Herbify: an ensemble deep learning framework integrating convolutional neural networks and vision transformers for precise herb identification

The accuracy of a classifier, whether it is an ensemble or not, is directly influenced by the training data used in learning. In remote sensing, training data mislabeling is inevitable and faces a major challenge. This paper proposes a versatile data cleaning which handles the mislabeling problem by exploiting the ensemble concepts for identifying, then eliminating or correcting the mislabeled training data. A powerful ensemble method, random forest, is at the core of our filter design and helps to distinguish mislabeled data from uncorrupted data more accurately. The major contribution of this work lies on the explicit use of the hypothesis margin as a decision means to identify and eliminate or correct mislabeled training data in an ensemble learning framework. Another key development that makes our algorithm superior to existing approaches is a design that avoids rare class instances to be mistaken for class noise. This fundamental aspect makes our data cleaning system particularly suitable for remote sensing classification tasks which usually suffer from both mislabeling and imbalance problems. The effectiveness of our algorithm is demonstrated in performing mapping of land covers. The generalization performance of two major supervised noise-sensitive classifiers, boosting and K-nearest neighbors, is strengthened by effective class noise reduction. A comparative analysis is conducted with respect to random forest, deep convolutional neural networks, as well as two well-established ensemble-based class noise filters, the majority vote and the consensus vote filters. This analysis demonstrates that our approach is more accurate than deep convolutional neural networks (one-dimensional CNN, AlexNet, EfficientNet, ResNet50 and ShuffletNet) and the reference ensemble methods.

Social networks are popular for advertising, idea sharing, and opinion formation. Due to COVID-19, coronavirus information disseminated on social media affects people’s lives directly. Individuals sometimes managed it well, but it often hampered daily activities. As a result, analyzing people’s feelings is important. Sentiment analysis identifies opinions or sentiments from text. In this paper, we present an effective model that leverages the benefits of Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) to categorize Arabic tweets using a stacked ensemble learning model. First, the tweets are represented as vectors using a word embedding model, then the text feature is extracted by CNN, and finally the context information of the text is acquired by BiLSTM. Aravec, FastText, and ArWordVec are employed separately to assess the impact of the word embedding on the our model. We also compare the proposed method to various deep learning models: CNN, LSTM, and BiLSTM. Experiments are performed on three different Arabic datasets related to COVID-19 and vaccines. Empirical findings show that the proposed model outperformed the other models’ results by achieving F-measures of 76.76%, 87.%, and 80.5% on the SenWave, AraCOVID19-SSD, and ArCovidVac datasets, respectively.

The leather industry plays a crucial role in the production of high-quality leather products to remain competitive in the market. Due to various reasons such as material handling and turning, leather defects can appear at different stages of the production process. This paper proposed deep learning techniques, specifically transfer learning, deep convolutional neural networks (D-CNN), and ensemble learning for automating leather defect detection and classification. The dataset was taken from the open source Kaggle. The dataset was preprocessed from the beginning. However, we performed Error Level Analysis (ELA) to detect possible changes or fraud in the image. We also used data augmentation techniques (resizing, rescaling, flipping, rotation, zooming, and contrasting) on the dataset to improve the model performance. This study investigated deep learning architectures for automatic leather defect detection, comparing them with seven deep convolutional neural networks (D-CNN) models and transfer learning architectures. Furthermore, we introduce and evaluate two novel ensemble models: MER built on transfer learning architecture (MobileNetV2, EfficientNetB0, and ResNet50) and AZL built on D-CNN architecture (AlexNet, ZfNet, and LeNet). A key contribution of this study is the demonstration of significant performance improvements achieved by strategically integrating squeeze and excitation and label smoothing techniques into transfer learning, D-CNN, and ensemble frameworks. Our proposed MER ensemble method achieves a state-of-the-art of 96.39% for leather defect detection. The proposed model was evaluated using performance metrics including precision, recall, F1-score, AUC, ROC, and Matthews Correlation Coefficient (MCC). Additionally, Grad-CAM was employed for explainable AI to visualize and interpret the model’s decision-making process. Comparative analysis against separate transfer learning and D-CNN models fails to establish the superior performance of our MER ensemble, which highlights the potential to significantly advance automated quality inspection in leather production. This study provides insights into evaluating the best deep learning techniques for leather defect classification, paving the way for more accurate, efficient, and reliable industrial applications. Proposes a robust deep learning-based framework for automated leather defect detection using transfer learning, deep CNNs, and two novel ensemble models (MER and AZL), enhanced with squeeze-and-excitation and label smoothing techniques. Achieves state-of-the-art performance with the proposed MER ensemble model, achieve 96.39% accuracy, and demonstrates superior results across precision, recall, F1-score, AUC–ROC, and MCC. Integrates Error Level Analysis (ELA), extensive data augmentation, and Grad-CAM–based explainable AI to ensure data reliability, improve generalization, and provide transparent visual interpretation of defect regions for industrial quality inspection.

Although many machine learning methods have been widely used to predict PM2.5 concentrations, these single or hybrid methods still have some shortcomings. This study integrated the advantages of convolutional neural network (CNN) feature extraction and the regression ability of random forest (RF) to propose a novel CNN-RF ensemble framework for PM2.5 concentration modeling. The observational data from 13 monitoring stations in Kaohsiung in 2021 were selected for model training and testing. First, CNN was implemented to extract key meteorological and pollution data. Subsequently, the RF algorithm was employed to train the model with five input factors, namely the extracted features from the CNN and spatiotemporal factors, including the day of the year, the hour of the day, latitude, and longitude. Independent observations from two stations were used to evaluate the models. The findings demonstrated that the proposed CNN-RF model had better modeling capability compared with the independent CNN and RF models: the average improvements in root mean square error (RMSE) and mean absolute error (MAE) ranged from 8.10% to 11.11%, respectively. In addition, the proposed CNN-RF hybrid model has fewer excess residuals at thresholds of 10 μg/m3, 20 μg/m3, and 30 μg/m3. The results revealed that the proposed CNN-RF ensemble framework is a stable, reliable, and accurate method that can generate superior results compared with the single CNN and RF methods. The proposed method could be a valuable reference for readers and may inspire researchers to develop even more effective methods for air pollution modeling. This research has important implications for air pollution research, data analysis, model estimation, and machine learning.

Accurate, early diagnosis of dental diseases such as caries, gingivitis, and tooth discoloration is critical for effective treatment and the prevention of long-term complications. Despite the convolutional neural networks’ (CNNs) potential in medical image analysis, single-model methodologies frequently exhibit insufficient resilience and generalizability. Adopting a late-fusion strategy, this study integrates DenseNet121 and InceptionV3 into an ensemble deep learning framework for multi-class dental disease classification. The convolutional bases of both models are partially frozen to maintain most of the layers as nontrainable and thus preserve the low- and mid-level feature representations. The last 50 layers are unfrozen and fine-tuned to adapt to task-specific dental features. Custom classification heads comprising of global average pooling, dense layers with corrected linear unit activation, batch normalization, dropout regularization, and a SoftMax output were optimized for the target task. To enhance its robustness, the framework incorporates extensive data augmentation, class balancing with weights, and focal loss as an optional optimization strategy. The impact of architectural complexity on model performance was systematically evaluated across classification heads with varying depths (1–4 dense layers). On both validation and test datasets the proposed ensemble consistently outperformed the individual models in terms of accuracy, precision, recall, and F1-score, achieving an impressive accuracy of 88.24%, a recall of 88.24%, a precision of 88.67%, and an F1-score of 88.32% on the test set. Furthermore, explainable artificial intelligence (XAI) was incorporated using Gradient-weighted Class Activation Mapping, providing interpretable visual explanations of the discriminative regions contributing to the predictions. Besides enhancing clinical trust in the framework, these visual insights validate the ensemble’s focus on pathology-relevant areas within dental images. Combining the complementary strengths of DenseNet121 and InceptionV3, systematic architectural exploration, and XAI, the proposed framework is a potentially reliable and interpretable tool for computer-assisted dental diagnostics and advanced dental-image classification.

Agricultural crop mapping has advanced over the last decades due to improved approaches and the increased availability of image datasets at various spatial and temporal resolutions. Considering the spatial and temporal dynamics of different crops during a growing season, multi-temporal classification frameworks are well-suited for mapping crops at large scales. Addressing the challenges posed by imbalanced class distribution, our approach combines the strengths of different deep learning models in an ensemble learning framework, enabling more accurate and robust classification by capitalizing on their complementary capabilities. This research aims to enhance the crop classification of maize, soybean, and wheat in Bei’an County, Northeast China, by developing a novel deep learning architecture that combines a three-dimensional convolutional neural network (3D-CNN) with a variant of convolutional recurrent neural networks (ConvRNN). The proposed method integrates multi-temporal Sentinel-1 polarimetric features with Sentinel-2 surface reflectance data for multi-source fusion and achieves an overall accuracy of 91.7%, a Kappa coefficient of 85.7%, and F1 scores of 93.7%, 92.2%, and 90.9% for maize, soybean, and wheat, respectively. Our proposed model is also compared with alternative data augmentation techniques, maintaining the highest mean F1 score (87.7%). The best performer was weakly supervised with ten per cent of ground truth data collected in Bei’an in 2017 and used to produce an annual crop map for measuring the model’s generalizability. The model learning reliability of the proposed method is interpreted through the visualization of model soft outputs and saliency maps.

Predictive models for early identification of heart disease must be precise and efficient because it is a major worldwide health concern. To improve classification performance while protecting data privacy, this study investigated a combined method that uses ensemble learning, feature selection, and federated learning (FL). The ensemble-based approaches proved the most predictive after testing several different machine learning (ML) models, including random forests, the light gradient boosting machine, support vector machines, k-nearest neighbors, convolutional neural networks, and long short-term memory. We used particle swarm optimization (PSO) for feature selection, which optimized the most relevant features in conjunction with voting and stacking approaches to further increase the model's performance. In addition, federated learning was implemented to allow decentralized training while preserving sensitive medical data. The results highlight the effectiveness of combining these techniques in the detection of heart disease, providing a scalable and privacy-preserving solution for real-world healthcare applications. Two benchmark datasets were used to validate the proposed approach, ensuring the reliability and generalizability of the findings. Furthermore, we used four performance metrics, namely accuracy, precision, recall, and F1score, to evaluate the selected models. Finally, federated learning was included to handle privacy issues and guarantee safe access to private medical data. This distributed method allows model training without centralizing patient data, so it is compatible with strict data privacy rules. With up to 95% precision, our method shows a notable increase in prediction accuracy according to the testing results. This work offers a strong, scalable, and safe solution for the early identification of cardiovascular diseases by combining ensemble learning, feature selection, and federated learning, opening the way for more general uses in medical diagnostics.

The transformation of traditional energy systems into smart energy systems has ushered in an era of efficiency, sustainability and technological growth. In this paper, we propose a new definition for “Quality of Energy Service” that focuses on ensuring optimal power-supply quality, encompassing factors such as availability, speed (i.e., the time to restore or adjust supply following interruptions or load changes) and reliability of supply. We explore the integration of advanced algorithms specifically tailored to enhance the Quality of Energy Services. By concentrating on key aspects—reliability, availability and operational efficiency—the study reviews how various algorithmic approaches, from machine learning models to classical optimisation techniques, can significantly improve power grid management. These algorithms are evaluated for their potential to optimise load distribution, predict system failures and manage real-time adjustments in power supply, thereby ensuring higher service quality and grid stability. The findings aim to provide actionable insights for policymakers, engineers and industry stakeholders seeking to advance smart grid technologies and meet global energy standards. Furthermore, we present a case study to demonstrate how these models can be integrated to optimise grid management, forecast energy demand and enhance operational efficiency. We employ multiple machine learning models—including Random Forest, XGBoost version 1.6.1 and Long Short-Term Memory (LSTM) networks—to predict future energy demand. These models are then combined within an ensemble learning framework to improve both the accuracy and robustness of the forecasts. Our ensemble framework not only predicts energy consumption but also optimises battery storage utilisation, ensuring continuous energy availability and reducing reliance on external energy sources. The proposed stacking ensemble achieved a forecasting accuracy of 99.06%, with a Mean Absolute Percentage Error (MAPE) of 0.9364% and a Coefficient of Determination (R2) of 0.998345, highlighting its superior performance compared to each individual base model.

Control Chart Pattern Recognition (CCPR) plays a crucial role in maintaining product quality. This paper presents some recognition models that utilizes a custom convolutional network alongside pretrained VGG19, MobileNet, and LeNet networks to identify Control Chart Patterns (CCP) for autocorrelated processes. The suggested architectures autonomously extract features from input data, in contrast to conventional methods that necessitate manual feature engineering. This study addresses the challenge of training deep networks with insufficient training data by employing transfer learning to refine a VGG19, MobileNet, and LeNet model, for a new CCPR task. The comparison of the performance between pre-trained networks and the extended convolutional network as a 2D CNN, which does not utilize transfer learning, indicates that pre-trained networks attain superior recognition accuracies with a smaller training data. On the other hand, the pretrained VGG19 demonstrates superior performance when compared to 1D CNN and conventional machine learning techniques, highlighting the advantages of utilizing transfer learning. Additionally, the utilization of pre-trained models addresses the challenge of the intricate design involved in the feature extraction component of deep networks that possess numerous hyperparameters. The suggested method has demonstrated significant potential for identifying CCPs, as evidenced by comparative experimental findings and a real-world case study.

Waste detection plays an essential role in ensuring efficient waste management. Convolutional Neural Networks are used in visual waste detection to improve waste management. This study uses a data set that covers various categories of waste, such as plastic, paper, metal, glass, trash, and cardboard. Convolutional Neural Networks are created and trained with refined architecture to achieve precise classification results. During the model development stage, the focus is on utilizing transfer learning techniques to implement Convolutional Neural Networks. Utilizing pre-trained models will speed up and improve the learning process by enriching the representation of waste features. By using the information embedded in the trained model, the Convolutional Neural Network can differentiate the specific attributes of various waste categories more accurately. Utilizing transfer learning allows models to adapt to real-world scenarios, thereby improving their ability to generalize and accurately identify waste that may exhibit significant variation in appearance. Combining these methodologies enhances the ability to identify waste in diverse environmental conditions, facilitates efficient waste management, and can be adapted to contemporary needs in environmental remediation. The model evaluation shows satisfactory performance, with a recognition accuracy of about 73%. Additionally, experiments are conducted under authentic circumstances to assess the reliability of the system under realistic circumstances. This study provides a valuable contribution to the advancement of waste detection systems that can be integrated into waste management with optimal efficiency.

Automated accurate emotion recognition system using rhythm-specific deep convolutional neural network technique with multi-channel EEG signals

Numerous deep learning architectures have been designed for a variety of tasks in the past few years. However, it is almost impossible for one model to work well for all kinds of scenarios and datasets. Therefore, we present an ensemble deep learning framework in this paper, which not only decreases the information loss and over-fitting problems caused by single models, but also overcomes the imbalanced data issue in multimedia big data. First, a suite of deep learning algorithms are utilized for deep feature selection. Thereafter, an enhanced ensemble algorithm is developed based on the performance of each single Support Vector Machine classifier on each deep feature set. We evaluate our proposed ensemble deep learning framework on a large and highly imbalanced video dataset containing natural disaster events. Experimental results demonstrate the effectiveness of the proposed framework for semantic event detection, and show how it outperforms several state-of-the-art deep learning architectures, as well as handcrafted features integrated with ensemble and non-ensemble algorithms.

An optimized ensemble learning framework for lithium-ion Battery State of Health estimation in energy storage system

Accumulating evidences have indicated that microRNA (miRNA) plays a crucial role in the pathogenesis and progression of various complex diseases. Inferring disease-associated miRNAs is significant to explore the etiology, diagnosis and treatment of human diseases. As the biological experiments are time-consuming and labor-intensive, developing effective computational methods has become indispensable to identify associations between miRNAs and diseases. We present an Ensemble learning framework with Resampling method for MiRNA-Disease Association (ERMDA) prediction to discover potential disease-related miRNAs. Firstly, the resampling strategy is proposed for building multiple different balanced training subsets to address the challenge of sample imbalance within the database. Then, ERMDA extracts miRNA and disease feature representations by integrating miRNA-miRNA similarities, disease-disease similarities and experimentally verified miRNA-disease association information. Next, the feature selection approach is applied to reduce the redundant information and increase the diversity among these subsets. Lastly, ERMDA constructs an individual learner on each subset to yield primitive outcomes, and the soft voting method is introduced for making the final decision based on the prediction results of individual learners. A series of experimental results demonstrates that ERMDA outperforms other state-of-the-art methods on both balanced and unbalanced testing sets. Besides, case studies conducted on the three human diseases further confirm the ERMDA's prediction capability for identifying potential disease-related miRNAs. In conclusion, these experimental results demonstrate that our method can serve as an effective and reliable tool for researchers to explore the regulatory role of miRNAs in complex diseases.

Vibrant communities are essential for sustainable urban development, yet prior research has often overlooked subjective spatial perceptions and lacked spatiotemporal analysis of vitality drivers in non-core areas. To address this, we introduce the concepts of hidden vibrant communities (HVN) and hidden vibrant areas (HVA), expanding a framework for constructing community vitality (NV) by integrating multimodal data. Combining group perceptions with spatiotemporal dynamics, we identify HVNs in non-core areas and classify them into HVAs based on spatial features. We further introduce a stacking ensemble learning framework to improve both NV prediction accuracy and the methodology. We then explore the impact of community environmental features on HVNs and HVAs, identify vitality drivers and thresholds, and reveal differences and commonalities in vitality across regions. The results find that HVN vitality varies significantly over time and space, while core areas remain stable. Point of interest density (POID) is the primary driver and other factors, such as POI mix (ME) and housing price index (HPI), exhibit region-specific effects. Threshold analysis identifies critical points where these factors positively or negatively impact vitality. District-level classification is used to further analyze the commonalities and differences in vitality drivers across HVAs, informing targeted urban planning strategies to enhance livability.