Ensemble Accuracy Research Articles

Effsemble: faster, smaller and more accurate ensemble networks for thoracic disease classification

Effsemble: faster, smaller and more accurate ensemble networks for thoracic disease classification

A hybrid-extreme learning machine based ensemble method for online dynamic security assessment of power systems

• A hybrid model is developed, blending ELM and Levenberg-Marquardt. • Particle swarm optimization-based weighted averaging ensemble is developed. • Energy margin and time margin are calculated for better application of techniques. • Hybrid ELM is found most suited technique for online DSA application. This manuscript develops a new hybrid-extreme learning machine (ELM) based ensemble model for real-time dynamic security assessment (DSA) of power systems. In order to boost the forecasting accuracy of ELM algorithm, a Levenberg-Marquardt (LM) backpropagation algorithm is used. The Ensemble strategy takes advantage of complementary information to improve the generalization capability of model. To realize an accurate ensemble model, particle swarm optimization (PSO) based weighted averaging is used to combine the individual forecasts. Transient energy function (TEF) terms, generator real and reactive power outputs and fault location are taken as input features for classification as well as prediction of transient energy margin (TEM) and time margin (TM) for a given operating condition and fault location. The potential of the proposed technique is validated on New England 39-bus and 68-bus test power systems. The results demonstrate that the developed model outperforms other state-of-the-art methods.

Effsemble: faster, smaller and more accurate ensemble networks for thoracic disease classification

Effsemble: faster, smaller and more accurate ensemble networks for thoracic disease classification

Quantifying the evolution of ensemble water temperature forecasts as a function of weather forecast lead-time: case study on the Nechako River watershed

Producing and improving hydrological and hydrodynamic forecasts while accounting for uncertainty through a probabilistic approach is useful in various applications, such as for water temperature forecasting. To produce such ensembles, probabilistic meteorological forecasts can be fed into hydrological and water temperature models for different lead-times. This study aims to gauge the impact of the meteorological forecast quality on the accuracy and reliability of water temperature forecast ensembles generated through the HEC-RAS process-based hydrothermal model. The Nechako River, a managed river system in British Columbia, Canada, is used in this case study. The thermal forecasts were generated and evaluated from 2017 to 2020. The tested hypothesis is that improvements in the meteorological forecasts would result in reducing the uncertainty and improving the accuracy of the water temperature forecast ensembles at various lead-times. The results of this study show that the thermal forecasts were indeed improved in terms of their sharpness and their individual accuracy, but no significant impacts were noted over the accuracy of the ensembles. Reliability was also investigated, and it was revealed that water temperature forecast ensembles were initially under-dispersed over the Nechako River, and that this issue was exacerbated when the HEC-RAS model was forced with better quality ensemble weather forecasts. The presence of lakes along the river and the meteorological forecast’s reliability are considered and discussed as causes for this issue. Overall, it was concluded that a reduction of the meteorological inputs’ uncertainty did not improve the uncertainty representation of water temperature forecasts for this system.

On the intensity of light scattered by molecular liquids-Comparison of experiment and quantum chemical calculations.

The intensity of light scattered by liquids has been studied for over a century since the valuable microscopic information about the molecules can be obtained, such as the anisotropy of the molecular polarizability tensor or preferred orientations of neighboring molecules. However, in modern dynamic light scattering experiments, the scattering intensity is usually disregarded, unlike in dielectric spectroscopy, which can be considered as a complementary experimental method, where the dielectric strength is routinely evaluated. The reason lies partly on the fact that the exact form of the equations relating the macroscopically measured light scattering intensity to the microscopic properties of the molecules is debated in the literature. Therefore, as a first step, we compare anisotropy parameters from the literature, calculated from light scattering intensities using different equations, with quantum chemical calculations for over 150 medium-sized molecules. This allows us to identify a consistent form of equations. In a second part, we turn to the depolarized light scattering spectra of 13 van der Waals liquids and some mixtures thereof, recorded with a combination of Tandem-Fabry-Perót and Raman spectroscopies, giving direct access to the reorientational dynamics of the molecules. We discuss how the strength of the structural α-relaxation is connected to the anisotropy parameter, what implication this has for the shape of the α-relaxation, how the components of a mixture-also for the case of ionic liquids-can be identified in this way, and how orientational correlation parameters can be extracted. Additionally, we point out for the example of n-alkanes that for highly flexible molecules, the reorientational motion might not be the decisive source of the depolarized scattered light. We also show that light scattering might serve as a sensitive tool to check the accuracy of a conformer ensemble obtained by quantum chemical calculations.

Fast and Accurate Artificial Compressibility Ensemble Algorithms for Computing Parameterized Stokes–Darcy Flow Ensembles

Fast and Accurate Artificial Compressibility Ensemble Algorithms for Computing Parameterized Stokes–Darcy Flow Ensembles

Research on and application of the operational storm surge ensemble forecast model in the Bay of Bengal

The coastal region of the Bay of Bengal (BoB) is one of the most vulnerable areas to storm surges caused by tropical cyclones (TCs). Historical data indicate that the BoB experiences two seasons in the annual cycle of TC formation, namely, the pre-monsoon season (April to May) and the post-monsoon season (October–December). The frequency of post-monsoon cyclones is significantly higher than that of pre-monsoon cyclones. Very severe cyclonic storms and stronger-intensity storms can cause disastrous storm surges. Based on the accuracy of current forecasts of the path, intensity and radius of maximum wind of TCs in the BoB, a total of 153 ensemble members are automatically generated according to the subjective deterministic forecast results of the TCs, basically covering all possible TCs. A graphics processing unit (GPU)-based parallel storm surge forecast model is established and validated for the coastal areas of the BoB. The speedup ratio of the model can exceed 300, which can meet the computational time requirements of large-scale ensemble numerical forecasts. For the 2020 Amphan cyclone, the costliest cyclone ever recorded in the North Indian Ocean (NIO), the numerical ensemble forecast model can provide rapid and accurate storm surge ensemble and numerical probability forecast products for the BoB. The probability of coastal flooding risk can be evaluated for the coastal areas of the BoB, enabling reasonable and valuable recommendations to be made to evacuate people and protect property near dangerous areas.

Computer-aided diagnostic for classifying chest X-ray images using deep ensemble learning

BackgroundNowadays doctors and radiologists are overwhelmed with a huge amount of work. This led to the effort to design different Computer-Aided Diagnosis systems (CAD system), with the aim of accomplishing a faster and more accurate diagnosis. The current development of deep learning is a big opportunity for the development of new CADs. In this paper, we propose a novel architecture for a convolutional neural network (CNN) ensemble for classifying chest X-ray (CRX) images into four classes: viral Pneumonia, Tuberculosis, COVID-19, and Healthy. Although Computed tomography (CT) is the best way to detect and diagnoses pulmonary issues, CT is more expensive than CRX. Furthermore, CRX is commonly the first step in the diagnosis, so it’s very important to be accurate in the early stages of diagnosis and treatment.ResultsWe applied the transfer learning technique and data augmentation to all CNNs for obtaining better performance. We have designed and evaluated two different CNN-ensembles: Stacking and Voting. This system is ready to be applied in a CAD system to automated diagnosis such a second or previous opinion before the doctors or radiology’s. Our results show a great improvement, 99% accuracy of the Stacking Ensemble and 98% of accuracy for the the Voting Ensemble.ConclusionsTo minimize missclassifications, we included six different base CNN models in our architecture (VGG16, VGG19, InceptionV3, ResNet101V2, DenseNet121 and CheXnet) and it could be extended to any number as well as we expect extend the number of diseases to detected. The proposed method has been validated using a large dataset created by mixing several public datasets with different image sizes and quality. As we demonstrate in the evaluation carried out, we reach better results and generalization compared with previous works. In addition, we make a first approach to explainable deep learning with the objective of providing professionals more information that may be valuable when evaluating CRXs.

A Hybrid Approach for Mass Valuation of Residential Properties through Geographic Information Systems and Machine Learning Integration

Geographic Information Systems (GIS) and Machine Learning methods are now widely used in mass property valuation using the physical attributes of properties. However, locational criteria, such as as proximity to important places, sea or forest views, flat topography are just some of the spatial factors that affect property values and, to date, these have been insufficiently used as part of the valuation process. In this study, a hybrid approach is developed by integrating GIS and Machine Learning for mass valuation of residential properties. GIS‐based Nominal Valuation Method was applied to carry out proximity, terrain, and visibility analyses using Ordnance Survey and OpenStreetMap data, than land value map of Great Britain was produced. Spatial criteria scores obtained from the GIS analyses were included in the price prediction process in which global and spatially clustered local regression models are built for England and Wales using Price Paid Data and Energy Performance Certificates data. Results showed that adding locational factors to the property price data and applying a novel nominally weighted spatial clustering algorithm for creating a local regression increased the prediction accuracy by about 45%. It also demonstrated that Random Forest was the most accurate ensemble model.

Enhanced machine learning tree classifiers for lithology identification using Bayesian optimization

Lithology identification is a fundamental activity in oil and gas exploration. The application of artificial intelligence (AI) is currently being adopted as a state-of-the-art means of automating lithology identification. One aspect of this AI approach is the application of population search algorithms to optimise hyperparameters for enhanced prediction performance. For the first time, Bayesian optimization is deployed to determine the optimal learning parameters for more accurate tree and tree ensemble lithology classifiers. The aim is to rely on the ability of Bayesian optimization to consider previous classification results to improve the output of decision and ensemble tree lithology models using well logs as inputs. The proposed Bayesian optimised decision tree (BODT) generated the best classification accuracy of 89.8% as compared to 86.9%, 83.3% and 81.2% for fine, medium and coarse trees. For the ensembled trees, the Bayesian optimised AdaBoost (BO-AdaBoost) classifier generated the highest improved prediction accuracy of 94.2% while Bayesian optimised Bagged (BO-Bagged) and Bayesian optimised RUSBoost (BO-RUSBoost) had a lower accuracy rate of 94.0% and 77.1% respectively. Additionally, the performance of the Bayesian optimised classifiers offered higher reliability when compared with particle swarm optimization-based artificial neural networks (PSO-ANN). Hence, incorporating Bayesian optimization as a hyperparameter search algorithm will improve litholofacies recognition, leading to a higher accuracy rate and thereby provide an improved alternative for intelligent lithology identification.

Prediction of Household Food Security Status using Ensemble Learning Models

Background:This research uses the Ethiopian HICE survey dataset. Predicting food insecurity is critical in presenting the household's situation to the appropriate agencies that take preventative and intervention measures.Objective:This research paper's primary goal is to predict households' food security status using ensemble learning models.Methods:We use five base classifiers and a voting strategy for ensemble classification to enhance the performance of different base classifiers. Backward feature elimination and hard and soft voting-based ensemble learning are used to evaluate household food security. The training set for the basic classifiers is composed of the features that have been selected. Each ML classifier makes its prediction about the class label with the help of an ensemble learning method. For making decisions, hard voting uses a simple majority, whereas soft vote employs a weighted probability. To determine the final prediction. Ethiopian household income, consumption, and expenditure dataset are used to test the proposed ensemble learning approach. The backward feature elimination approach improved the model's performance by removing irrelevant and redundant features. Random forest, gradient boosting, multi-layer perceptron, K-nearest Neighbor, and Extra Tree classifiers were used to predict the family's level of food security. Finally, the authors compare the accuracy of ensemble and base classifiers.Result:The experiment result shows that the RF classifier surpasses the other base and ensemble classifiers and scored 99.98% accuracy. Because a Random forest classifier is an ensemble learning classifier that uses several decision trees, the final prediction is computed based on the majority vote of the several trees. The comparison result of hard and soft voting reveals that soft voting outperforms hard voting before and after feature selection with accuracies of 99.79% and 99.77%, respectively.Conclusion:Based on the result obtained, ensemble learning plays a significant role in predicting household food security status and implementing hard and soft voting. The RF classifier surpasses the other base and ensemble classifiers with an accuracy of 99.98%. From ensemble methods, soft voting surpasses hard voting with an accuracy score of 99.79%.

Spatial mode-based calibration (SMoC) of forecast precipitation fields from numerical weather prediction models

Statistical calibration of precipitation forecasts from numerical weather prediction (NWP) models is routinely performed grid-cell by grid-cell, aiming to produce accurate and reliable ensemble forecasts for precipitation fields. Calibrated ensemble members from different grid-cells are then connected using ensemble reordering to form spatially structured ensemble forecasts. However, ensemble reordering approaches, such as the well-known Schaake shuffle method, are often criticized for not considering real physical atmospheric states of precipitation events. In this paper, we propose a spatial mode-based calibration (SMoC) model for post-processing forecast precipitation fields and producing ensemble forecasts with an inbuilt spatial structure, so that ensemble reordering is not required. The SMoC model is developed based on spatial modes derived from empirical orthogonal function (EOF) analyses of precipitation fields and linear regressions of derived EOF expansion coefficients of the first few dominant modes. Unlike conventional calibration models that are applied to forecast grid-cells individually, SMoC is applied to the whole forecast fields, and the spatial structure is inherently present in calibrated ensemble forecasts. There is therefore no need to rely on ensemble reordering for spatial reconstruction. The performance of SMoC is evaluated by applying it to NWP forecasts of substantive precipitation events over the Brisbane drainage basin in eastern Australia. Cross-validated results show that SMoC calibrated forecasts are of high quality at both grid-cell and basin scales. The spatial structure of precipitation is found to be well embedded in the ensemble members of the calibrated forecasts for the basin.

Effective machine learning model combination based on selective ensemble strategy for time series forecasting

The success of ensemble forecasting heavily depends on the selection and combination of component models as proven by numerous studies that show the superior performance of ensemble forecasting in modeling volatile time series. This work develops a selective machine learning ensemble (SMLE) model that evaluates the validation accuracy of the model ensemble by applying a time series cross-validation method, and establish an efficient model combination process by using a novel soft selection algorithm for selecting and weighing several machine learning models, such as support vector regression, feedforward neural network, random forest, and gradient boosting decision tree. The well-known NN3 time series forecasting experiment and a real-world crude oil price forecasting application are used to verify the performance of SMLE. Numerical results and in-depth analysis indicate that the proposed model can provide more accurate and reliable predictions compared with those obtained via individual forecasts, advanced forecasting techniques, and ensemble strategies.

The role of structural dynamics in the thermal adaptation of hyperthermophilic enzymes.

Proteins from hyperthermophilic organisms are evolutionary optimised to adopt functional structures and dynamics under conditions in which their mesophilic homologues are generally inactive or unfolded. Understanding the nature of such adaptation is of crucial interest to clarify the underlying mechanisms of biological activity in proteins. Here we measured NMR residual dipolar couplings of a hyperthermophilic acylphosphatase enzyme at 80°C and used these data to generate an accurate structural ensemble representative of its native state. The resulting energy landscape was compared to that obtained for a human homologue at 37°C, and additional NMR experiments were carried out to probe fast (15N relaxation) and slow (H/D exchange) backbone dynamics, collectively sampling fluctuations of the two proteins ranging from the nanosecond to the millisecond timescale. The results identified key differences in the strategies for protein-protein and protein-ligand interactions of the two enzymes at the respective physiological temperatures. These include the dynamical behaviour of a β-strand involved in the protection against aberrant protein aggregation and concerted motions of loops involved in substrate binding and catalysis. Taken together these results elucidate the structure-dynamics-function relationship associated with the strategies of thermal adaptation of protein molecules.

Developing an ensemble machine learning model for early prediction of sepsis-associated acute kidney injury

SummarySepsis-associated acute kidney injury (S-AKI) is very common and early prediction is beneficial. This study aiming to develop an accurate ensemble model to predict the risk of S-AKI based on easily available clinical information. Patients with sepsis from the United States (US) database Medical Information Mart for Intensive Care-IV were used as a modeling cohort to predict the occurrence of AKI by combining Support Vector Machine, Random Forest, Neural Network, and Extreme Gradient Boost as four first-level learners via stacking algorithm. The external validation databases were the eICU Collaborative Research Database from US and Critical Care Database comprising infection patients at Zigong Fourth People’s Hospital from China, whose AUROC values for the ensemble model 48–12 h before the onset of AKI were 0.774–0.788 and 0.756–0.813, respectively. In this study, an ensemble model for early prediction of S-AKI onset was developed and it demonstrated good performance in multicenter external datasets.

A dynamic ensemble method for residential short-term load forecasting

In this paper, a dynamic ensemble method is proposed to forecast the residential short-term load accurately. The main idea is to utilize the state-space approaches to dynamically adjust the weight coefficients used to combine the base models. The dynamic ensemble method is divided into two stages. In the first stage, the least square method is utilized to estimate the weight coefficients. In the second stage, the particle filter is employed to exclude the estimation error in the first stage and update the weight coefficients dynamically to improve the ensemble accuracy. Then, three heterogeneous models (i.e., vector autoregression model, Gaussian process regression model, and the long short-term memory neural network) are employed as the base models and combined based on the weight coefficients to forecast the residential load. The numerical tests are conducted considering two scenarios in two public datasets to investigate the performance of the proposed method. The first scenario assumes the base models are accurate and stable, while the second scenario considers the dynamics of the base models. Root mean square error, mean absolute error, and mean absolute percentage error are used to evaluate the performances. These indicators show that the proposed method outperforms in both scenarios and datasets.

An Ensemble-Based Analysis of a Liminal Extreme Rainfall Event near Taiwan

This study analyzes an ensemble of numerical simulations of a heavy rainfall event east of Taiwan on 9 June 2020. Heavy rainfall was produced by quasi-stationary back-building mesoscale convective systems (MCS) associated with a mei-yu front. Global model forecast skill was poor in location and intensity of rainfall. The mesoscale ensemble showed liminal conditions between heavy rainfall or little to no rainfall. The two most accurate and two least accurate ensemble members are selected for analysis via validation against radar-estimated rainfall observations. All members feature moist soundings with low levels of free convection (LFC) and sufficient instability for deep convection. We find that stronger gradients in 100-m θe and θv in the most accurate members associated with a near-surface frontal boundary focus the lifting mechanism for deep, moist convection and enhanced rainfall. As the simulations progress, stronger southerly winds in the least accurate members advect drier mid-level air into the region of interest and shift the near-surface boundary further north and west. Analysis of the verification ensemble mean analysis reveals a strong near-surface frontal boundary similarly positioned as in the most accurate members and dry air aloft more similar to that in the least accurate members, suggesting that the positioning of the frontal boundary is more critical to accurately reproducing rainfall patterns and intensity in this case. The analyses suggest that subtle details in the simulation of frontal boundaries and mesoscale flow structures can lead to bifurcations in producing extreme or almost no rainfall. Implications for improved probabilistic forecasts of heavy rainfall events will be discussed.

Evaluation and Statistical Post‐Processing of Two Precipitation Reforecast Products During Summer in the Mainland of China

AbstractRaw weather forecasts from numerical prediction models usually suffer from systematic bias, which can be removed by statistical post‐processing methods to achieve accurate and reliable ensemble forecasts. The post‐processing of precipitation forecasts often requires reforecasts with long historical archives to ensure sufficient sample size. In this work, we provide an evaluation of two widely used short‐range reforecast products during summer in the mainland of China, including reforecasts from the Integrated Forecasting System of the European Center for Medium‐Range Weather Forecasts (ECMWF), and reforecasts from Global Ensemble Forecast System version 12 (GEFSv12) of the National Centers for Environmental Prediction. The results suggest that ECMWF reforecasts outperform GEFSv12 reforecasts in accuracy and discrimination in most regions of China, especially for heavy rain. On the other hand, the post‐processed GEFSv12 reforecasts are better than post‐processed ECMWF reforecasts in reliability for light rain in several dry regions. Moreover, we combine ECMWF and GEFSv12 reforecasts by Bayesian model averaging (BMA). The results show that BMA is able to combine the advantages of two reforecasts. Post‐processed forecasts from BMA perform as well as ECMWF reforecasts in accuracy and discrimination skill for heavy rain in wet regions of southern China. BMA results are also as reliable as post‐processed GEFSv12 reforecasts in dry regions of northern China. The evaluation results in this study could serve as a useful guide for further applications of those reforecast products in China.

A Comparative Study of Multi-Model Ensemble Forecasting Accuracy between Equal- and Variant-Weight Techniques

Previous studies on multi-model ensemble forecasting mainly focused on the weight allocation of each model, but did not discuss how to suppress the reduction of ensemble forecasting accuracy when adding poorer models. Based on a variant weight (VW) method and the equal weight (EW) method, this study explored this topic through theoretical and real case analyses. A theoretical proof is made, showing that this VW method can improve the forecasting accuracy of a multi-model ensemble, in the case of either the same models combination or adding an even worse model into the original multi-model ensemble, compared to the EW method. Comparative multi-model ensemble forecasting experiments against a real case between the VW and EW methods show that the forecasting accuracy of a multi-model ensemble applying the VW method is better than that of each individual model (including the model from the European Centre for Medium-Range Weather Forecasts). The 2 m temperature forecasting applying the VW method is superior to that applying the EW method for all the multi-model ensembles. Both theoretical proof and numerical experiments show that an improved forecast, better than a best model, is generally possible.

Visualizing RNA Structures by SAXS-Driven MD Simulations.

The biological role of biomolecules is intimately linked to their structural dynamics. Experimental or computational techniques alone are often insufficient to determine accurate structural ensembles in atomic detail. We use all-atom molecular dynamics (MD) simulations and couple it to small-angle X-ray scattering (SAXS) experiments to resolve the structural dynamics of RNA molecules. To accomplish this task, we utilize a set of re-weighting and biasing techniques tailored for RNA molecules. To showcase our approach, we study two RNA molecules: a riboswitch that shows structural variations upon ligand binding, and a two-way junction RNA that displays structural heterogeneity and sensitivity to salt conditions. Integration of MD simulations and experiments allows the accurate construction of conformational ensembles of RNA molecules. We observe a dynamic change of the SAM-I riboswitch conformations depending on its binding partners. The binding of SAM and Mg2+ cations stabilizes the compact state. The absence of Mg2+ or SAM leads to the loss of tertiary contacts, resulting in a dramatic expansion of the riboswitch conformations. The sensitivity of RNA structures to the ionic strength demonstrates itself in the helix junction helix (HJH). The HJH shows non-monotonic compaction as the ionic strength increases. The physics-based picture derived from the experimentally guided MD simulations allows biophysical characterization of RNA molecules. All in all, SAXS-guided MD simulations offer great prospects for studying RNA structural dynamics.