Ensemble Learning Algorithm Research Articles

Which classifiers are connected to others? An optimal connection framework for multi-layer ensemble systems

Ensemble learning is a powerful machine learning strategy that combines multiple models e.g. classifiers to improve predictions beyond what any single model can achieve. Until recently, traditional ensemble methods typically use only one layer of models which limits the exploration of different aspects in the classifiers’ predictions. On the other hand, the rise of deep learning has introduced multi-layer architectures that can learn complex functions by transforming data into multiple levels of representation. This characteristic of deep learning suggests that multi-layer ensembles may potentially provide better performance compared to single-layer ensembles. However, a problem which might arise is that in the subsequent layers, not all the inputs to a classifier are desirable, leading to lower performance. In this paper, we introduce a novel multi-layer ensemble of classifiers named COME in which each classifier at a specific layer is connected to multiple classifiers in the previous layer. These connections signify the use of the previous-layer-classifiers’ outputs as inputs for training the current layer's classifier. Each classifier can be connected to different classifiers in the previous layer, which allows inputs in each layer to be optimally selected. We propose a binary encoding scheme to encode the topology of the proposed multi-layer ensemble with defined connections between layers. Differential Evolution, a popular evolutionary computation method, is used as the optimisation algorithm to search for the optimal set of connections. Experimental results on 30 datasets from the UCI Machine Learning Repository and OpenML demonstrate that our proposed ensemble outperforms many state-of-the-art ensemble learning algorithms.

Spatial impulse response analysis and ensemble learning for efficient precision level sensing

In this paper, we propose an innovative method for determining the fill level of containers, such as trash cans, addressing a critical aspect of waste management. The method combines spatial impulse response analysis with machine learning (ML) techniques, offering a unique and effective approach for sound-based classification that can be extended to various domains beyond waste management. By employing a buzzer-generated sine sweep signal, we create a distinctive signature specific to the fill level of the waste container. This signature, once accurately decoded, is then interpreted by a specially developed ensemble learning algorithm. Our approach achieves a classification accuracy of over 90% when implemented locally on a development board, optimizing operational efficiencies and eliminating the need to delegate complex classification tasks to external entities. Using low-cost and energy-efficient hardware components, our method offers a cost-effective approach that contributes to sustainable and efficient waste management practices, providing a reliable and locally deployable solution.

Ensemble machine learning algorithm for cost-effective and timely detection of diabetes in Maiduguri, Borno State

Diabetes is a serious medical condition that severely hinders the body's ability to produce or properly regulate insulin, leading to detrimental carbohydrate metabolism and dangerously high blood sugar levels. This ultimately causes inadequate carbohydrate metabolism and heightened blood glucose levels. Alarmingly, from 2000 to 2019, diabetes-related mortality rates rose by 3%. In the year 2019 alone, diabetes was tragically responsible for nearly 2 million deaths. This groundbreaking research introduces the improved weighted average ensemble learning (WAEL) model as an innovative solution for detecting diabetes. The enhanced WAEL model effectively addresses the overfitting challenge by integrating multiple models that have gained unique insights from the data. The proposed WAEL model ingeniously combines five feature spaces through the grey wolf optimisation (GWO) algorithm to uncover the optimal weight combination. GWO plays a vital role in weight optimization, enabling the reduction of weights in models that are particularly sensitive to noise. The results demonstrated that the improved WAEL achieved an astounding level of accuracy, soaring to 98.90%. The LGBM algorithm followed closely, achieving an impressive accuracy of 85.00%. The RF method recorded an accuracy of 81.00%. When it comes to accurately identifying diabetes, the improved WAEL ensemble model significantly outperformed the other five individual models, as evidenced by metrics such as accuracy, precision, recall, and F1-score. Therefore, the proposed model stands as a compelling alternative tool for healthcare professionals in the early detection of diabetes.

Enhanced intrusion detection model based on principal component analysis and variable ensemble machine learning algorithm

The intrusion detection system (IDS) model, which can identify the presence of intruders in the network and take some predefined action for safe data transit across the network, is advantageous in achieving security in both simple and advanced network systems. Several IDS models have various security problems, such as low detection accuracy and high false alarms, which can be caused by the network traffic dataset's excessive dimensionality and class imbalance in the creation of IDS models. Principal Component Analysis (PCA) has proven to be a helpful feature selection technique for dimensionality reduction. As a result, because it is a linear transformation, it has challenges capturing non-linear relationships between feature properties in the network traffic datasets. This paper proposes a variable ensemble machine learning method to solve the problem and achieve a low variance model with high accuracy and low false alarm. First, PCA is combined with the AdaBoost ensemble machine learning algorithm, which acts as stagewise additive modelling to compensate for PCA's deficiency in feature selection in network traffic by minimizing the exponential loss function. Secondly, PCA is used for feature selection, and a LogitBoost classifier algorithm can be used for multiclass classification and acts as an additive tree regression to compensate for the PCA's weakness by minimizing the Logistic Loss to provide an optimal classifier output. Finally, the low variance ability of RandomForest, which employs the bagging approach, is applied to eliminate overfittings. The experiments of the IDS model developed from the proposed methods were evaluated on the WSN-DS, NSL-KDD, and UNSW-N15 datasets. The performance of the methods, PCA with AdaBoost, on the WSN-DS dataset has an accuracy score of 92.3 %, an 89.0 % accuracy score on the NSL-KDD dataset, and a 67.9 % accuracy score on UNSW-N15, which is the least accurate score. PCA and RandomForest surpassed them by scoring 100 % accuracy on all three datasets. PCA and Bagging have an accuracy score of 99.8 % on the WSN-DS dataset, 100 % on the NSL-KDD dataset, and 93.4 % on the UNSW-N15 dataset. In comparison, PCA and LogitBoost have an accuracy score of 98.9 % on the WSN-DS dataset, 100 % on the NSL-KDD dataset, and 88.7 % on the UNSW-N15 dataset.

Utilizing molecular simulation, ideal adsorbed solution theory and ensemble learning algorithms to investigate adsorption and separation of sulfides on amorphous nanoporous materials

Using grand canonical Monte Carlo method, we investigated the adsorption of pure H2S and SO2 gases on amorphous materials, and the separation of CH4-H2S and CO2-SO2 mixtures. At 303 K, the optimal adsorbent for both gases was found to be HCP-Colina-id016, with 16 mmol/g. For CH4-H2S mixture, despite aCarbon-Marks-id002 exhibiting the highest selectivity (approximately 80), the H2S adsorption was low (around 1 mmol/g), while Kerogen-Coasne-id013 demonstrated a high H2S adsorption of 12 mmol/g with a selectivity of 20. In the case of CO2-SO2, HCP-Colina-id018 exhibited a SO2 selectivity exceeding 30, with a high SO2 adsorption of 12 mmol/g. The Ideal Adsorbed Solution Theory underestimated the adsorption and selectivity of both mixtures, particularly evident in CO2-SO2. Molecular simulations revealed that, for the CO2-SO2 system, CO2 underwent condensation, resulting in a sudden drop in the SO2 adsorption isotherm. However, IAST accurately predicted this abrupt change. Based on the adsorption data obtained from molecular simulations, we compared the predictive performance of four ensemble learning algorithms, namely Random Forest (RF), Gradient Boosted Decision Trees (GBDT), Extreme Gradient Boosting (XGBoost), and CatBoost, for H2S and SO2 pure gases in amorphous porous materials. The rankings were observed to be XGBoost > GBDT > RF > CatBoost.

Monitoring of plant diseases caused by Fusarium commune and Rhizoctonia solani in bok choy using hyperspectral remote sensing and machine learning.

Local vegetable production is susceptible to various fungal pathogens, the most common and lethal of which are Fusarium commune and Rhizoctonia solani. Early detection of these pathogens is challenging, and by the time visual symptoms appear, the pathogens may have already spread extensively, causing massive damage to the production. In this study, we explored the use of hyperspectral data for early detection of diseases caused by F. commune or R. solani in bok choy Brassica rapa subsp. chinensis by collecting hyperspectral data from healthy plants and plants inoculated with either fungal pathogen in a controlled experimental set-up. Based on the collected data, we employed various tree-based, distribution-based, geometric, neural networks and ensemble learning algorithms to train detection models. Among the trained models, Multi-Layer Perceptron (MLP) models performed the best with overall accuracy reaching 95.9 ± 0.26%. MLP models could differentiate between healthy and infected plants with 99% precision after 1 day of infection, and distinguish between different fungal pathogens with 99% precision after 2 days. During this period, no visible symptoms of fungal infection could be observed. Further analysis into trained MLP models and general reflectance profiles of plants also revealed a high correlation of the spectral regions 445-460, 560-595, 606-620 and 719-728 nm with fungal infection in bok choy plants. Our findings highlight the potential of hyperspectral imaging as a highly precise early detection tool for fungal diseases in plants. © 2024 Society of Chemical Industry.

A stacking-based model for the spread of Botryosphaeria laricina

Botryosphaeria laricina (larch shoot blight) was first identified in 1973 in Jilin Province, China. The disease spread rapidly and caused considerable damage because its pathogenesis was unknown at the time and there were no effective controls or quarantine methods. At present, it shows a spreading trend, but most research can only conduct physiological analyses within a relatively short period, combining individual influencing factors. Nevertheless, methods such as neural network models, ensemble learning algorithms, and Markov models are used in pest and disease prediction and forecasting. However, there may be fitting issues or inherent limitations associated with these methods. This study obtained B. laricina data at the county level from 2003 to 2021. The dataset was augmented using the SMOTE algorithm, and then algorithms such as XGBoost were used to select the significant features from a combined set of 12 features. A new stacking fusion model has been proposed to predict the status of B. laricina. The model is based on random forest, gradient boosted decision tree, CatBoost and logistic regression algorithms. The accuracy, recall, specificity, precision, F1 value and AUC of the model reached 90.9%, 91.6%, 90.4%, 88.8%, 90.2% and 96.2%. The results provide evidence of the strong performance and stability of the model. B. laricina is mainly found in the northeast and this study indicates that it is spreading northwest. Reasonable means should be used promptly to prevent further damage and spread.

IG-ANGO: a novel ensemble learning algorithm for breast cancer prediction using genomic data

IG-ANGO: a novel ensemble learning algorithm for breast cancer prediction using genomic data

Confirmatory Study on Prediction of Maximal Oxygen Uptake by Ensemble Learning Algorithm

Objective: The prediction model of maximal oxygen uptake (VO2max) was established by different ensemble learning methods. Compared with multiple linear regression, a common prediction method of VO2max, the effectiveness of the ensemble learning model is verified, and screen out the optimal VO2max prediction model. Methods: A total of 84 male college students, were recruited, informed of the purpose and procedure of the experiment, and passed the PAR-Q questionnaire screening. Cardiopulmonary Exercise Test (CPET), lung function, body composition test, VO2max, minute ventilation volume (VE), tidal volume (Vt), weight, and other related indicators. After the correlation analysis of the experimental results, a VO2max multiple linear regression model and an ensemble learning prediction model are computed. Results: The indicators that finally enter the multiple linear regression equation include: weight, Vt, VE. The regression equation is: VO2max (ml/kg/min) = 52.455+0.005×VT (ml/min) + 0.111×VE (L/min)-0.441×weight (kg), and the regression coefficient test results is extremely significant (P<0.01). The VO2max prediction model is computed by using the ensemble learning algorithm and among the five prediction models, it is found that Weighted average (R^2=0.7715) has the best effect, and higher than multiple regression model, there is no significant difference between the actual value and the predicted value VO2max (P>0.05). Conclusion: In the case of a small sample size, the weighted average model predicts VO2max the most accurately, compared with multiple regression models, the ensemble learning model can include more basic indicators, the evaluation of VO2max in the body is more comprehensive, and the prediction results are more accurate.

Enhancing Cybersecurity in Healthcare: Evaluating Ensemble Learning Models for Intrusion Detection in the Internet of Medical Things.

This study investigates the efficacy of machine learning models for intrusion detection in the Internet of Medical Things, aiming to enhance cybersecurity defenses and protect sensitive healthcare data. The analysis focuses on evaluating the performance of ensemble learning algorithms, specifically Stacking, Bagging, and Boosting, using Random Forest and Support Vector Machines as base models on the WUSTL-EHMS-2020 dataset. Through a comprehensive examination of performance metrics such as accuracy, precision, recall, and F1-score, Stacking demonstrates exceptional accuracy and reliability in detecting and classifying cyber attack incidents with an accuracy rate of 98.88%. Bagging is ranked second, with an accuracy rate of 97.83%, while Boosting yielded the lowest accuracy rate of 88.68%.

Detecting refactoring type of software commit messages based on ensemble machine learning algorithms

Refactoring is a well-established topic in contemporary software engineering, focusing on enhancing software's structural design without altering its external behavior. Commit messages play a vital role in tracking changes to the codebase. However, determining the exact refactoring required in the code can be challenging due to various refactoring types. Prior studies have attempted to classify refactoring documentation by type, achieving acceptable results in accuracy, precision, recall, F1-Score, and other performance metrics. Nevertheless, there is room for improvement. To address this, we propose a novel approach using four ensemble Machine Learning algorithms to detect refactoring types. Our experimentation utilized a dataset containing 573 commits, with text cleaning and preprocessing applied to address data imbalances. Various techniques, including hyperparameter optimization, feature engineering with TF-IDF and bag-of-words, and binary transformation using one-vs-one and one-vs-rest classifiers, were employed to enhance accuracy. Results indicate that the experiment involving feature engineering using the TF-IDF technique outperformed other methods. Notably, the XGBoost algorithm with the same technique achieved superior performance across all metrics, attaining 100% accuracy. Moreover, our results surpass the current state-of-the-art performance using the same dataset. Our proposed approach bears significant implications for software engineering, particularly in enhancing the internal quality of software.

Developing an integrated prediction model for daylighting, thermal comfort, and energy consumption in residential buildings based on the stacking ensemble learning algorithm

Developing an integrated prediction model for daylighting, thermal comfort, and energy consumption in residential buildings based on the stacking ensemble learning algorithm

Detecting schizophrenia, bipolar disorder, psychosis vulnerability and major depressive disorder from 5 minutes of online-collected speech.

Psychosis poses substantial social and healthcare burdens. The analysis of speech is a promising approach for the diagnosis and monitoring of psychosis, capturing symptoms like thought disorder and flattened affect. Recent advancements in Natural Language Processing (NLP) methodologies enable the automated extraction of informative speech features, which has been leveraged for early psychosis detection and assessment of symptomology. However, critical gaps persist, including the absence of standardized sample collection protocols, small sample sizes, and a lack of multi-illness classification, limiting clinical applicability. Our study aimed to (1) identify an optimal assessment approach for the online and remote collection of speech, in the context of assessing the psychosis spectrum and evaluate whether a fully automated, speech-based machine learning (ML) pipeline can discriminate among different conditions on the schizophrenia-bipolar spectrum (SSD-BD-SPE), help-seeking comparison subjects (MDD), and healthy controls (HC) at varying layers of analysis and diagnostic complexity. We adopted online data collection methods to collect 20 minutes of speech and demographic information from individuals. Participants were categorized as "healthy" help-seekers (HC), having a schizophrenia-spectrum disorder (SSD), bipolar disorder (BD), major depressive disorder (MDD), or being on the psychosis spectrum with sub-clinical psychotic experiences (SPE). SPE status was determined based on self-reported clinical diagnosis and responses to the PHQ-8 and PQ-16 screening questionnaires, while other diagnoses were determined based on self-report from participants. Linguistic and paralinguistic features were extracted and ensemble learning algorithms (e.g., XGBoost) were used to train models. A 70%-30% train-test split and 30-fold cross-validation was used to validate the model performance. The final analysis sample included 1140 individuals and 22,650 minutes of speech. Using 5-minutes of speech, our model could discriminate between HC and those with a serious mental illness (SSD or BD) with 86% accuracy (AUC = 0.91, Recall = 0.7, Precision = 0.98). Furthermore, our model could discern among HC, SPE, BD and SSD groups with 86% accuracy (F1 macro = 0.855, Recall Macro = 0.86, Precision Macro = 0.86). Finally, in a 5-class discrimination task including individuals with MDD, our model had 76% accuracy (F1 macro = 0.757, Recall Macro = 0.758, Precision Macro = 0.766). Our ML pipeline demonstrated disorder-specific learning, achieving excellent or good accuracy across several classification tasks. We demonstrated that the screening of mental disorders is possible via a fully automated, remote speech assessment pipeline. We tested our model on relatively high number conditions (5 classes) in the literature and in a stratified sample of psychosis spectrum, including HC, SPE, SSD and BD (4 classes). We tested our model on a large sample (N = 1150) and demonstrated best-in-class accuracy with remotely collected speech data in the psychosis spectrum, however, further clinical validation is needed to test the reliability of model performance.

Introducing HeliEns: A Novel Hybrid Ensemble Learning Algorithm for Early Diagnosis of Helicobacter pylori Infection

The Gram-negative bacterium Helicobacter pylori (H. infection) infects the human stomach and is a major cause of gastritis, peptic ulcers, and gastric cancer. With over 50% of the global population affected, early and accurate diagnosis of H. infection infection is crucial for effective treatment and prevention of severe complications. Traditional diagnostic methods, such as endoscopy with biopsy, serology, urea breath tests, and stool antigen tests, are often invasive, costly, and can lack precision. Recent advancements in machine learning (ML) and quantum machine learning (QML) offer promising non-invasive alternatives capable of analyzing complex datasets to identify patterns not easily discernible by human analysis. This research aims to develop and evaluate HeliEns, a novel quantum hybrid ensemble learning algorithm designed for the early and accurate diagnosis of H. infection infection. HeliEns combines the strengths of multiple quantum machine learning models, specifically Quantum K-Nearest Neighbors (QKNN), Quantum Naive Bayes (QNB), and Quantum Logistic Regression (QLR), to enhance diagnostic accuracy and reliability. The development of HeliEns involved rigorous data preprocessing steps, including data cleaning, encoding of categorical variables, and feature scaling, to ensure the dataset’s suitability for quantum machine learning algorithms. Individual models (QKNN, QNB, and QLR) were trained and evaluated using metrics such as accuracy, precision, recall, and F1-score. The ensemble model was then constructed by integrating these quantum models using a hybrid approach that leverages their diverse strengths. The HeliEns model demonstrated superior performance compared to individual models, achieving an accuracy of 94%, precision of 97%, recall of 92%, and an F1-score of 94% in detecting H. infection infection. The quantum ensemble approach effectively mitigated the limitations of individual models, providing a robust and reliable diagnostic tool. HeliEns significantly improved diagnostic accuracy and reliability for early H. infection detection. The integration of multiple quantum ML algorithms within the HeliEns framework enhanced overall model performance. The non-invasive nature of the HeliEns model offers a cost-effective and user-friendly alternative to traditional diagnostic methods. This research underscores the transformative potential of quantum machine learning in healthcare, particularly in enhancing diagnostic efficiency and patient outcomes. HeliEns represents a significant advancement in the early diagnosis of H. infection infection, leveraging quantum machine learning to provide a non-invasive, accurate, and reliable diagnostic tool. This research highlights the importance of QML-driven solutions in healthcare and sets the stage for future research to further refine and validate the HeliEns model in real-world clinical settings.

Predicting non-suicidal self-injury among Chinese adolescents: The application of ten algorithms of machine learning

Background and aimsHigh non-suicidal self-injury (NSSI) prevalence among adolescents is a global health issue. However, current prediction models for adolescent NSSI rely on a limited set of algorithms, resulting in biased predictions. Therefore, the aim of this study is to develop multiple machine learning models to enhance prediction accuracy and mitigate biases among Chinese adolescents. MethodsA total of 4487 junior and senior high school students in China were recruited. Multiple algorithms were included, such as logistic regression, decision tree, support vector machine, Naive Bayes, multi-layer perceptron, K-nearest neighbors, and ensemble learning algorithm like random forest, bagging, AdaBoost, and stacking to build predictive models. Data processing techniques, including standardization and the synthetic minority oversampling technique, were employed to optimize the predictive model. The model was trained on 70 % of the data, reserving 30 % for testing. ResultsThe ten prediction models achieved a good performance, with area under the receiver operating characteristic curve (AUC) scores above 0.700 in the test set. The stacking and random forest models achieved AUC scores of 0.904 and 0.898, respectively. The prediction performance of the Naive Bayes model was relatively poor. The top five important variables were resilience, bully, suicidal ideation, internet addiction, and depression. ConclusionsThe ensemble machine learning algorithm showed promising results predicting NSSI among adolescents. Such algorithms should be recommended for future NSSI research to enhance predictive accuracy. Identification of important features in NSSI prediction can help develop screening protocols and lay a foundation for clinical diagnosis and intervention in adolescent populations.

Data analytics in ensemble learning for effective crop yield prediction

In agricultural research, Crop Yield Prediction (CYP) offers the best decision-making to assist farmers in agricultural yield forecasting efficiently. Most of the existing studies have not considered the exhaustive exploration of data analytics techniques in Machine Learning (ML) for CYP due to which the existing models have not given the optimal results. The main objective of this study is to investigate the effectiveness of data analytics in Ensemble Learning (EL) techniques for more reliable and high-performance CYP models. This article proposes an expert system model, Blended Expert System for Crop Yield Prediction (BESCYP), designed to predict the precise crop yields for specific agricultural lands in the Assam state of India. The proposed BESCYP employs Expectation–Maximization (EM) algorithm to treat missing values, the Isolation Forest (IF) technique to analyze outliers, the Genetic Algorithm (GA) to perform feature selection, Robust Scaling (RS) technique to perform data normalization and the Extra Tree (ET) for classification that overcome the variance and overfitting problem commonly associated with standard ML algorithms. The evaluation of the proposed BESCYP model has been performed using accuracy, precision, recall, and F-1 score on a dataset obtained from International Crops Research Institute for Semi-Arid Tropics (ICRISAT). The proposed model is compared against different standard ML algorithms, EL algorithms and various existing models available in the literature, and the experimental results show that the proposed BESCYP model outperforms other models.

High-noise solar panel defect identification method based on the improved EfficientNet-V2

As a crucial element in photovoltaic power generation systems, the condition of solar panels significantly impacts the efficiency of power generation. The ability to accurately and promptly detect defects in solar panels is essential for enhancing system performance. This study introduces a novel model for identifying defects in photovoltaic modules, leveraging an enhanced version of EfficientNet-V2. This model aims to address challenges in identifying defects in infrared images of solar panels under conditions of high-noise and low-model efficiency. To address the challenges of high image noise and blur, this article initially presents a methodology that combines the Db4 wavelet transform with a blind deconvolution algorithm for comprehensive preprocessing of the original image. Furthermore, this study optimizes the model's feature representation capabilities by implementing key transformations within the EfficientNet-V2 network framework. Notably, we replaced the traditional SE block with the more efficient channel attention (ECA) mechanism module. Due to its lightweight structure and effective performance, ECA substantially improves the model's capacity to extract complex and abstract image features, while also accelerating the training process's convergence speed and enhancing overall computational efficiency. At the classifier level, this paper innovatively integrates the XGBoost ensemble learning algorithm into the model, substituting the conventional softmax classifier used in traditional convolutional neural network (CNN). With its superior generalization capabilities, robust nonlinear modeling skills, and efficient computational characteristics, XGBoost can more accurately detect minute defects in solar panels based on the deep features produced by EfficientNet-V2, thereby significantly improving the accuracy and robustness of defect detection. Simulation results demonstrate that the proposed model structure outperforms traditional CNN models in terms of accuracy and stability, underscoring the efficacy of the enhanced EfficientNet-V2 model in detecting solar panel defects under high-noise conditions.

AdaBoost algorithm for the recognition of young women’s body shapes based on 2D images

Classifying and recognizing the human body shape during human body measurements based on 2D images helps to improve measurement accuracy. In this paper, 430 young women's 2D body images were selected to establish 2D body datasets. The characteristic indices used to represent the body shape in 2D images were extracted by computer vision technology, namely the body height pixel value, projected unit area, and projected area ratio of the front and side of the body. The two-step cluster model was used to classify the body shape into three clusters: the tall, flat, and medium fatness type; the short, thin, and medium roundness type; and the round, fat, and medium height type. Then, the decision tree model and AdaBoost algorithm, an ensemble learning algorithm with the decision tree as the weak classifier, were used to recognize the body shape. The results show that the recognition accuracy of the decision tree recognition model was 93.19%. The body shape recognition method using AdaBoost achieved a better recognition effect than the decision tree model, and the recognition accuracy of the test set reached 97.35%. Through comparative study, we found that the recognition accuracy of the 2D body shape recognition method based on AdaBoost was improved and that the recognition accuracy was relatively stable. This study provides a new method for the recognition of human body shape in clothing customization and online shopping.

Transfer of near-infrared analysis models for gasoline RON based on ensemble learning

The research octane number (RON) has guiding significance for evaluating the quality of gasoline, while near-infrared (NIR) spectroscopy analysis technology provides an important means for the detection of RON non-destructively and rapidly. When using a near-infrared spectrometer to obtain the RON of gasoline, if the analysis model can be shared among different instruments, it will greatly reduce the cost of re-modeling or model maintenance. Aiming to achieve the sharing of a NIR spectroscopy analysis model for RON between two portable near-infrared spectrometers of the same model, two ensemble learning algorithms, random forest (RF) and extreme gradient boosting (XGBoost), were employed for investigation, as well as two other machine learning algorithms, support vector regression (SVR) and decision tree (DT). Based on the RON of 120 gasoline samples and their NIR spectroscopy collected on the two spectrometers, hybrid and pure models were established to evaluate their sharing performance among SVR, DT, RF and XGBoost. In order to further simplify the model and improve its robustness and prediction accuracy, the characteristic wavelength selection strategies, including elimination of uninformative variables (UVE), successive projections algorithm (SPA), and competitive adaptive reweighted sampling (CARS), were also adopted to optimize the model. The results showed that the hybrid model based on the CARS-RF method yielded the best prediction performance, with the coefficient of determination (R2) of 0.96, 0.86, and 0.94 for the single prediction sets of instrument A, instrument B, and the hybrid prediction set of the two instruments, respectively. Therefore, the hybrid modeling method based on ensemble learning algorithms combined with an appropriate wavelength selection strategy can effectively improve the robustness and universality of the model, and achieve the sharing of gasoline RON models on two near-infrared spectrometers of the same model.

Brain age prediction via cross-stratified ensemble learning

As an important biomarker of neural aging, the brain age reflects the integrity and health of the human brain. Accurate prediction of brain age could help to understand the underlying mechanism of neural aging. In this study, a cross-stratified ensemble learning algorithm with staking strategy was proposed to obtain brain age and the derived predicted age difference (PAD) using T1-weighted magnetic resonance imaging (MRI) data. The approach was characterized as by implementing two modules: one was three base learners of 3D-DenseNet, 3D-ResNeXt, 3D-Inception-v4; another was 14 secondary learners of liner regressions. To evaluate performance, our method was compared with single base learners, regular ensemble learning algorithms, and state-of-the-art (SOTA) methods. The results demonstrated that our proposed model outperformed others models, with three metrics of mean absolute error (MAE), root mean-squared error (RMSE), and coefficient of determination (R2) of 2.9405 years, 3.9458 years, and 0.9597, respectively. Furthermore, there existed significant differences in PAD among the three groups of normal control (NC), mild cognitive impairment (MCI) and Alzheimer's disease (AD), with an increased trend across NC, MCI, and AD. It was concluded that the proposed algorithm could be effectively used in computing brain aging and PAD, and offering potential for early diagnosis and assessment of normal brain aging and AD.