Ensemble Model Research Articles

A high-precision automatic diagnosis method of maize developmental stage based on ensemble deep learning with IoT devices

Accurately determining the stage of crop development holds significant importance for field crop management. With the advancement of smart agriculture, an increasing number of Internet of Things (IoT) devices are being integrated into agricultural production, enabling more efficient acquisition of high-precision crop images. Currently, research on detecting crop growth stages based on IoT device images remains relatively scarce. Most existing studies rely on a single network model for detection, often encountering issues such as low accuracy and overfitting. Therefore, in this study, we collected maize images using IoT devices and constructed an integrated deep learning model by utilizing four convolutional neural networks (CNNs) to detect the growth period of maize in real time. Additionally, we implemented several improvements on these four CNNs and subsequently tested the performance of the ensemble model on the maize dataset. Regarding the ensemble strategy for the ensemble model, we proposed a dynamic weighted voting method, building upon the original voting approach, which can mitigate model training fluctuations and expedite model convergence. Ultimately, we manually simulated various lighting conditions to assess their impact on the ensemble model. Experimental results demonstrate that the ensemble deep model proposed in this paper represents a robust method for detecting maize growth stages, achieving an accuracy rate of 0.976 on the maize dataset, effectively facilitating high-precision detection of maize growth stages in complex backgrounds.

Enhanced detection of tomato leaf diseases using ensemble deep learning: INCVX-NET model

<p align="justify"><span lang="EN-US">Automated leaf disease detection quickly identifies early symptoms, and saves time on large farms. Traditional methods like visual inspection and laboratory detection are prevalent despite being labor-intensive, time-consuming, and susceptible to human error. Recently, deep learning (DL) has emerged as a promising alternative for crop disease recognition. However, these models usually demand extensive training data and face problems in generalization due to the diverse features among different crop diseases. This complexity makes it difficult to achieve optimal recognition performance across all scenarios. To solve this issue, a novel ensemble approach INCVX-Net is proposed to integrate the three DL models, ‘Inception, visual geometry group (VGG)-16, and Xception’ using weighted averaging ensemble for tomato crop leaf disease detection. This approach utilizes the strengths of three DL models to recognize a wide range of disease patterns and captures even slight changes in leaf characteristics. INCVX-Net achieves an impressive 99.5% accuracy in disease detection, outperforming base models such as InceptionV2 (93.4%), VGG-16 Net (92.7%), and Xception (95.2%). This significant leap in accuracy demonstrates the growing power of ensemble DL models in disease detection compared to standalone DL models. The research paves the groundwork for future advancements in disease detection, enhancing precision agriculture through ensemble models.</span></p>

Towards optimizing conservation planning: A performance evaluation of modeling techniques for predicting Mediterranean native species distribution

Understanding the intricate relationships between environmental conditions and key variables influencing species distribution is critical for ecology and conservation, especially in the face of escalating environmental challenges intensified by climate change. Species Distribution Models (SDMs) are valuable tools for predicting current and future habitat suitability, aiding in conservation planning. This study specifically sought to evaluate the predictive performance of different modeling techniques in determining the geographical distribution of three Mediterranean native species with different ecological flexibility in the western Mediterranean coastal land of Egypt namely Thymelaea hirsuta (L.) Endl., Ononis vaginalis Vahl, and Limoniastrum monopetalum (L.) Boiss. An ensemble model incorporating three modeling algorithms -the generalized linear model, boosted regression trees, and random forests- was juxtaposed against the Maxent non-parametric machine-learning modeling technique. The models integrated diverse variables, encompassing climatic, edaphic, habitat, and topographic factors. The results unveiled a high degree of similarity and agreement between the Maxent and ensemble models, both exhibiting outstanding fits and performance metrics. The ensemble model demonstrated remarkable accuracy across various measures evaluating model performance. However, Maxent showcased a comparatively strong performance, particularly in identifying critical areas for the distribution of the studied Mediterranean species. These findings emphasize the importance of employing a suite of modeling techniques to refine conservation planning and management strategies. At the local scale, the outcomes can guide targeted actions such as habitat restoration and species monitoring programs. Regionally, and globally it can inform policy development and cross-border conservation initiatives that address shared environmental challenges and contribute to frameworks for studies of similar ecosystems worldwide. Such a comprehensive approach ensures more effective safeguarding of Mediterranean native species under changing environmental conditions.

Ensemble stacking classifier model for prediction of diabetes

Diabetes, being a chronic condition, possesses the capacity to instigate a global healthcare catastrophe. This condition can be managed and potentially cured with prompt diagnosis and treatment. Integrating machine learning technology with medical science enables precise prognosis of an individual’s susceptibility to diabetes. The proposed work presents the ensemble stacking classifier model. This efficient and effective diabetes prediction model predicts a patient’s diabetes risk by combining the output of multiple machine-learning techniques into a single model. The performance parameters of four distinct machine learning classification algorithms K-nearest neighbors (KNN), random forest (RF), support vector machine (SVM), and decision tree (DT) are compared in this study with those of the proposed stacked classifier model. The suggested model is developed using ensemble methods, where the previously discussed algorithms are integrated to create the base classifier layer of the stack classifier. The meta-classifier is implemented in the form of the logistic regression (LR) algorithm. Upon evaluating the performance of both the developed model and its algorithms, it is proved that the proposed model attains a testing accuracy of 88.5%, surpassing the accuracy of all baseline classification algorithms. As a result, this work determines that the ensemble stacking classifier model exhibits higher prediction accuracy than the base classifier algorithms. This finding underscores the model’s potential as a viable instrument for predicting diabetes in individuals.

Integrating ensemble learning and meta bagging techniques for temperature-specific State of Health prediction in Lithium-ion Batteries

Accurately predicting the State of Health (SOH) of lithium-ion batteries is crucial in the field of battery technology to guarantee operational dependability and durability. This study introduces a novel methodology that integrates ensemble learning and meta-bagging techniques to enhance the precision of SOH prediction, particularly in scenarios characterized by fluctuating temperature conditions. We start our methodology by conducting an extensive data-gathering phase, specifically targeting crucial variables: temperature, voltage, current, and capacity. Subsequently, a thorough data preparation stage ensues, encompassing tasks such as cleansing, standardization, and curating relevant features, with particular attention given to temperature characteristics. Next, we employ a temperature-adaptive ensemble learning model that combines different prediction algorithms, such as Linear Regression (LR) and Long Short-Term Memory (LSTM) networks. These algorithms are trained on temperature-stratified data subsets using a Meta-Bagging Estimator (MBE). This strategy enhances the precision of predictions and the ability to overcome the obstacle of temperature changes impacting battery performance. The efficacy of the ensemble model is thoroughly assessed utilizing diverse performance criteria, showcasing its improved predictive capability compared to conventional approaches. Our findings indicate that this customized strategy is better equipped to manage the intricate dynamics of lithium-ion battery behavior, resulting in more dependable and precise assessments of battery SOH. This work substantially contributes to battery health monitoring and can significantly advance battery management systems.

Enhancing ultrasound-guided brachial plexus nerve localization with ResNet50 and support vector machine

<p>Medical image segmentation and classification plays a vital role in nerve block/region identification, particularly for anesthesiologists relying on instinctual judgments. However, due to patient-specific anatomical variations, these methods sometimes lack precision. This research focuses on addressing the problem, by incorporating novel ensembling method of ResNet-50 and support vector machine (SVM) to achieve segmentation of dataset images and classification of nerve blocks respectively. The said novel ensemble model is trained on a publicly available dataset consisting of more than 16,800 images. The sole purpose of this work is to address the problem of peripheral nerve blocking (PNB) with the usage of ensemble modelling, while achieving the highest possible accuracy. This research will help practitioners in accurately identifying the location of brachial plexus and distinguishing the type of nerve block to be injected – interscalene and supraclavicular. The model, which integrates ResNet50 and SVM classifier, achieved a commendable 99.27% accuracy in identifying and classifying the brachial plexus region.</p>

Probability density function based adaptive ensemble learning with global convergence for wind power prediction

Accurate wind power prediction is highly significant to the safety, stability, and economic operation of power grids. Currently, typical ensemble methods for wind power forecasting are widely designed based on the mean square error (MSE) loss, which are very suitable for the assumption that the error distribution obeys the Gaussian distribution. However, the complex nonlinear conversion of wind energy into wind power may change the statistical characteristics of errors, thus the prediction model based on the traditional MSE loss may lead to the performance degradation of the forecasting model. To address these problems, a probability density function based adaptive ensemble learning with global convergence is proposed for wind power prediction, which comprises three modules: a data preprocessing module, a prediction module, and a combination module. Firstly, an effective feature generation mechanism is employed to extract the multi-mode characteristics of wind data. Then, an auxiliary error based adaptive global convergence model is developed as benchmark predictor in the prediction module, where an adaptive updating algorithm is derived based on the Lyapunov approach to ensure the global convergence of the model weights. Moreover, considering the asymmetric characteristic of modeling error, a probability density function (PDF) based ensemble learning is created to integrate the results of benchmark predictors. Specifically, the ensemble model parameter updating is transformed into the shape control for the modeling error PDF, which can break through the limitation of MSE loss capturing only the second moment information, and emphasize the spatial distribution of errors to make an unbiased estimate of wind power. Experimental results show that the proposed ensemble model has significant advantages over other models involved in this study.

Weighted ensemble deep learning approach for classification of gastrointestinal diseases in colonoscopy images Aided by explainable AI

Gastrointestinal diseases are significant health issues worldwide, requiring early diagnosis due to their serious health implications. Therefore, detecting these diseases using artificial intelligence-based medical decision support systems through colonoscopy images plays a critical role in early diagnosis. In this study, a deep learning-based method is proposed for the classification of gastrointestinal diseases and colon anatomical landmarks using colonoscopy images. For this purpose, five different Convolutional Neural Network (CNN) models, namely Xception, ResNet-101, NASNet-Large, EfficientNet, and NASNet-Mobile, were trained. An ensemble model was created using class-based recall values derived from the validation performances of the top three models (Xception, ResNet-101, NASNet-Large). A user-friendly Graphical User Interface (GUI) was developed, allowing users to perform classification tasks and use Gradient-weighted Class Activation Mapping (Grad-CAM), an explainable AI tool, to visualize the regions from which the model derives information. Grad-CAM visualizations contribute to a better understanding of the model’s decision-making processes and play an important role in the application of explainable AI. In the study, eight labels, including anatomical markers such as z-line, pylorus, and cecum, as well as pathological findings like esophagitis, polyps, and ulcerative colitis, were classified using the KVASIR V2 dataset. The proposed ensemble model achieved a 94.125% accuracy on the KVASIR V2 dataset, demonstrating competitive performance compared to similar studies in the literature. Additionally, the precision and F1 score values ​​of this model are equal to 94.168% and 94.125%, respectively. These results suggest that the proposed method provides an effective solution for the diagnosis of GI diseases and can be beneficial for medical education.

SnapE-ResNet: A novel electronic nose classification algorithm for gas data collected by open sampling systems

Increasing the depth of the neural network and implementing the ensemble of neural networks are two main methods to improve the accuracy of gas recognition algorithms. However, the issues of optimization difficulties or excessive resource consumption may occur when training deep neural networks or integrating multiple neural networks. In this study, an algorithm of snapshot ensemble is combined with Residual network (ResNet) to form a SnapE-ResNet which could simplify the training of the network due to the residual structure of ResNet, while realizes the ensemble of multiple ResNet models without additional time consumption and helps the network escape from local minimum using cyclic cosine annealing algorithm. The effectiveness of the proposed SnapE-ResNet is verified through classification experiments on the public dataset. Furthermore, the long-term drift and short-term drift are reduced by applying a zero-offset baseline compensation algorithm and a gaussian noise that added to the baseline signal. The experimental result indicating a high gas classification accuracy of 99.8% from SnapE-ResNet is obtained, outperforming the comparative models. In addition, the necessity for snapshot ensemble is validated by network ablation experiments. This study could offer a reference for gas classification tasks in the gas detection fields.

AB-SEL model for the prediction of coronary heart disease

Coronary heart disease (CHD) stands as the leading cause of silent and noncommunicable deaths. Early detection is essential for slowing the progression of death and saving lives. Machine learning is of interest to medical researchers and has been used to predict heart disease. This article proposes an Accuracy-Based Stacked Ensemble Learning (AB-SEL) Model to predict coronary heart disease while minimizing computational time. The dataset undergoes the Logistic Regression Recursive Feature Elimination (LR-RFE) method to identify the important features. LR, RF and AdaBoost classifiers are chosen to build Ensemble machine-learning models including techniques like bagging, majority voting, and stacking. Random search and grid search techniques have also been employed to optimize the hyperparameters. The authors applied the Cleveland dataset accessible on Kaggle and data scaling was performed using the standard scalar method. Performance measures were used to assess effectiveness, including accuracy, precision, recall, F1 score, and computational time, and were validated through 5-Fold cross-validation. The suggested approach achieved a 90.16% accuracy rate, requiring only 0.2 seconds of computational time, and yielded an AUC of 0.892.

Analysis of vegetation dynamics from 2001 to 2020 in China's Ganzhou rare earth mining area using time series remote sensing and SHAP-enhanced machine learning

Rare earth mining, essential for modern industries and economic growth, often leads to severe environmental degradation. Previous research has explored the ecological impacts of rare earth mining but has not fully investigated the intricate interplay and subdivisions of environmental and anthropogenic factors driving vegetation changes over extended periods. This study addresses this gap by employing time series remote sensing and SHAP-enhanced machine learning to analyze vegetation dynamics in China's Ganzhou rare earth mining area from 2001 to 2020. Using the kNDVI derived from Landsat data, we identified three distinct vegetation trajectory types: pro-environment, des-environment, and res-environment. An ensemble machine learning model combined with SHAP analysis revealed the cropland area proportion, PM10 levels, and shrubland area proportion as the most influential factors affecting vegetation across all mining types. Additionally, after 2012, the palmer drought severity index and downward surface shortwave radiation emerged as positive contributors to vegetation health, while population pressure had a more substantial negative influence in des-environment areas. Our findings highlight spatial heterogeneity in vegetation recovery patterns and highlight the complex interactions among land cover changes, air quality, climate factors, and human activities in shaping vegetation dynamics. This study provides valuable insights for developing targeted, context-specific restoration strategies in rare earth mining areas, contributing to more sustainable mining practices and global environmental management.

Unveiling the spatially varied nonlinear effects of urban built environment on housing prices using an interpretable ensemble learning model

The relationship between urban built environment (UBE) and housing prices manifests as complex, exhibiting significant nonlinearities and spatial heterogeneity that remain inadequately understood. Taking Shanghai as a testbed, this study employs a novel ensemble learning approach, augmented by Bayesian optimization and Monte Carlo simulation, to decipher the intricate and nonlinear impacts of UBE factors on housing markets across diverse urban geographies. Our analysis unveils substantial spatial variations in how transit accessibility, amenities, residential density, and green/blue spaces influence real estate values. Proximity to metro stations and bike-sharing facilities exerts a more pronounced positive effect than bus stops. Moreover, residents in central areas demonstrate a higher willingness-to-pay for public service amenities, while those in outer suburbs prioritize access to public transportation infrastructure. Intriguingly, it invokes an optimal threshold range of urban density for properties in new cities, thereby increasing the vitality and dense socio-economic networks. Furthermore, the sprawling suburbs have identified an adverse economic impact of large conservation green/blue spaces. These insights can guide policymakers in crafting spatially-tailored strategies that harness localized built environment drivers to catalyse equitable and prosperous urban development. Tailored policies informed by this spatially explicit understanding of nonlinear built environment-housing interactions can foster more sustainable, liveable, and inclusive cities.

A drift-aware dynamic ensemble model with two-stage member selection for carbon price forecasting

A drift-aware dynamic ensemble model with two-stage member selection for carbon price forecasting

Unveiling personality traits through Bangla speech using Morlet wavelet transformation and BiG

Speech serves as a potent medium for expressing a wide array of psychologically significant attributes. While earlier research on deducing personality traits from user-generated speech predominantly focused on other languages, there is a noticeable absence of prior studies and datasets for automatically assessing user personalities from Bangla speech. In this paper, our objective is to bridge the research gap by generating speech samples, each imbued with distinct personality profiles. These personality impressions are subsequently linked to OCEAN (Openness, Conscientiousness, Extroversion, Agreeableness, and Neuroticism) personality traits. To gauge accuracy, human evaluators, unaware of the speaker’s identity, assess these five personality factors. The dataset is predominantly composed of around 90% content sourced from online Bangla newspapers, with the remaining 10% originating from renowned Bangla novels. We perform feature level fusion by combining MFCCs with LPC features to set MELP and MEWLP features. We introduce MoMF feature extraction method by transforming Morlet wavelet and fusing MFCCs feature. We develop two soft voting ensemble models, DistilRo (based on DistilBERT and RoBERTa) and BiG (based on Bi-LSTM and GRU), for personality classification in speech-to-text and speech modalities, respectively. The DistilRo model has gained F-1 score 89% in speech-to-text and the BiG model has gained F-1 score 90% in speech modality.

Comparative exploration of deep convolutional neural networks using real-time endoscopy images

Until now various deep convolutional neural networks are designed and trained for the purpose of classifying different medical conditions related to the domain of gastroenterology. Most of the study carried out have considered publicly available datasets to train the classification networks. Nevertheless, the main motive for carrying out different works in the field gastroenterology is to administer the developed models in healthcare centers in real-world set-ups. For doing so, it is important to check the generalizing ability of the designed systems by regulating them so as to classify endoscopy images captured in a specific hospital. In this regard, the foremost work completed is the collection of the endoscopy data from the hospital and then correctly annotating the images taking the help of a senior endoscopist with experience of more than 5 years. Once the data annotation is completed, the images are segregated into the class of normal and abnormal endoscopy images. Four different models are designed in the current work based on deep learning models, transfer learning models and ensemble approaches, and trained to classify the hospital endoscopy data as normal or abnormal. The models are then tested and evaluated based on various performance measures. It is observed from the comparative analysis that the transfer learning-based ensemble model has the best generalizing ability and gives the best specificity of 100 ​%. It is believed that deep learning-based models can assist endoscopists in add-on to human prediction efficiency.

Multi-model fusion method for predicting CO2 concentration in greenhouse tomatoes

With the rapid development of greenhouse agriculture, accurate prediction of environmental parameters such as temperature, humidity, and carbon dioxide concentration is crucial for optimal crop growth. Traditional forecasting models struggle with the nonlinear and complex nature of greenhouse data, leading to challenges in model robustness. This study addresses these issues by proposing a multi-model fusion strategy for predicting CO2 concentration in greenhouse tomatoes. The proposed method integrates wavelet denoising (WT), variational mode decomposition (VMD), and long short-term memory networks (LSTM). This innovative nonlinear ensemble model effectively extracts key time series features and removes noise, while an introduced attention mechanism enhances the model’s focus on essential time steps, improving prediction accuracy. Experimental results demonstrate that the multi-model fusion approach significantly outperforms single models in terms of accuracy and stability, achieving mean absolute error (MAE) and root mean square error (RMSE) of 0.0117 and 0.0194, respectively. The proposed method offers significant advantages for CO2 prediction in greenhouse crops, providing a theoretical basis and technical support for optimizing and controlling greenhouse parameters. This contributes to the advancement of smart agriculture by offering an efficient environmental monitoring and prediction tool. Additionally, the study presents new ideas and technical solutions for addressing similar agricultural environment prediction challenges, optimizing greenhouse environment control strategies, and improving crop production efficiency.

Pneumonia stage analyzes through image processing

A physical examination and diagnostic imaging techniques including lung biopsies, ultrasounds, and chest X-rays are typically used to make the diagnosis of pneumonia infection, an infectious disease that has the potential to be life-threatening. The objective of this research is to categorize the stages of pneumonia through image processing methods. Before that, an ensemble model for diagnosing pneumonia infections is created utilizing the transfer learning algorithms ResNet50V2 and DenseNet201. The 5,857 images were taken from the PAUL MOONEY dataset for this research. The proposed ensemble averaging model recognizes lung infection appropriately and accurately. By applying a contour detection approach, the left and right chests are separated and the affected pixels from there to analyze the stage of pneumonia. It is very crucial to identify the stage for treatment purposes.

Cross-condition fault diagnosis of chillers based on an ensemble approach with adaptive weight allocation

The Heating, Ventilation and Air Conditioning (HVAC) systems are complex and prone to failures during operation, often leading to significant energy waste. Timely and accurate Fault Detection and Diagnosis (FDD) can enhance energy efficiency. The HVAC system operates under diverse conditions, data-driven models trained under existing conditions may experience performance degradation when faced with new conditions. Transfer learning offers an effective solution to this issue. This study proposes a novel transfer learning ensemble model based on adaptive weights, leveraging different transfer learning strategies to improve diagnosis performance under new conditions. Multiple cross-condition transfer learning tasks were implemented to test the proposed method, and its effectiveness was validated through multiple experiments to minimize the impact of randomness. Results showed that, compared to fine-tuning (FT), domain-adversarial neural network (DANN), and baseline models, the proposed method outperforms the other models. The average accuracy of multiple experiments improved by 0.21 % to 2.34 % compared to FT. Additionally, modifying DANN to utilize a small amount of labeled information from the target domain has led to greater overlap between the feature distributions of the source and target domains, resulting in improved performance that is close to that of FT. Finally, we analyzed the impact of target domain data volume on the performance of the four methods. The performance of the baseline model improved significantly with the increase in data volume, while the other models showed less improvement. Meanwhile, the diagnostic results of the baseline model were significantly influenced by experimental randomness when there is less training data, whereas the FT diagnostic results were relatively more stable.

Enhancing Cardiovascular Disease Prediction Accuracy through an Ensemble Machine Learning Approach

This study explores the efficacy of an ensemble machine learning approach, specifically a Voting Classifier combining Decision Tree, k-Nearest Neighbors, and Gaussian Naive Bayes classifiers, in predicting cardiovascular diseases (CVDs). Utilizing a dataset consisting of 70,000 clinical records, the model was rigorously tested through 5-fold cross-validation, achieving remarkable results with average accuracies, precision, recall, and F1-scores all exceeding 99%. The findings validate the hypothesis that ensemble models, due to their capacity to leverage multiple learning algorithms, provide superior prediction accuracy and reliability compared to single predictor models. This research not only confirms the effectiveness of ensemble methods in medical diagnostics but also highlights their potential to enhance decision-making in clinical settings. Given the model's success in identifying various stages of cardiovascular conditions with high accuracy, it offers significant implications for early intervention and personalized patient management. Future research should aim to validate these results across more diverse populations and explore the integration of additional predictive factors that could refine the model's applicability. This study contributes to the computational health field by demonstrating how advanced machine learning techniques can be effectively applied in predicting health outcomes.

SDN-Driven Security Framework for DDOS Attack Detection and Mitigation

The rising complexity and persistence of Distributed Denial of Service (DDoS) attacks, particularly low-rate variants, present significant challenges in detection and mitigation within Software-Defined Networking (SDN) environments. Existing detection systems often flood networks with alerts, burdening security personnel and delaying timely mitigation. Furthermore, many solutions are designed and tested in simulated conditions, limiting their real-world applicability. To address these challenges, we propose an SDN-based security framework enhanced with automated monitoring, detection, and mitigation capabilities, optimized for slow-rate DDoS attacks. Our framework was rigorously evaluated on a physical testbed, achieving mitigation efficiencies between 91.66% and 100% under varied attack conditions, thus proving its robustness in practical settings. Additionally, we introduce the SDN-SlowRate-DDoS dataset, designed to assist researchers and industry professionals in developing and testing intrusion detection solutions in more realistic scenarios. To further improve DDoS defense in SDN, we incorporate an ensemble online machine learning model that dynamically adapts to evolving attack patterns, enhancing accuracy across attack types and outperforming traditional models with a detection rate of 99.2% on benchmark datasets. This dual approach leverages both real-world testing and adaptive machine learning, advancing proactive DDoS threat management in SDN environments