborates with the data fusion procedure in order to maintain the current environment of Da Vinci platform. The proposed solution ensures secure and immutable data sharing and exchange across interconnected components, including robotic arms, haptic feedback systems, and imaging equipment, by leveraging a blockchain consortium network. Technology like Hyperledger Sawtooth, is used because of its flexible modular hierarchy due to the predefined consensus mechanism and protocol design, which allow for low-latency operations that are essential for real-time surgical applications.

Nail diseases, including such common conditions as fungus, and more serious issues like melanoma, may be important clues to the overall health and require a clear diagnosis to be treated. The purpose of the paper is to create a nail disease detection system based on the advanced machine learning methods, including transfer learning and federated learning. The research seeks to show how machine learning and federated learning can be combined to detect nail disease performance with high accuracy without having to share data. The data include pictures of diverse nail conditions including Acral Lentiginous Melanoma, Onychogryphosis, and Pitting among others that are checked to maintain the quality of data in a uniform manner to facilitate the effective training of the models. The most common feature extraction models are ResNet152V2, DenseNet201, MobileNetV2, and InceptionResNetV2 that produce between 1,280 and 2,048 features per image. These characteristics are then pooled to create a unified feature space of 6,784 dimensions which is further narrowed to five major characteristics with Linear Discriminant Analysis (LDA) to create an efficient form of classification. A range of classification models, including Deep Neural Networks (DNN) and Long Short-Term Memory (LSTM) and Bidirectional LSTM (Bi-LSTM) are compared, with the last one reaching the highest classification accuracy of 91.8%. The federated learning strategy enables the joint training of DL models by different clients to ensure data-privacy and has validation-accuracy rates exceeding 99-percent in both uniformly random and structured data distributions. The proposed federated learning-based models resulted high in both uniformly random and structured data distributions.

Skin cancer is one of the most common and fatal malignancies worldwide, and reliable, early diagnosis systems are required to provide effective clinical intervention. Nevertheless, even with recent progress in deep learning, current models tend to overfit and lack generalizability and robustness across various lesion types. This paper overcomes these challenges by presenting a late-fusion ensemble of pre-trained convolutional neural networks (CNNs) integrated with Diverse Convolution Networks (DCNs) to build a complementary and discriminative skin cancer detection framework from dermoscopic images. The study utilized a curated dataset of 10,600 dermoscopic images from the ISIC Archive, comprising 9600 images for training and 1000 images for testing, encompassing both malignant and benign lesion classes. The proposed late-fusion model combines the predictions of several pre-trained architectures, leveraging their strengths while mitigating bias. Experimental results show excellent performance with an accuracy of 99.7%, an F1-score of 99.75%, a precision of 99.12%, and a recall of 99.34% on the test set, and an accuracy of 99.1% and an F1-score exceeding state-of-the-art models of 99.8% on the melanoma dataset. The conclusions validate the robustness and diagnostic accuracy of the framework in complex clinical situations. Future work will focus on model explainability, integration with clinical decision support systems, and validation using larger, multi-source datasets to enhance real-world applicability.

We propose TourOptiGuide, a hybrid tourism recommendation prototype that integrates deep learning and fuzzy logic for context-aware point-of-interest (POI) suggestion. The system generates next-POI recommendations by jointly considering four factors: the tourist’s current location, inferred preferences, estimated age, and historical visit context. Deep learning models are employed to detect POIs from images and estimate tourist age from facial images, while a trajectory data warehouse structures historical visit information for contextual filtering. A fuzzy inference system is used to combine these heterogeneous inputs and produce interpretable recommendation decisions under uncertainty. The proposed framework is evaluated as a proof-of-concept prototype, demonstrating coherent system behavior and functional integration of perception, historical context, and fuzzy reasoning.

The increased use of AI in SQA requires a strict analysis of AI models in TC generation. This paper examines the factuality of the advanced reasoning of Open AI (O1 series) being the distinguishing factor between Open AI and non reasoning models GPT 4o and human Quality Assurance (QA) engineers. To address this, we evaluated the performance of 12 experimental groups: human testers, eight variants of GPT 4o which used a variety of prompt engineering techniques (e.g., chain of thought, few shot, self consistency), and three variants of the O1 model in the generation of TCs on 11 software applications. The 3682 TCs were tested with the mixtral-8x7B-32768 model according to four criteria: Coverage, Clarity, ENC and NFC. The descriptive statistics indicated that there was consistently high coverage and high clarity among the groups but there was a significant difference in the capture of edge, negative and non-functional scenarios. Statistical analyses using both parametric and non parametric methods identified significant differences across several TC quality metrics depending on prompt ing strategy and underlying model architecture. Interestingly, AI models not only produced more TCs but also had a better QTQ ratio than human engineers under the conditions of receiving little contextual information. Finally, reasoning models have more structured outputs but may not yield significant performance improvement given the increased computational costs. Enhanced prompt engineering enables non-reasoning models to offer an efficient and effective alternative for TC generation, thereby underscoring the potential for AI integration in optimizing software testing processes. To facilitate replication and further research, an extensive dataset and fully documented codebase are available at the GitHub repository provided in the Contributions section.

Scientific workflows are crucial for handling extensive datasets and facilitating large-scale scientific research. They are time-consuming and resource-intensive applications, rendering dispersed technologies like cloud computing particularly appropriate for their implementation. Nonetheless, cloud systems provide distinct issues, especially in task scheduling and data location, which must be resolved for optimal workflow execution. Despite much research on workflow scheduling, the integrated optimization of task scheduling and data placement requires additional literature research. This research introduces an innovative scheduling framework that concurrently tackles task scheduling and data placement for scientific workflows in cloud environments. The proposed work incorporates a genetic algorithm for optimizing scheduling and data placement, combined with a fuzzy data placement strategy utilizing the interval type-2 fuzzy C-means clustering method, which adeptly addresses data uncertainty in cloud storage. The suggested scheduler significantly minimizes data transmission time and enhances overall workflow performance by dynamically coordinating data allocation with task execution. This study's optimization methodology integrates evolutionary computation with fuzzy clustering. It offers a more comprehensive and adaptable method for managing scientific workflows in the cloud. The proposed scheduler was executed using a simulated cloud environment on real-world workflows, such as Epigenomics and LIGO from Pegasus. The experimental results show that the proposed approach reduces data movements significantly compared to the literature.

Conjunctivitis is one of the most common eye disorders, and timely diagnoses are crucial to ensure no complications arise from such disorders. Manually examining the images to diagnose the disease often takes too much time and may result in inconsistent diagnosis. To overcome this problem, this paper presents a framework referred to as MOGONET, which is based on deep learning (DL) techniques and involves image preprocessing, image segmentation, data augmentation, transfer learning, and multi-objective optimization for diagnosing conjunctivitis eye disease. To increase the quality of the image for analysis, image contrast enhancements are employed. Furthermore, image multi-threshold segmentation techniques are used to identify vital areas of the eye. Also, an advanced Generalized Normal Distribution Optimization (GNDO) algorithm combined with a CNN is used to improve the efficiency and minimize computational complexity. The proposed model is tested using conjunctivitis eye image dataset from Kaggle and Shutterstock. According to the result findings, the accuracy, precision, recall, and F-score of the proposed model is 98.32

Multiclass classification in educational data mining presents persistent challenges including class imbalance, miscalibrated probability outputs, and insufficient statistical validation. We proposed RXK-VEM, a hybrid ensemble framework that integrates random forest (RF), extreme gradient boosting (XGBoost), and K-nearest neighbors (KNN) through a formally defined vote-entropy-weighted meta-fusion (VEM) operator, followed by meta-level calibration using multinomial logistic regression. The VEM operator is defined as a mapping on the probability simplex triangle C-1, aggregating heterogeneous base learner outputs into a unified probabilistic representation with provable closure properties. We further established a Rademacher complexity-based generalization bound showing that operating in the compressed C-dimensional probability space (C << d) tightens the generalization gap relative to classifiers trained directly on raw features, providing theoretical justification for the stacking architecture. We validated RXK-VEM on two structurally distinct educational datasets: A primary academic performance dataset (N = 560, five classes) from Universiti Teknologi Malaysia and a secondary student dropout dataset (N = 4,424, three classes) from the University of California, Irvine (UCI) repository. On the primary dataset, RXK-VEM achieves 91.07% accuracy, 91.22% precision, and an 86.21% Matthews correlation coefficient (MCC), outperforming all individual base learners and conventional ensemble strategies. On the secondary dataset, the model achieves 77.30% accuracy and a 62.33% MCC, maintaining competitive performance across all metrics. Statistical validation through five-fold stratified cross-validation, paired t-tests, and Wilcoxon signed-rank tests confirms that improvements over weaker baselines are consistent and not attributable to random variation. A systematic ablation study quantifies the complementary contribution of each base learner, and Shapley additive explanations analysis validates the interpretability of the identified predictors. The proposed framework offers a mathematically rigorous, empirically validated, and interpretable architecture for probabilistic ensemble integration in multiclass educational prediction tasks.

Sustainable energy systems will entail a change in the carbon intensity projections, which should be carried out in a proper manner to facilitate the smooth running of the grid and reduce greenhouse emissions. The present article outlines the TransCarbonNet, a novel hybrid deep learning framework with self-attention characteristics added to the bidirectional Long Short-Term Memory (Bi-LSTM) network to forecast the carbon intensity of the grid several days. The proposed temporal fusion model not only learns the local temporal interactions but also the longterm patterns of the carbon emission data; hence, it is able to give suitable forecasts over a period of seven days. TransCarbonNet takes advantage of a multi-head self-attention element to identify significant temporal connections, which means the Bi-LSTM element calculates sequential dependencies in both directions. Massive tests on two actual data sets indicate much improved results in comparison with the existing results, with mean relative errors of 15.3 percent and 12.7 percent, respectively. The framework has given explicable weights of attention that reveal critical periods that influence carbon intensity alterations, and informed decisions on the management of carbon sustainability. The effectiveness of the proposed solution has been validated in numerous cases of operations, and TransCarbonNet is established to be an effective tool when it comes to carbon-friendly optimization of the grid.

Early detection of depression from social media text requires models that balance predictive accuracy with interpretability. This paper presents an explainable transformer-based framework fine-tuning BERT with CNN and BiLSTM extensions for detecting depressive cues in online language. Using a Twitter dataset of 7,769 annotated posts, three architectures are evaluated: baseline BERT, BERT–CNN, and BERT–BiLSTM. The BERT–CNN model achieves the best performance with 91.81% accuracy and F1-score of 0.9185. LIME-based explanations confirm that predictions are grounded in semantically meaningful depressive cues rather than surface-level keywords, supporting transparent early mental-health monitoring.

Accurate multi-step building load forecasting is critical for optimizing energy management in smart grids and reducing operational costs. However, existing forecasting methods struggle with complex temporal dependencies, seasonal variations, and robust performance under noisy conditions. This paper proposes R-SATNet (Robust Self-Attention Transformer Network), a novel deep learning architecture that integrates multi-head self-attention mechanisms with robust optimization techniques for enhanced building load prediction. The proposed framework incorporates temporal feature extraction modules, adaptive noise suppression layers, and multi-scale attention blocks to capture both short-term fluctuations and long-term seasonal patterns. Extensive experiments on real-world building load datasets demonstrate that R-SATNet achieves superior forecasting accuracy with 15.7% lower RMSE and 12.3% improved MAPE compared to state-of-the-art methods. The model maintains robust performance under various noise conditions and provides reliable multi-step predictions up to 24 h ahead, making it highly suitable for practical smart energy system deployments. The proposed framework is validated across six diverse building datasets spanning commercial, residential, industrial, campus, mixed-use, and healthcare facilities, confirming its generalizability and practical applicability in heterogeneous smart energy environments.

Background: Volumetric three-dimensional (3D) magnetic resonance imaging (MRI) data are inherently complex, making feature extraction a significant challenge. The large volume of information in 3D MRI requires advanced algorithms to efficiently extract relevant features, which is crucial for identifying abnormal brain activity such as epileptic seizures. Methods: We propose a novel 3D methodology named the Cubic Decimal Descriptor Pattern (C-DDP) for enhanced feature extraction from 3D MRI data. To evaluate its effectiveness, we conducted experiments using a publicly available 3D MRI dataset comprising 85 individuals with focal cortical dysplasia type II (FCD II) and 85 healthy controls. Both 3D Fluid Attenuated Inversion Recovery (FLAIR) and T1-weighted isotropic image sequences were processed, and features were analyzed using various machine learning classifiers. Results: C-DDP consistently demonstrated superior performance in extracting discriminative features across different classifiers. The method improved the detection of FCD II and enabled more accurate identification of epileptogenic lesions compared to conventional feature extraction approaches. Conclusions: The proposed C-DDP approach represents a notable advancement in 3D MRI feature extraction, offering potential for enhanced diagnostic accuracy in patients with epileptogenic lesions. This methodology may contribute to more effective clinical decision-making and improved patient outcomes.

Cerebral palsy is a prevalent neurodevelopmental syndrome that disrupts motor development in children, making early detection vital for effective intervention. Traditional clinical assessments rely on subjective observations, often missing minor motor abnormalities until they become severe, typically after 12 months of age. This article presents a novel deep learning model, TransCP-Net (Transformer-based Cerebral Palsy Network), designed for early detection of infant cerebral palsy through spatiotemporal pose representation learning. The architecture employs hierarchical spatial and temporal attention to analyze complex motion patterns in video sequences, integrating multi-modal data for improved accuracy. TransCP-Net incorporates specialized preprocessing, including temporal smoothing and trajectory encoding, to enhance feature learning. Tests on 1370 infant movement videos yielded impressive results: 94.7% sensitivity, 92.3% specificity, and an AUC-ROC of 0.968, outperforming ten state-of-the-art methods. Notably, it achieved a sensitivity of 96.3% within the critical 9-15 weeks range of fidgety movements, enabling timely interventions. Attention visualization highlights key areas such as the hips and shoulders, reinforcing clinical relevance. TransCP-Net demonstrates effectiveness across diverse clinical settings, serving as a viable, non-invasive tool for early cerebral palsy detection.

IntroductionHeart disease is a leading cause of death worldwide, necessitating accurate early diagnosis. Although machine learning (ML) shows potential for this task, many current models are hindered by data inconsistencies, poor feature selection, and limited robustness.MethodsThis study proposes a novel, robust diagnostic framework. It employs advanced data preprocessing using Padding Interpolation for missing values, Median Imputation for outliers, and RobustScalar for scaling to ensure data integrity. A key innovation is an Improved Particle Swarm Optimization (IPSO) algorithm, enhanced with dynamic inertia weight and a mutation operator to avoid premature convergence. This IPSO performs dual optimization: selecting optimal features and tuning the hyperparameters of five classifiers (Logistic Regression, Linear Discriminant Analysis, Gaussian Naïve Bayes, Support Vector Classifier, and XGBoost).ResultsThe framework was evaluated on a composite dataset from five public repositories. The proposed IPSO-optimized XGBoost model achieved superior performance at a 90:10 train-test split, with an accuracy of 91.3%, sensitivity of 88.37%, specificity of 93.88%, precision of 92.68%, F1-score of 90.48%, and a Diagnostic Odds Ratio of 116.53. Statistical tests (p < 0.05) confirmed these improvements over baselines were significant. The model also demonstrated consistent generalizability on independent Cleveland and Statlog datasets.DiscussionThe results establish that the integrated framework of rigorous preprocessing and the hybrid IPSO optimization-classification model creates a highly effective and generalizable pipeline for automated heart disease diagnosis, addressing key limitations of existing approaches.

The analysis of Android malware shows that this threat is constantly increasing and is a real threat to mobile devices since traditional approaches, such as signature-based detection, are no longer effective due to the continuously advancing level of sophistication. To resolve this problem, efficient and flexible malware detection tools are needed. This work examines the possibility of employing deep CNNs to detect Android malware by transforming network traffic into image data representations. Moreover, the dataset used in this study is the CIC-AndMal2017, which contains 20,000 instances of network traffic across five distinct malware categories: a. Trojan, b. Adware, c. Ransomware, d. Spyware, e. Worm. These network traffic features are then converted to image formats for deep learning, which is applied in a CNN framework, including the VGG16 pre-trained model. In addition, our approach yielded high performance, yielding an accuracy of 0.92, accuracy of 99.1%, precision of 98.2%, recall of 99.5%, and F1 score of 98.7%. Subsequent improvements to the classification model through changes within the VGG19 framework improved the classification rate to 99.25%. Through the results obtained, it is clear that CNNs are a very effective way to classify Android malware, providing greater accuracy than conventional techniques. The success of this approach also shows the applicability of deep learning in mobile security along with the direction for the future advancement of the realtime detection system and other deeper learning techniques to counter the increasing number of threats emerging in the future.

The critical need for early and precise detection of dementia, a crippling cognitive illness that primarily affects women, is addressed by this study. Global healthcare systems face enormous hurdles as dementia becomes more common. The need for non-invasive and effective alternatives is highlighted because current diagnostic techniques are frequently invasive, expensive, and imprecise. To address this issue, our work presents a unique method for female dementia identification from speech using multimodal Siamese networks. In contrast to earlier models, our approach uses both transcript and audio data, utilizing the complementary information present in both modalities. Improving dementia detection accuracy and reliability is the main driving force for this study, particularly in the early stages when intervention can be more successful. Additionally, the information used in this study includes 104 people in the control group, 208 people with a dementia diagnosis, and 85 whose diagnosis is uncertain. There are 238 control files and 298 dementia files in the audio dataset, and 243 control files and 306 dementia files in the transcript database. This extensive dataset makes it possible to evaluate our suggested model with confidence. Moreover, multimodal Siamese networks—a cutting-edge technique that captures relationships between multimodal data—are a part of the basic methodology used. Our model has a greater accuracy of 99% on the Dementia Bank Database, demonstrating considerable improvements over earlier approaches. The assessment parameters, encompassing an Area Under the Curve (AUC) of 0.99, bolster the efficacy of our methodology. This paper will improve at-risk individuals’ quality of life by developing non-intrusive dementia detection for early diagnosis and intervention.

As one of the major threats to women's health worldwide, breast cancer requires early diagnosis and accurate classification, since they are key to optimizing therapeutic interventions and ensuring precise prognosis. Recently, deep learning has demonstrated notable advantages in breast cancer image classification. However, their performance heavily relies on the proper configuration of hyperparameters. To overcome the inefficiencies and weaknesses of conventional hyperparameter optimization methods, like limited effectiveness and vulnerability to premature convergence, this research proposes a Multi-Strategy Parrot Optimizer (MSPO) and applies it to breast cancer image classification tasks. Based on the original Parrot Optimizer, MSPO integrates several strategies, including Sobol sequence initialization, nonlinear decreasing inertia weight, and a chaotic parameter to enhance global exploration ability and convergence steadiness. Tests using the CEC 2022 benchmark functions reveal that MSPO surpasses leading algorithms regarding optimization precision and convergence rate. An ablation study was conducted on three variants of MSPO through CEC 2022 to further validate the effectiveness of each key strategy. Furthermore, MSPO is combined with the ResNet18 model and applied to the BreaKHis breast cancer image dataset. Results indicate that the model optimized by MSPO notably surpasses both the non-optimized version and other alternative optimization algorithms using four assessment indicators: accuracy, precision, recall, and F1-score. This validates the promising application potential and practical significance of MSPO in medical image classification tasks.

This study aims to develop a lightweight convolutional network for the classification of multiple retinal diseases using fundus images and to evaluate cross-dataset generalization under strict label alignment. We introduce OcuMDNet (Ocular Multi-Disease Net), a compact convolutional neural network (CNN) specifically designed for fundus imagery, incorporating batch normalization and dropout for regularization. A standardized processing pipeline is employed, which includes cropping and resizing images to 224 × 224 pixels, applying contrast-limited adaptive histogram equalization (CLAHE), and performing per-channel normalization. The training process utilizes the AdamW optimizer and incorporates early stopping to enhance model performance. We propose a label-aligned evaluation protocol: (i) assesses 4-class performance (Normal, diabetic retinopathy (DR), Glaucoma, age-related macular degeneration (AMD)) on a combined dataset assembled from public sources; (ii) reports disease-specific results based on the native labels of each dataset (DR: EyePACS, Messidor; Glaucoma: ORIGA; AMD: AREDS); and (iii) evaluates cross-dataset transfer for DR (training on EyePACS and testing on Messidor). Patient-level splits are implemented to prevent data leakage, and class counts are reported for each split. Performance metrics such as accuracy, macro-F1 score, and one-vs-rest ROC-AUC are calculated with 95 % confidence intervals using stratified bootstrap (n = 1000). Paired comparisons are conducted using McNemar's test for accuracy and DeLong's method for AUC, with multiplicity control applied. The OcuMDNet demonstrates strong performance on both combined and disease-specific benchmarks, maintaining robust discrimination in cross-dataset evaluations for DR while ensuring computational efficiency suitable for large-scale screening applications. Ablation studies confirm the significance of preprocessing steps and architectural choices. In conjunction with a label-aligned protocol, the OcuMDNet provides an accurate and efficient baseline for multi-disease fundus analysis, facilitating a transparent assessment of cross-dataset reliability. The code, scripts, and trained weights will be made available to support reproducibility.