Combined prediction of rockburst based on multiple factors and stacking ensemble algorithm

Rockburst is a kind of common geological disaster in deep tunnel engineering.It has the characteristics of causing great harm and occurring at random locations and times.These characteristics seriously affect tunnel construction and threaten the physical and mental health and safety of workers.Therefore,it is of great significance to study the tendency of rockburst in the early stage of tunnel survey,design and construction.At present,there is no unified method and selected parameters for rockburst prediction.In view of the large difference of different rockburst criteria and the imbalance of rockburst database categories,this paper presents a two-step rockburst prediction method based on multiple factors and the stacking ensemble algorithm.Considering the influence of rock physical and mechanical parameters,tunnel face conditions and excavation disturbance,multiple rockburst criteria are predicted by integrating multiple machine learning algorithms.A combined prediction model of rockburst criteria is established,and the results of each rockburst criterion index are weighted and combined,with the weight updated using the field rockburst record.The dynamic weight is combined with the cloud model to comprehensively evaluate the regional rockburst risk.Field results from applying the model in the Grand Canyon tunnel show that the rockburst prediction method proposed in this paper has better applicability and higher accuracy than the single rockburst criterion.

Ligand Based Virtual Screening of Molecular Compounds in Drug Discovery Using GCAN Fingerprint and Ensemble Machine Learning Algorithm

The drug development process takes a long time since it requires sorting through a large number of inactive compounds from a large collection of compounds chosen for study and choosing just the most pertinent compounds that can bind to a disease protein.The use of virtual screening in pharmaceutical research is growing in popularity.During the early phases of medication research and development,it is crucial.Chemical compound searches are nowmore narrowly targeted.Because the databases containmore andmore ligands,thismethod needs to be quick and exact.Neural network fingerprints were created more effectively than the well-known Extended Connectivity Fingerprint(ECFP).Only the largest sub-graph is taken into consideration to learn the representation,despite the fact that the conventional graph network generates a better-encoded fingerprint.When using the average or maximum pooling layer,it also contains unrelated data.This article suggested the Graph Convolutional Attention Network(GCAN),a graph neural network with an attention mechanism,to address these problems.Additionally,it makes the nodes or sub-graphs that are used to create the molecular fingerprint more significant.The generated fingerprint is used to classify drugs using ensemble learning.As base classifiers,ensemble stacking is applied to Support Vector Machines(SVM),Random Forest,Nave Bayes,Decision Trees,AdaBoost,and Gradient Boosting.When compared to existing models,the proposed GCAN fingerprint with an ensemble model achieves relatively high accuracy,sensitivity,specificity,and area under the curve.Additionally,it is revealed that our ensemble learning with generated molecular fingerprint yields 91%accuracy,outperforming earlier approaches.

The study of intelligent algorithm in particle identification of heavy-ion collisions at low and intermediate energies

Traditional particle identification methods face timeconsuming,experience-dependent,and poor repeatability challenges in heavy-ion collisions at low and intermediate energies.Researchers urgently need solutions to the dilemma of traditional particle identification methods.This study explores the possibility of applying intelligent learning algorithms to the particle identification of heavy-ion collisions at low and intermediate energies.Multiple intelligent algorithms,including XgBoost and TabNet,were selected to test datasets from the neutron ion multi-detector for reaction-oriented dynamics(NIMROD-ISiS)and Geant4 simulation.Tree-based machine learning algorithms and deep learning algorithms e.g.TabNet show excellent performance and generalization ability.Adding additional data features besides energy deposition can improve the algorithm’s performance when the data distribution is nonuniform.Intelligent learning algorithms can be applied to solve the particle identification problem in heavy-ion collisions at low and intermediate energies.

Structural Damage Identification Using Ensemble Deep Convolutional Neural Network Models

The existing strategy for evaluating the damage condition of structures mostly focuses on feedback supplied by traditional visualmethods,which may result in an unreliable damage characterization due to inspector subjectivity or insufficient level of expertise.As a result,a robust,reliable,and repeatable method of damage identification is required.Ensemble learning algorithms for identifying structural damage are evaluated in this article,which use deep convolutional neural networks,including simple averaging,integrated stacking,separate stacking,and hybridweighted averaging ensemble and differential evolution(WAE-DE)ensemblemodels.Damage identification is carried out on three types of damage.The proposed algorithms are used to analyze the damage of 4585 structural images.The effectiveness of the ensemble learning techniques is evaluated using the confusion matrix.For the testing dataset,the confusion matrix achieved an accuracy of 94 percent and a minimum recall of 92 percent for the best model(WAE-DE)in distinguishing damage types as flexural,shear,combined,or undamaged.

Assimilation of Remote Sensing and Crop Model for LAI Estimation Based on Ensemble Kalman Filter

Data assimilation in agricultural remote sensing research is of great significance to integrate with remote sensing observations and model simulations for parameters estimation. The present investigation not only designed and realized the Ensemble Kalman Filtering algorithm （EnKF） assimilation by combing the crop growth model （CERES-Wheat） with remote sensing data, but also optimized and updated the key parameters （LAI） of winter wheat by using remote sensing data. Results showed that the assimilation LAI and the observation ones agreed with each other, and the R2 reached 0.8315. So assimilation remote sensing and crop model could provide reference data for the agricultural production.

Novel Partitioned Time-Stepping Algorithms for Fast Computation of Random Interface-Coupled Problems with Uncertain Parameters

The simulation of multi-domain,multi-physics mathematical models with uncertain parameters can be quite demanding in terms of algorithm design and com-putation costs.Our main objective in this paper is to examine a physical interface coupling between two random dissipative systems with uncertain parameters.Due to the complexity and uncertainty inherent in such interface-coupled problems,un-certain diffusion coefficients or friction parameters often arise,leading to consid-ering random systems.We employ Monte Carlo methods to produce independent and identically distributed deterministic heat-heat model samples to address ran-dom systems,and adroitly integrate the ensemble idea to facilitate the fast calcu-lation of these samples.To achieve unconditional stability,we introduce the scalar auxiliary variable(SAV)method to overcome the time constraints of the ensemble implicit-explicit algorithm.Furthermore,for a more accurate and stable scheme,the ensemble data-passing algorithm is raised,which is unconditionally stable and convergent without any auxiliary variables.These algorithms employ the same co-efficient matrix for multiple linear systems and enable easy parallelization,which can significantly reduce the computational cost.Finally,numerical experiments are conducted to support the theoretical results and showcase the unique features of the proposed algorithms.

Ensemble hybrid machine learning methods for gully erosion susceptibility mapping: K-fold cross validation approach

Gully erosion is one of the important problems creating barrier to agricultural development.The present research used the radial basis function neural network(RBFnn)and its ensemble with random sub-space(RSS)and rotation forest(RTF)ensemble Meta classifiers for the spatial mapping of gully erosion susceptibility(GES)in Hinglo river basin.120 gullies were marked and grouped into four-fold.A total of 23 factors including topographical,hydrological,lithological,and soil physio-chemical properties were effectively used.GES maps were built by RBFnn,RSS-RBFnn,and RTF-RBFnn models.The very high susceptibility zone of RBFnn,RTF-RBFnn and RSS-RBFnn models covered 6.75%,6.72%and 6.57%in Fold-1,6.21%,6.10%and 6.09%in Fold-2,6.26%,6.13%and 6.05%in Fold-3 and 7%,6.975%and 6.42%in Fold-4 of the basin.Receiver operating characteristics(ROC)curve and statistical techniques such as mean-absolute-error(MAE),root-mean-absolute-error(RMSE)and relative gully density area(R-index)methods were used for evaluating the GES maps.The results of the ROC,MAE,RMSE and R-index methods showed that the models of susceptibility to gully erosion have excellent predictive efficiency.The simulation results based on machine learning are satisfactory and outstanding and could be used to forecast the areas vulnerable to gully erosion.

A SVD-based ensemble projection algorithm for calculating the conditional nonlinear optimal perturbation

Conditional nonlinear optimal perturbation(CNOP) is an extension of the linear singular vector technique in the nonlinear regime.It represents the initial perturbation that is subjected to a given physical constraint,and results in the largest nonlinear evolution at the prediction time.CNOP-type errors play an important role in the predictability of weather and climate.Generally,when calculating CNOP in a complicated numerical model,we need the gradient of the objective function with respect to the initial perturbations to provide the descent direction for searching the phase space.The adjoint technique is widely used to calculate the gradient of the objective function.However,it is difficult and cumbersome to construct the adjoint model of a complicated numerical model,which imposes a limitation on the application of CNOP.Based on previous research,this study proposes a new ensemble projection algorithm based on singular vector decomposition(SVD).The new algorithm avoids the localization procedure of previous ensemble projection algorithms,and overcomes the uncertainty caused by choosing the localization radius empirically.The new algorithm is applied to calculate the CNOP in an intermediate forecasting model.The results show that the CNOP obtained by the new ensemble-based algorithm can effectively approximate that calculated by the adjoint algorithm,and retains the general spatial characteristics of the latter.Hence,the new SVD-based ensemble projection algorithm proposed in this study is an effective method of approximating the CNOP.

A better carbon-water flux simulation in multiple vegetation types by data assimilation

Background:The accurate estimation of carbon-water flux is critical for understanding the carbon and water cycles of terrestrial ecosystems and further mitigating climate change.Model simulations and observations have been widely used to research water and carbon cycles of terrestrial ecosystems.Given the advantages and limitations of each method,combining simulations and observations through a data assimilation technique has been proven to be highly promising for improving carbon-water flux simulation.However,to the best of our knowledge,few studies have accomplished both parameter optimization and the updating of model state variables through data assimilation for carbon-water flux simulation in multiple vegetation types.And little is known about the variation of the performance of data assimilation for carbon-water flux simulation in different vegetation types.Methods:In this study,we assimilated leaf area index(LAI)time-series observations into a biogeochemical model(Biome-BGC)using different assimilation algorithms(ensemble Kalman filter algorithm(EnKF)and unscented Kalman filter(UKF))in different vegetation types(deciduous broad-leaved forest(DBF),evergreen broad-leaved forest(EBF)and grassland(GL))to simulate carbon-water flux.Results:The validation of the results against the eddy covariance measurements indicated that,overall,compared with the original simulation,assimilating the LAI into the Biome-BGC model improved the carbon-water flux simulations(R^(2)increased by 35%,root mean square error decreased by 10%;the sum of the absolute error decreased by 8%)but more significantly,improved the water flux simulations(R^(2)increased by 31%,root mean square error decreased by 18%;the sum of the absolute error decreased by 16%).Among the different forest types,the data assimilation techniques(both EnKF and UKF)achieved the best performance towards carbon-water flux in EBF(R^(2)increased by 44%,root mean square error decreased by 24%;the sum of the absolute error decreased by 28%),and the performances of EnKF and UKF showed slightly different when simulating carbon fluxes.Conclusion:We suggest that to reduce the uncertainty in global carbon-water flux quantification,forthcoming data assimilation treatment should consider the vegetation types where the data assimilation experiments are carried out,the simulated objectives and the assimilation algorithms.

Adaptive Fault Detection Scheme Using an Optimized Self-healing Ensemble Machine Learning Algorithm

This paper proposes a new cost-efficient,adaptive,and self-healing algorithm in real time that detects faults in a short period with high accuracy,even in the situations when it is difficult to detect.Rather than using traditional machine learning(ML)algorithms or hybrid signal processing techniques,a new framework based on an optimization enabled weighted ensemble method is developed that combines essential ML algorithms.In the proposed method,the system will select and compound appropriate ML algorithms based on Particle Swarm Optimization(PSO)weights.For this purpose,power system failures are simulated by using the PSCA D-Python co-simulation.One of the salient features of this study is that the proposed solution works on real-time raw data without using any pre-computational techniques or pre-stored information.Therefore,the proposed technique will be able to work on different systems,topologies,or data collections.The proposed fault detection technique is validated by using PSCAD-Python co-simulation on a modified and standard IEEE-14 and standard IEEE-39 bus considering network faults which are difficult to detect.

Novel machine learning framework for thermal conductivity prediction by crystal graph convolution embedded ensemble

With the advent of the big data era,artificial intelligence technology has penetrated and deeply affected our daily life.In addition,data-based machine learning algorithms have been applied to physics,chemistry,material science,and other basic science fields.However,the scarcity of data sets is known as the main obstacle to its development.Mining effective information from the limited data samples and building an appropriate machine learning algorithms framework are the major breakthroughs.For solid materials,the intrinsic properties are closely related to their atomic composition and relative positions,namely crystal structures.Here,inspired by the emerging of graph convolution neural network and material crystal graph,we proposed an integrated algorithms framework embedded crystal graph to train and predict the lattice thermal conductivities of crystal materials.This machine learning algorithms framework showed superior learning and generalization ability.In addition,not only in predicting thermal conductivities,but our framework also has great performance in predicting other phonon or electron-related properties.This strategy provided a new approach in the design of machine learning framework,which indicated the great potential for the application of machine learning in material science.

Ensemble Prediction of Monsoon Index with a Genetic Neural Network Model

After the consideration of the nonlinear nature changes of monsoon index,and the subjective determination of network structure in traditional artificial neural network prediction modeling,monthly and seasonal monsoon intensity index prediction is studied in this paper by using nonlinear genetic neural network ensemble prediction（GNNEP）modeling.It differs from traditional prediction modeling in the following aspects： （1）Input factors of the GNNEP model of monsoon index were selected from a large quantity of preceding period high correlation factors,such as monthly sea temperature fields,monthly 500-hPa air temperature fields,monthly 200-hPa geopotential height fields,etc.,and they were also highly information-condensed and system dimensionality-reduced by using the empirical orthogonal function（EOF）method,which effectively condensed the useful information of predictors and therefore controlled the size of network structure of the GNNEP model.（2）In the input design of the GNNEP model,a mean generating function（MGF）series of predictand（monsoon index）was added as an input factor;the contrast analysis of results of predic- tion experiments by a physical variable predictor-predictand MGF GNNEP model and a physical variable predictor GNNEP model shows that the incorporation of the periodical variation of predictand（monsoon index）is very effective in improving the prediction of monsoon index.（3）Different from the traditional neural network modeling,the GNNEP modeling is able to objectively determine the network structure of the GNNNEP model,and the model constructed has a better generalization capability.In the case of identical predictors,prediction modeling samples,and independent prediction samples,the prediction accuracy of our GNNEP model combined with the system dimensionality reduction technique of predictors is clearly higher than that of the traditional stepwise regression model using the traditional treatment technique of predictors,suggesting that the GNNEP model opens up a vast range of possibilities for operational weather prediction.

Integration of Landsat time-series vegetation indices improves consistency of change detection

Vegetation indices(VIs)were used to detect when and where vegetation changes occurred.However,different VIs have different or even diametrically opposite results,which obstructed the in-depth understanding of vegetation change.Therefore,this study examined the spatial and temporal consistency offive VIs(EVI;NBR;NDMI;NDVI;and NIRv)in detecting abrupt and gradual vegetation changes,and provided an ensemble algorithm which integrated the change detection results of thefive indices to reduce the uncertainty of change detection using a single index.The spatial consistency of thefive indices in abrupt change detection accounted for 50.6%of the study area,but the temporal consistency was low(21.6%).Wetness indices(NBR,NDMI)were more sensitive to negative abrupt changes,greenness indices(EVI,NDVI,NIRv)were more sensitive to positive abrupt changes;and both types of indices were similar in detecting gradual and total changes.The overall accuracy of the ensemble method was 81.60%and higher than that of any single index in abrupt change detection.This study provides a comprehensive evaluation of the spatial and temporal inconsistencies of change detection in model-fitting errors and various types of vegetation changes.The proposed ensemble method can support robust change-detection.