Albacore tuna(Thunnus alalunga)is one of the target species of tuna longline fishing,and waters near the Cook Islands are a vital albacore tuna fishing ground.Marine environmental data are usually presented with diffe...Albacore tuna(Thunnus alalunga)is one of the target species of tuna longline fishing,and waters near the Cook Islands are a vital albacore tuna fishing ground.Marine environmental data are usually presented with different spatial resolutions,which leads to different results in tuna fishery prediction.Study on the impact of different spatial resolutions on the prediction accuracy of albacore tuna fishery to select the best spatial resolution can contribute to better management of albacore tuna resources.The nominal catch per unit effort(CPUE)of albacore tuna is calculated according to vessel monitor system(VMS)data collected from Chinese distantwater fishery enterprises from January 1,2017 to May 31,2021.A total of 26 spatiotemporal and environmental factors,including temperature,salinity,dissolved oxygen of 0–300 m water layer,chlorophyll-a concentration in the sea surface,sea surface height,month,longitude,and latitude,were selected as variables.The temporal resolution of the variables was daily and the spatial resolutions were set to be 0.5°×0.5°,1°×1°,2°×2°,and 5°×5°.The relationship between the nominal CPUE and each individual factor was analyzed to remove the factors irrelavant to the nominal CPUE,together with a multicollinearity diagnosis on the factors to remove factors highly related to the other factors within the four spatial resolutions.The relationship models between CPUE and spatiotemporal and environmental factors by four spatial resolutions were established based on the long short-term memory(LSTM)neural network model.The mean absolute error(MAE)and root mean square error(RMSE)were used to analyze the fitness and accuracy of the models,and to determine the effects of different spatial resolutions on the prediction accuracy of the albacore tuna fishing ground.The results show the resolution of 1°×1°can lead to the best prediction accuracy,with the MAE and RMSE being 0.0268 and 0.0452 respectively,followed by 0.5°×0.5°,2°×2°and 5°×5°with declining prediction accuracy.The results suggested that 1)albacore tuna fishing ground can be predicted by LSTM;2)the VMS records the data in detail and can be used scientifically to calculate the CPUE;3)correlation analysis,and multicollinearity diagnosis are necessary to improve the prediction accuracy of the model;4)the spatial resolution should be 1°×1°in the forecast of albacore tuna fishing ground in waters near the Cook Islands.展开更多
The alpha–beta filter algorithm has been widely researched for various applications,for example,navigation and target tracking systems.To improve the dynamic performance of the alpha–beta filter algorithm,a new pred...The alpha–beta filter algorithm has been widely researched for various applications,for example,navigation and target tracking systems.To improve the dynamic performance of the alpha–beta filter algorithm,a new prediction learning model is proposed in this study.The proposed model has two main components:(1)the alpha–beta filter algorithm is the main prediction module,and(2)the learning module is a feedforward artificial neural network(FF‐ANN).Furthermore,the model uses two inputs,temperature sensor and humidity sensor data,and a prediction algorithm is used to predict actual sensor readings from noisy sensor readings.Using the novel proposed technique,prediction accuracy is significantly improved while adding the feed‐forward backpropagation neural network,and also reduces the root mean square error(RMSE)and mean absolute error(MAE).We carried out different experiments with different experimental setups.The proposed model performance was evaluated with the traditional alpha–beta filter algorithm and other algorithms such as the Kalman filter.A higher prediction accuracy was achieved,and the MAE and RMSE were 35.1%–38.2%respectively.The final proposed model results show increased performance when compared to traditional methods.展开更多
Crop Yield Prediction(CYP)is critical to world food production.Food safety is a top priority for policymakers.They rely on reliable CYP to make import and export decisions that must be fulfilled before launching an ag...Crop Yield Prediction(CYP)is critical to world food production.Food safety is a top priority for policymakers.They rely on reliable CYP to make import and export decisions that must be fulfilled before launching an agricultural business.Crop Yield(CY)is a complex variable influenced by multiple factors,including genotype,environment,and their interactions.CYP is a significant agrarian issue.However,CYP is the main task due to many composite factors,such as climatic conditions and soil characteristics.Machine Learning(ML)is a powerful tool for supporting CYP decisions,including decision support on which crops to grow in a specific season.Generally,Artificial Neural Networks(ANN)are usually used to predict the behaviour of complex non-linear models.As a result,this research paper attempts to determine the correlations between climatic variables,soil nutrients,and CYwith the available data.InANN,threemethods,Levenberg-Marquardt(LM),Bayesian regularisation(BR),and scaled conjugate gradient(SCG),are used to train the neural network(NN)model and then compared to determine prediction accuracy.The performance measures of the training,as declared above,such as Mean Squared Error(MSE)and correlation coefficient(R),were determined to assess the ANN models that had been built.The experimental study proves that LM training algorithms are better,while BR and SCG have minimal performance.展开更多
A hard problem that hinders the movement of waxy crude oil is wax deposition in oil pipelines.To ensure the safe operation of crude oil pipelines,an accurate model must be developed to predict the rate of wax depositi...A hard problem that hinders the movement of waxy crude oil is wax deposition in oil pipelines.To ensure the safe operation of crude oil pipelines,an accurate model must be developed to predict the rate of wax deposition in crude oil pipelines.Aiming at the shortcomings of the ENN prediction model,which easily falls into the local minimum value and weak generalization ability in the implementation process,an optimized ENN prediction model based on the IRSA is proposed.The validity of the new model was confirmed by the accurate prediction of two sets of experimental data on wax deposition in crude oil pipelines.The two groups of crude oil wax deposition rate case prediction results showed that the average absolute percentage errors of IRSA-ENN prediction models is 0.5476% and 0.7831%,respectively.Additionally,it shows a higher prediction accuracy compared to the ENN prediction model.In fact,the new model established by using the IRSA to optimize ENN can optimize the initial weights and thresholds in the prediction process,which can overcome the shortcomings of the ENN prediction model,such as weak generalization ability and tendency to fall into the local minimum value,so that it has the advantages of strong implementation and high prediction accuracy.展开更多
Background:Recently,machine learning(ML)has become attractive in genomic prediction,but its superiority in genomic prediction over conventional(ss)GBLUP methods and the choice of optimal ML methods need to be investig...Background:Recently,machine learning(ML)has become attractive in genomic prediction,but its superiority in genomic prediction over conventional(ss)GBLUP methods and the choice of optimal ML methods need to be investigated.Results:In this study,2566 Chinese Yorkshire pigs with reproduction trait records were genotyped with the GenoBaits Porcine SNP 50 K and PorcineSNP50 panels.Four ML methods,including support vector regression(SVR),kernel ridge regression(KRR),random forest(RF)and Adaboost.R2 were implemented.Through 20 replicates of fivefold cross-validation(CV)and one prediction for younger individuals,the utility of ML methods in genomic prediction was explored.In CV,compared with genomic BLUP(GBLUP),single-step GBLUP(ssGBLUP)and the Bayesian method BayesHE,ML methods significantly outperformed these conventional methods.ML methods improved the genomic prediction accuracy of GBLUP,ssGBLUP,and BayesHE by 19.3%,15.0% and 20.8%,respectively.In addition,ML methods yielded smaller mean squared error(MSE)and mean absolute error(MAE)in all scenarios.ssGBLUP yielded an improvement of 3.8% on average in accuracy compared to that of GBLUP,and the accuracy of BayesHE was close to that of GBLUP.In genomic prediction of younger individuals,RF and Adaboost.R2_KRR performed better than GBLUP and BayesHE,while ssGBLUP performed comparably with RF,and ssGBLUP yielded slightly higher accuracy and lower MSE than Adaboost.R2_KRR in the prediction of total number of piglets born,while for number of piglets born alive,Adaboost.R2_KRR performed significantly better than ssGBLUP.Among ML methods,Adaboost.R2_KRR consistently performed well in our study.Our findings also demonstrated that optimal hyperparameters are useful for ML methods.After tuning hyperparameters in CV and in predicting genomic outcomes of younger individuals,the average improvement was 14.3% and 21.8% over those using default hyperparameters,respectively.Conclusion:Our findings demonstrated that ML methods had better overall prediction performance than conventional genomic selection methods,and could be new options for genomic prediction.Among ML methods,Adaboost.R2_KRR consistently performed well in our study,and tuning hyperparameters is necessary for ML methods.The optimal hyperparameters depend on the character of traits,datasets etc.展开更多
Data is always a crucial issue of concern especially during its prediction and computation in digital revolution.This paper exactly helps in providing efficient learning mechanism for accurate predictability and reduc...Data is always a crucial issue of concern especially during its prediction and computation in digital revolution.This paper exactly helps in providing efficient learning mechanism for accurate predictability and reducing redundant data communication.It also discusses the Bayesian analysis that finds the conditional probability of at least two parametric based predictions for the data.The paper presents a method for improving the performance of Bayesian classification using the combination of Kalman Filter and K-means.The method is applied on a small dataset just for establishing the fact that the proposed algorithm can reduce the time for computing the clusters from data.The proposed Bayesian learning probabilistic model is used to check the statistical noise and other inaccuracies using unknown variables.This scenario is being implemented using efficient machine learning algorithm to perpetuate the Bayesian probabilistic approach.It also demonstrates the generative function forKalman-filer based prediction model and its observations.This paper implements the algorithm using open source platform of Python and efficiently integrates all different modules to piece of code via Common Platform Enumeration(CPE)for Python.展开更多
Among the advantages of using industrial robots for machining applications instead of machine tools are flexibility, cost effectiveness, and versatility. Due to the kinematics of the articulated robot, the system beha...Among the advantages of using industrial robots for machining applications instead of machine tools are flexibility, cost effectiveness, and versatility. Due to the kinematics of the articulated robot, the system behaviour is quite different compared with machine tools. Two major questions arise in implementing robots in machining tasks: one is the robot’s stiffness, and the second is the achievable machined part accuracy, which varies mainly due to the huge variety of robot models. This paper proposes error prediction model in the application of industrial robot for machining tasks, based on stiffness and accuracy limits. The research work includes experimental and theoretical parts. Advanced machining and inspection tools were applied, as well as a theoretical model of the robot structure and stiffness based on the form-shaping function approach. The robot machining performances, from the workpiece accuracy point of view were predicted.展开更多
Determining an optimal sample size is a key step in designing field surveys,and is particularly important for detecting the spatial pattern of highly variable properties such as soil organic carbon(SOC).Based on 550 s...Determining an optimal sample size is a key step in designing field surveys,and is particularly important for detecting the spatial pattern of highly variable properties such as soil organic carbon(SOC).Based on 550 soil sampling points in the nearsurface layer(0 to 20 cm)in a representative region of northern China's agro-pastoral ecotone,we studied effects of four interpolation methods such as ordinary kriging(OK),universal kriging(UK),inverse distance weighting(IDW)and radial basis function(RBF)and random subsampling(50,100,200,300,400,and 500)on the prediction accuracy of SOC estimation.When the Shannon's Diversity Index(SHDI)and Shannon's Evenness Index(SHEI)was 2.01 and 0.67,the OK method appeared to be a superior method,which had the smallest root mean square error(RMSE)and the mean error(ME)nearest to zero.On the contrary,the UK method performed poorly for the interpolation of SOC in the present study.The sample size of 200 had the most accurate prediction;50 sampling points produced the worst prediction accuracy.Thus,we used 200 samples to estimate the study area's soil organic carbon density(SOCD)by the OK method.The total SOC storage to a depth of 20 cm in the study area was 117.94 Mt,and its mean SOCD was 2.40 kg/m2.The SOCD kg/(C⋅m2)of different land use types were in the following order:woodland(3.29)>grassland(2.35)>cropland(2.19)>sandy land(1.55).展开更多
Stock market trends forecast is one of the most current topics and a significant research challenge due to its dynamic and unstable nature.The stock data is usually non-stationary,and attributes are non-correlative to...Stock market trends forecast is one of the most current topics and a significant research challenge due to its dynamic and unstable nature.The stock data is usually non-stationary,and attributes are non-correlative to each other.Several traditional Stock Technical Indicators(STIs)may incorrectly predict the stockmarket trends.To study the stock market characteristics using STIs and make efficient trading decisions,a robust model is built.This paper aims to build up an Evolutionary Deep Learning Model(EDLM)to identify stock trends’prices by using STIs.The proposed model has implemented the Deep Learning(DL)model to establish the concept of Correlation-Tensor.The analysis of the dataset of three most popular banking organizations obtained from the live stock market based on the National Stock exchange(NSE)-India,a Long Short Term Memory(LSTM)is used.The datasets encompassed the trading days from the 17^(th) of Nov 2008 to the 15^(th) of Nov 2018.This work also conducted exhaustive experiments to study the correlation of various STIs with stock price trends.The model built with an EDLM has shown significant improvements over two benchmark ML models and a deep learning one.The proposed model aids investors in making profitable investment decisions as it presents trend-based forecasting and has achieved a prediction accuracy of 63.59%,56.25%,and 57.95%on the datasets of HDFC,Yes Bank,and SBI,respectively.Results indicate that the proposed EDLA with a combination of STIs can often provide improved results than the other state-of-the-art algorithms.展开更多
Background:The suitability and performance of the bankruptcy prediction models is an empirical question.The aim of this paper is to develop a bankruptcy prediction model for Indian manufacturing companies on a sample ...Background:The suitability and performance of the bankruptcy prediction models is an empirical question.The aim of this paper is to develop a bankruptcy prediction model for Indian manufacturing companies on a sample of 208 companies consisting of an equal number of defaulted and non-defaulted firms.Out of 208 companies,130 are used for estimation sample,and 78 are holdout for model validation.The study reestimates the accounting based models such as Altman EI(Journal of Finance 23:19189-209,1968)Z-Score,Ohlson JA(Journal of Accounting Research 18:109-131,1980)Y-Score and Zmijewski ME(Journal of Accounting Research 22:59-82,1984)X-Score model.The paper compares original and re-estimated models to explore the sensitivity of these models towards the change in time periods and financial conditions.Methods:Multiple Discriminant Analysis(MDA)and Probit techniques are employed in the estimation of Z-Score and X-Score models,whereas Logit technique is employed in the estimation of Y-Score and the newly proposed models.The performance of all the original,re-estimated and new proposed models are assessed by predictive accuracy,significance of parameters,long-range accuracy,secondary sample and Receiver Operating Characteristic(ROC)tests.Results:The major findings of the study reveal that the overall predictive accuracy of all the three models improves on estimation and holdout sample when the coefficients are re-estimated.Amongst the contesting models,the new bankruptcy prediction model outperforms other models.Conclusions:The industry specific model should be developed with the new combinations of financial ratios to predict bankruptcy of the firms in a particular country.The study further suggests the coefficients of the models are sensitive to time periods and financial condition.Hence,researchers should be cautioned while choosing the models for bankruptcy prediction to recalculate the models by looking at the recent data in order to get higher predictive accuracy.展开更多
To realize data sharing,and to fully use the data value,breaking the data island between institutions to realize data collaboration has become a new sharing mode.This paper proposed a distributed data security sharing...To realize data sharing,and to fully use the data value,breaking the data island between institutions to realize data collaboration has become a new sharing mode.This paper proposed a distributed data security sharing scheme based on C/S communication mode,and constructed a federated learning architecture that uses differential privacy technology to protect training parameters.Clients do not need to share local data,and they only need to upload the trained model parameters to achieve data sharing.In the process of training,a distributed parameter update mechanism is introduced.The server is mainly responsible for issuing training commands and parameters,and aggregating the local model parameters uploaded by the clients.The client mainly uses the stochastic gradient descent algorithm for gradient trimming,updates,and transmits the trained model parameters back to the server after differential processing.To test the performance of the scheme,in the application scenario where many medical institutions jointly train the disease detection system,the model is tested from multiple perspectives by taking medical data as an example.From the testing results,we can know that for this specific test dataset,when the parameters are properly configured,the lowest prediction accuracy rate is 90.261%and the highest accuracy rate is up to 94.352.It shows that the performance of the model is good.The results also show that this scheme realizes data sharing while protecting data privacy,completes accurate prediction of diseases,and has a good effect.展开更多
In recent years,real-time video streaming has grown in popularity.The growing popularity of the Internet of Things(IoT)and other wireless heterogeneous networks mandates that network resources be carefully apportioned...In recent years,real-time video streaming has grown in popularity.The growing popularity of the Internet of Things(IoT)and other wireless heterogeneous networks mandates that network resources be carefully apportioned among versatile users in order to achieve the best Quality of Experience(QoE)and performance objectives.Most researchers focused on Forward Error Correction(FEC)techniques when attempting to strike a balance between QoE and performance.However,as network capacity increases,the performance degrades,impacting the live visual experience.Recently,Deep Learning(DL)algorithms have been successfully integrated with FEC to stream videos across multiple heterogeneous networks.But these algorithms need to be changed to make the experience better without sacrificing packet loss and delay time.To address the previous challenge,this paper proposes a novel intelligent algorithm that streams video in multi-home heterogeneous networks based on network-centric characteristics.The proposed framework contains modules such as Intelligent Content Extraction Module(ICEM),Channel Status Monitor(CSM),and Adaptive FEC(AFEC).This framework adopts the Cognitive Learning-based Scheduling(CLS)Module,which works on the deep Reinforced Gated Recurrent Networks(RGRN)principle and embeds them along with the FEC to achieve better performances.The complete framework was developed using the Objective Modular Network Testbed in C++(OMNET++),Internet networking(INET),and Python 3.10,with Keras as the front end and Tensorflow 2.10 as the back end.With extensive experimentation,the proposed model outperforms the other existing intelligentmodels in terms of improving the QoE,minimizing the End-to-End Delay(EED),and maintaining the highest accuracy(98%)and a lower Root Mean Square Error(RMSE)value of 0.001.展开更多
The social engineering cyber-attack is where culprits mislead the users by getting the login details which provides the information to the evil server called phishing.The deep learning approaches and the machine learn...The social engineering cyber-attack is where culprits mislead the users by getting the login details which provides the information to the evil server called phishing.The deep learning approaches and the machine learning are compared in the proposed system for presenting the methodology that can detect phishing websites via Uniform Resource Locator(URLs)analysis.The legal class is composed of the home pages with no inclusion of login forms in most of the present modern solutions,which deals with the detection of phishing.Contrarily,the URLs in both classes from the login page due,considering the representation of a real case scenario and the demonstration for obtaining the rate of false-positive with the existing approaches during the legal login pages provides the test having URLs.In addition,some model reduces the accuracy rather than training the base model and testing the latest URLs.In addition,a feature analysis is performed on the present phishing domains to identify various approaches to using the phishers in the campaign.A new dataset called the MUPD dataset is used for evaluation.Lastly,a prediction model,the Dense forward-backwards Long Short Term Memory(LSTM)model(d−FBLSTM),is presented for combining the forward and backward propagation of LSMT to obtain the accuracy of 98.5%on the initiated login URL dataset.展开更多
The prediction of mild cognitive impairment or Alzheimer’s disease(AD)has gained the attention of huge researchers as the disease occurrence is increasing,and there is a need for earlier prediction.Regrettably,due to...The prediction of mild cognitive impairment or Alzheimer’s disease(AD)has gained the attention of huge researchers as the disease occurrence is increasing,and there is a need for earlier prediction.Regrettably,due to the highdimensionality nature of neural data and the least available samples,modelling an efficient computer diagnostic system is highly solicited.Learning approaches,specifically deep learning approaches,are essential in disease prediction.Deep Learning(DL)approaches are successfully demonstrated for their higher-level performance in various fields like medical imaging.A novel 3D-Convolutional Neural Network(3D-CNN)architecture is proposed to predict AD with Magnetic resonance imaging(MRI)data.The proposed model predicts the AD occurrence while the existing approaches lack prediction accuracy and perform binary classification.The proposed prediction model is validated using the Alzheimer’s disease Neuro-Imaging Initiative(ADNI)data.The outcomes demonstrate that the anticipated model attains superior prediction accuracy and works better than the brain-image dataset’s general approaches.The predicted model reduces the human effort during the prediction process and makes it easier to diagnose it intelligently as the feature learning is adaptive.Keras’experimentation is carried out,and the model’s superiority is compared with various advanced approaches for multi-level classification.The proposed model gives better prediction accuracy,precision,recall,and F-measure than other systems like Long Short Term Memory-Recurrent Neural Networks(LSTM-RNN),Stacked Autoencoder with Deep Neural Networks(SAE-DNN),Deep Convolutional Neural Networks(D-CNN),Two Dimensional Convolutional Neural Networks(2D-CNN),Inception-V4,ResNet,and Two Dimensional Convolutional Neural Networks(3D-CNN).展开更多
A new model for predicting the total tree height for harvested stems from cut-to-length(CTL)harvester data was constructed for Pinus radiata(D.Don)following a conceptual analysis of relative stem profi les,comparisons...A new model for predicting the total tree height for harvested stems from cut-to-length(CTL)harvester data was constructed for Pinus radiata(D.Don)following a conceptual analysis of relative stem profi les,comparisons of candidate models forms and extensive selections of predictor variables.Stem profi les of more than 3000 trees in a taper data set were each processed 6 times through simulated log cutting to generate the data required for this purpose.The CTL simulations not only mimicked but also covered the full range of cutting patterns of nearly 0.45×106 stems harvested during both thinning and harvesting operations.The single-equation model was estimated through the multipleequation generalized method of moments estimator to obtain effi cient and consistent parameter estimates in the presence of error correlation and heteroscedasticity that were inherent to the systematic structure of the data.The predictive performances of our new model in its linear and nonlinear form were evaluated through a leave-one-tree-out cross validation process and compared against that of the only such existing model.The evaluations and comparisons were made through benchmarking statistics both globally over the entire data space and locally within specifi c subdivisions of the data space.These statistics indicated that the nonlinear form of our model was the best and its linear form ranked second.The prediction accuracy of our nonlinear model improved when the total log length represented more than 20%of the total tree height.The poorer performance of the existing model was partly attributed to the high degree of multicollinearity among its predictor variables,which led to highly variable and unstable parameter estimates.Our new model will facilitate and widen the utilization of harvester data far beyond the current limited use for monitoring and reporting log productions in P.radiata plantations.It will also facilitate the estimation of bark thickness and help make harvester data a potential source of taper data to reduce the intensity and cost of the conventional destructive taper sampling in the fi eld.Although developed for P.radiata,the mathematical form of our new model will be applicable to other tree species for which CTL harvester data are routinely captured during thinning and harvesting operations.展开更多
In order to study the spatial structure and dynamical mechanism of extreme precipitation in East Asia, a corresponding climate network is constructed by employing the method of event synchronization. It is found that ...In order to study the spatial structure and dynamical mechanism of extreme precipitation in East Asia, a corresponding climate network is constructed by employing the method of event synchronization. It is found that the area of East Asian summer extreme precipitation can be separated into two regions: one with high area-weighted connectivity receiving heavy precipitation mostly during the active phase of the East Asian Summer Monsoon(EASM),and another one with low area-weighted connectivity receiving heavy precipitation during both the active and the retreating phase of the EASM. Besides, a new way for the prediction of extreme precipitation is also developed by constructing a directed climate networks. The simulation accuracy in East Asia is 58% with a 0-day lead, and the prediction accuracy is 21% and average 12% with a 1-day and an n-day(2≤n≤10) lead, respectively. Compared to the normal EASM year, the prediction accuracy is low in weak years and high in strong years, which is relevant to the differences of correlations and extreme precipitation rates in different EASM situations. Recognizing and indentifying these effects is good for understanding and predicting extreme precipitation in East Asia.展开更多
Grey prediction is vital in statistical prediction with wide applications.However,most grey prediction methods focus on annual predictions of the monotonic time series instead of the seasonal time series.The paper use...Grey prediction is vital in statistical prediction with wide applications.However,most grey prediction methods focus on annual predictions of the monotonic time series instead of the seasonal time series.The paper uses the extended model of the grey GM(1,1)model to predict the seasonal time series.Some improvements have been made in two aspects to improve the prediction accuracy of the model.1)We introduce seasonal multiple factors to transform the original time series,which improves the adaptability of the seasonal data to the model.The transformed series conforms to the law presented by the model.2)The seasonal data are in superimposed sine and cosine fluctuations with tendencies.Therefore,the paper extends the grey action quantity of the traditional GM(1,1)model.The newly extended grey model is called the GM(1,1,exp×sin,exp×cos)model,which is provided with the parameter optimization methods and time response equations.According to the proposed modeling method,we establish a GM(1,1,exp×sin,exp×cos)model for China's quarterly gross domestic product(GDP)with high accuracy.展开更多
An optimized data-matching machine learning algorithm is developed to provide high-prediction accuracy of total burned areas for specific wildfire incidents.It is applied to a well-studied forest-fire dataset from Por...An optimized data-matching machine learning algorithm is developed to provide high-prediction accuracy of total burned areas for specific wildfire incidents.It is applied to a well-studied forest-fire dataset from Portugal Montesinho Natural Park considering 13 input variables.The total burned area distribution of the 517 burn events in that dataset is highly positively skewed.The model is transparent and avoids regressions and hidden layers.This increases its detailed datamining capabilities.It matches the highest burned-area prediction accuracy achieved for this datasetwith a wide range of traditionalmachine learning algorithms.The two-stage prediction process provides informative feature selection that establishes the relative influences of the input variables on burned-area predictions.Optimizing with mean absolute error(MAE)and root mean square error(RMSE)as separate objective functions provides complementary information with which to data mine each total burnedarea incident.Such insight offers potential agricultural,ecological,environmental and forestry benefits by improving the understanding of the key influences associated with each burn event.Data mining the differential trends of cumulative absolute error and squared error also provides detailed insight with which to determine the suitability of each optimized solution to accurately predict burned-areas events of specific types.Such prediction accuracy and insight leads to confidence in how each prediction is derived.It provides knowledge to make appropriate responses and mitigate specific burn incidents,as they occur.Such informed responses should lead to short-term and long-term multi-faceted benefits by helping to prevent certain types of burn incidents being repeated or spread.展开更多
The inverse of expected error variance is utilized to determine weights of individual ensemble members based on the THORPEX(The Observing System Research and Predictability Experiment) Interactive Grand Global Ensembl...The inverse of expected error variance is utilized to determine weights of individual ensemble members based on the THORPEX(The Observing System Research and Predictability Experiment) Interactive Grand Global Ensemble(TIGGE) forecast datasets. The weights of all ensemble members are thus calculated for summer 2012, with the NCEP final operational global analysis(FNL) data as the truth. Based on the weights of all ensemble members, the variable weighted ensemble mean(VWEM) of temperature of summer 2013 is derived and compared with that from the simple equally weighted ensemble mean. The results show that VWEM has lower root-mean-square error(RMSE) as well as absolute error, and has improved the temperature prediction accuracy. The improvements are quite notable over the Tibetan Plateau and its surrounding areas; specifically, a relative improvement rate of RMSE of more than 24% in 2-m temperature is demonstrated. Moreover, the improvement rates vary slightly with the prediction lead-time(24–96 h). It is suggested that the VWEM approach be employed in operational ensemble prediction to provide guidance for weather forecasting and climate prediction.展开更多
Aiming at the problems of low prediction accuracy and weak generalization ability of current reliability prediction models,this paper proposes a hybrid multi-layer heterogeneous particle swarm optimization algorithm(H...Aiming at the problems of low prediction accuracy and weak generalization ability of current reliability prediction models,this paper proposes a hybrid multi-layer heterogeneous particle swarm optimization algorithm(HMHPSO)that can simultaneously optimize the structure and parameters of the GRU neural network.It first introduced a multi-layer heteromass particle swarm optimization(MHPSO)algorithm,which sets the population topology as a hierarchical structure and introduces the concept of attractors,so as to improve the update formula of particle speed,and enhance the information interaction ability between particles,increase the diversity of the groups,thereby improving the optimization ability of the algorithm.Then the HMHPSO used the quantum particle swarm optimization(QPSO)algorithm to determine the structure of the GRU,that is,the number of hidden nodes.Experimental results show that the algorithm can generate GRU neural networks with high generalization performance and low architecture complexity,and has better prediction accuracy in software reliability prediction.展开更多
基金the National Natural Science Foundation of China(No.32273185)the National Key R&D Program of China(No.2020YFD0901205)the Marine Fishery Resources Investigation and Exploration Program of the Ministry of Agriculture and Rural Affairs of China in 2021(No.D-8006-21-0215)。
文摘Albacore tuna(Thunnus alalunga)is one of the target species of tuna longline fishing,and waters near the Cook Islands are a vital albacore tuna fishing ground.Marine environmental data are usually presented with different spatial resolutions,which leads to different results in tuna fishery prediction.Study on the impact of different spatial resolutions on the prediction accuracy of albacore tuna fishery to select the best spatial resolution can contribute to better management of albacore tuna resources.The nominal catch per unit effort(CPUE)of albacore tuna is calculated according to vessel monitor system(VMS)data collected from Chinese distantwater fishery enterprises from January 1,2017 to May 31,2021.A total of 26 spatiotemporal and environmental factors,including temperature,salinity,dissolved oxygen of 0–300 m water layer,chlorophyll-a concentration in the sea surface,sea surface height,month,longitude,and latitude,were selected as variables.The temporal resolution of the variables was daily and the spatial resolutions were set to be 0.5°×0.5°,1°×1°,2°×2°,and 5°×5°.The relationship between the nominal CPUE and each individual factor was analyzed to remove the factors irrelavant to the nominal CPUE,together with a multicollinearity diagnosis on the factors to remove factors highly related to the other factors within the four spatial resolutions.The relationship models between CPUE and spatiotemporal and environmental factors by four spatial resolutions were established based on the long short-term memory(LSTM)neural network model.The mean absolute error(MAE)and root mean square error(RMSE)were used to analyze the fitness and accuracy of the models,and to determine the effects of different spatial resolutions on the prediction accuracy of the albacore tuna fishing ground.The results show the resolution of 1°×1°can lead to the best prediction accuracy,with the MAE and RMSE being 0.0268 and 0.0452 respectively,followed by 0.5°×0.5°,2°×2°and 5°×5°with declining prediction accuracy.The results suggested that 1)albacore tuna fishing ground can be predicted by LSTM;2)the VMS records the data in detail and can be used scientifically to calculate the CPUE;3)correlation analysis,and multicollinearity diagnosis are necessary to improve the prediction accuracy of the model;4)the spatial resolution should be 1°×1°in the forecast of albacore tuna fishing ground in waters near the Cook Islands.
基金supported by the Institute of Information&communications Technology Planning&Evaluation(IITP)grant funded by the Korea government(MSIT)(No.2020‐0‐01441,Artificial Intelligence Convergence Research Center(Chungnam National University))“Regional Innovation Strategy(RIS)”through the National Research Foundation of Korea(NRF)funded by the Ministry of Education(MOE)(2021RIS‐004).
文摘The alpha–beta filter algorithm has been widely researched for various applications,for example,navigation and target tracking systems.To improve the dynamic performance of the alpha–beta filter algorithm,a new prediction learning model is proposed in this study.The proposed model has two main components:(1)the alpha–beta filter algorithm is the main prediction module,and(2)the learning module is a feedforward artificial neural network(FF‐ANN).Furthermore,the model uses two inputs,temperature sensor and humidity sensor data,and a prediction algorithm is used to predict actual sensor readings from noisy sensor readings.Using the novel proposed technique,prediction accuracy is significantly improved while adding the feed‐forward backpropagation neural network,and also reduces the root mean square error(RMSE)and mean absolute error(MAE).We carried out different experiments with different experimental setups.The proposed model performance was evaluated with the traditional alpha–beta filter algorithm and other algorithms such as the Kalman filter.A higher prediction accuracy was achieved,and the MAE and RMSE were 35.1%–38.2%respectively.The final proposed model results show increased performance when compared to traditional methods.
文摘Crop Yield Prediction(CYP)is critical to world food production.Food safety is a top priority for policymakers.They rely on reliable CYP to make import and export decisions that must be fulfilled before launching an agricultural business.Crop Yield(CY)is a complex variable influenced by multiple factors,including genotype,environment,and their interactions.CYP is a significant agrarian issue.However,CYP is the main task due to many composite factors,such as climatic conditions and soil characteristics.Machine Learning(ML)is a powerful tool for supporting CYP decisions,including decision support on which crops to grow in a specific season.Generally,Artificial Neural Networks(ANN)are usually used to predict the behaviour of complex non-linear models.As a result,this research paper attempts to determine the correlations between climatic variables,soil nutrients,and CYwith the available data.InANN,threemethods,Levenberg-Marquardt(LM),Bayesian regularisation(BR),and scaled conjugate gradient(SCG),are used to train the neural network(NN)model and then compared to determine prediction accuracy.The performance measures of the training,as declared above,such as Mean Squared Error(MSE)and correlation coefficient(R),were determined to assess the ANN models that had been built.The experimental study proves that LM training algorithms are better,while BR and SCG have minimal performance.
文摘A hard problem that hinders the movement of waxy crude oil is wax deposition in oil pipelines.To ensure the safe operation of crude oil pipelines,an accurate model must be developed to predict the rate of wax deposition in crude oil pipelines.Aiming at the shortcomings of the ENN prediction model,which easily falls into the local minimum value and weak generalization ability in the implementation process,an optimized ENN prediction model based on the IRSA is proposed.The validity of the new model was confirmed by the accurate prediction of two sets of experimental data on wax deposition in crude oil pipelines.The two groups of crude oil wax deposition rate case prediction results showed that the average absolute percentage errors of IRSA-ENN prediction models is 0.5476% and 0.7831%,respectively.Additionally,it shows a higher prediction accuracy compared to the ENN prediction model.In fact,the new model established by using the IRSA to optimize ENN can optimize the initial weights and thresholds in the prediction process,which can overcome the shortcomings of the ENN prediction model,such as weak generalization ability and tendency to fall into the local minimum value,so that it has the advantages of strong implementation and high prediction accuracy.
基金supported by grants from the National Key Research and Development Project(2019YFE0106800)Modern Agriculture Science and Technology Key Project of Hebei Province(19226376D)China Agriculture Research System of MOF and MARA.
文摘Background:Recently,machine learning(ML)has become attractive in genomic prediction,but its superiority in genomic prediction over conventional(ss)GBLUP methods and the choice of optimal ML methods need to be investigated.Results:In this study,2566 Chinese Yorkshire pigs with reproduction trait records were genotyped with the GenoBaits Porcine SNP 50 K and PorcineSNP50 panels.Four ML methods,including support vector regression(SVR),kernel ridge regression(KRR),random forest(RF)and Adaboost.R2 were implemented.Through 20 replicates of fivefold cross-validation(CV)and one prediction for younger individuals,the utility of ML methods in genomic prediction was explored.In CV,compared with genomic BLUP(GBLUP),single-step GBLUP(ssGBLUP)and the Bayesian method BayesHE,ML methods significantly outperformed these conventional methods.ML methods improved the genomic prediction accuracy of GBLUP,ssGBLUP,and BayesHE by 19.3%,15.0% and 20.8%,respectively.In addition,ML methods yielded smaller mean squared error(MSE)and mean absolute error(MAE)in all scenarios.ssGBLUP yielded an improvement of 3.8% on average in accuracy compared to that of GBLUP,and the accuracy of BayesHE was close to that of GBLUP.In genomic prediction of younger individuals,RF and Adaboost.R2_KRR performed better than GBLUP and BayesHE,while ssGBLUP performed comparably with RF,and ssGBLUP yielded slightly higher accuracy and lower MSE than Adaboost.R2_KRR in the prediction of total number of piglets born,while for number of piglets born alive,Adaboost.R2_KRR performed significantly better than ssGBLUP.Among ML methods,Adaboost.R2_KRR consistently performed well in our study.Our findings also demonstrated that optimal hyperparameters are useful for ML methods.After tuning hyperparameters in CV and in predicting genomic outcomes of younger individuals,the average improvement was 14.3% and 21.8% over those using default hyperparameters,respectively.Conclusion:Our findings demonstrated that ML methods had better overall prediction performance than conventional genomic selection methods,and could be new options for genomic prediction.Among ML methods,Adaboost.R2_KRR consistently performed well in our study,and tuning hyperparameters is necessary for ML methods.The optimal hyperparameters depend on the character of traits,datasets etc.
文摘Data is always a crucial issue of concern especially during its prediction and computation in digital revolution.This paper exactly helps in providing efficient learning mechanism for accurate predictability and reducing redundant data communication.It also discusses the Bayesian analysis that finds the conditional probability of at least two parametric based predictions for the data.The paper presents a method for improving the performance of Bayesian classification using the combination of Kalman Filter and K-means.The method is applied on a small dataset just for establishing the fact that the proposed algorithm can reduce the time for computing the clusters from data.The proposed Bayesian learning probabilistic model is used to check the statistical noise and other inaccuracies using unknown variables.This scenario is being implemented using efficient machine learning algorithm to perpetuate the Bayesian probabilistic approach.It also demonstrates the generative function forKalman-filer based prediction model and its observations.This paper implements the algorithm using open source platform of Python and efficiently integrates all different modules to piece of code via Common Platform Enumeration(CPE)for Python.
文摘Among the advantages of using industrial robots for machining applications instead of machine tools are flexibility, cost effectiveness, and versatility. Due to the kinematics of the articulated robot, the system behaviour is quite different compared with machine tools. Two major questions arise in implementing robots in machining tasks: one is the robot’s stiffness, and the second is the achievable machined part accuracy, which varies mainly due to the huge variety of robot models. This paper proposes error prediction model in the application of industrial robot for machining tasks, based on stiffness and accuracy limits. The research work includes experimental and theoretical parts. Advanced machining and inspection tools were applied, as well as a theoretical model of the robot structure and stiffness based on the form-shaping function approach. The robot machining performances, from the workpiece accuracy point of view were predicted.
基金This research was supported by the National Key R and D Program of China(2016YFC0500901 and 2016YFC0500907)the National Natural Science Foundation of China(Grant Nos.31971466 and 41807525)the One Hundred Person Project of the Chinese Academy of Sciences(Y551821).
文摘Determining an optimal sample size is a key step in designing field surveys,and is particularly important for detecting the spatial pattern of highly variable properties such as soil organic carbon(SOC).Based on 550 soil sampling points in the nearsurface layer(0 to 20 cm)in a representative region of northern China's agro-pastoral ecotone,we studied effects of four interpolation methods such as ordinary kriging(OK),universal kriging(UK),inverse distance weighting(IDW)and radial basis function(RBF)and random subsampling(50,100,200,300,400,and 500)on the prediction accuracy of SOC estimation.When the Shannon's Diversity Index(SHDI)and Shannon's Evenness Index(SHEI)was 2.01 and 0.67,the OK method appeared to be a superior method,which had the smallest root mean square error(RMSE)and the mean error(ME)nearest to zero.On the contrary,the UK method performed poorly for the interpolation of SOC in the present study.The sample size of 200 had the most accurate prediction;50 sampling points produced the worst prediction accuracy.Thus,we used 200 samples to estimate the study area's soil organic carbon density(SOCD)by the OK method.The total SOC storage to a depth of 20 cm in the study area was 117.94 Mt,and its mean SOCD was 2.40 kg/m2.The SOCD kg/(C⋅m2)of different land use types were in the following order:woodland(3.29)>grassland(2.35)>cropland(2.19)>sandy land(1.55).
基金Funding is provided by Taif University Researchers Supporting Project Number(TURSP-2020/10),Taif University,Taif,Saudi Arabia.
文摘Stock market trends forecast is one of the most current topics and a significant research challenge due to its dynamic and unstable nature.The stock data is usually non-stationary,and attributes are non-correlative to each other.Several traditional Stock Technical Indicators(STIs)may incorrectly predict the stockmarket trends.To study the stock market characteristics using STIs and make efficient trading decisions,a robust model is built.This paper aims to build up an Evolutionary Deep Learning Model(EDLM)to identify stock trends’prices by using STIs.The proposed model has implemented the Deep Learning(DL)model to establish the concept of Correlation-Tensor.The analysis of the dataset of three most popular banking organizations obtained from the live stock market based on the National Stock exchange(NSE)-India,a Long Short Term Memory(LSTM)is used.The datasets encompassed the trading days from the 17^(th) of Nov 2008 to the 15^(th) of Nov 2018.This work also conducted exhaustive experiments to study the correlation of various STIs with stock price trends.The model built with an EDLM has shown significant improvements over two benchmark ML models and a deep learning one.The proposed model aids investors in making profitable investment decisions as it presents trend-based forecasting and has achieved a prediction accuracy of 63.59%,56.25%,and 57.95%on the datasets of HDFC,Yes Bank,and SBI,respectively.Results indicate that the proposed EDLA with a combination of STIs can often provide improved results than the other state-of-the-art algorithms.
文摘Background:The suitability and performance of the bankruptcy prediction models is an empirical question.The aim of this paper is to develop a bankruptcy prediction model for Indian manufacturing companies on a sample of 208 companies consisting of an equal number of defaulted and non-defaulted firms.Out of 208 companies,130 are used for estimation sample,and 78 are holdout for model validation.The study reestimates the accounting based models such as Altman EI(Journal of Finance 23:19189-209,1968)Z-Score,Ohlson JA(Journal of Accounting Research 18:109-131,1980)Y-Score and Zmijewski ME(Journal of Accounting Research 22:59-82,1984)X-Score model.The paper compares original and re-estimated models to explore the sensitivity of these models towards the change in time periods and financial conditions.Methods:Multiple Discriminant Analysis(MDA)and Probit techniques are employed in the estimation of Z-Score and X-Score models,whereas Logit technique is employed in the estimation of Y-Score and the newly proposed models.The performance of all the original,re-estimated and new proposed models are assessed by predictive accuracy,significance of parameters,long-range accuracy,secondary sample and Receiver Operating Characteristic(ROC)tests.Results:The major findings of the study reveal that the overall predictive accuracy of all the three models improves on estimation and holdout sample when the coefficients are re-estimated.Amongst the contesting models,the new bankruptcy prediction model outperforms other models.Conclusions:The industry specific model should be developed with the new combinations of financial ratios to predict bankruptcy of the firms in a particular country.The study further suggests the coefficients of the models are sensitive to time periods and financial condition.Hence,researchers should be cautioned while choosing the models for bankruptcy prediction to recalculate the models by looking at the recent data in order to get higher predictive accuracy.
基金This work was supported by Funding of the Nanjing Institute of Technology(No.KE21-451).
文摘To realize data sharing,and to fully use the data value,breaking the data island between institutions to realize data collaboration has become a new sharing mode.This paper proposed a distributed data security sharing scheme based on C/S communication mode,and constructed a federated learning architecture that uses differential privacy technology to protect training parameters.Clients do not need to share local data,and they only need to upload the trained model parameters to achieve data sharing.In the process of training,a distributed parameter update mechanism is introduced.The server is mainly responsible for issuing training commands and parameters,and aggregating the local model parameters uploaded by the clients.The client mainly uses the stochastic gradient descent algorithm for gradient trimming,updates,and transmits the trained model parameters back to the server after differential processing.To test the performance of the scheme,in the application scenario where many medical institutions jointly train the disease detection system,the model is tested from multiple perspectives by taking medical data as an example.From the testing results,we can know that for this specific test dataset,when the parameters are properly configured,the lowest prediction accuracy rate is 90.261%and the highest accuracy rate is up to 94.352.It shows that the performance of the model is good.The results also show that this scheme realizes data sharing while protecting data privacy,completes accurate prediction of diseases,and has a good effect.
文摘In recent years,real-time video streaming has grown in popularity.The growing popularity of the Internet of Things(IoT)and other wireless heterogeneous networks mandates that network resources be carefully apportioned among versatile users in order to achieve the best Quality of Experience(QoE)and performance objectives.Most researchers focused on Forward Error Correction(FEC)techniques when attempting to strike a balance between QoE and performance.However,as network capacity increases,the performance degrades,impacting the live visual experience.Recently,Deep Learning(DL)algorithms have been successfully integrated with FEC to stream videos across multiple heterogeneous networks.But these algorithms need to be changed to make the experience better without sacrificing packet loss and delay time.To address the previous challenge,this paper proposes a novel intelligent algorithm that streams video in multi-home heterogeneous networks based on network-centric characteristics.The proposed framework contains modules such as Intelligent Content Extraction Module(ICEM),Channel Status Monitor(CSM),and Adaptive FEC(AFEC).This framework adopts the Cognitive Learning-based Scheduling(CLS)Module,which works on the deep Reinforced Gated Recurrent Networks(RGRN)principle and embeds them along with the FEC to achieve better performances.The complete framework was developed using the Objective Modular Network Testbed in C++(OMNET++),Internet networking(INET),and Python 3.10,with Keras as the front end and Tensorflow 2.10 as the back end.With extensive experimentation,the proposed model outperforms the other existing intelligentmodels in terms of improving the QoE,minimizing the End-to-End Delay(EED),and maintaining the highest accuracy(98%)and a lower Root Mean Square Error(RMSE)value of 0.001.
文摘The social engineering cyber-attack is where culprits mislead the users by getting the login details which provides the information to the evil server called phishing.The deep learning approaches and the machine learning are compared in the proposed system for presenting the methodology that can detect phishing websites via Uniform Resource Locator(URLs)analysis.The legal class is composed of the home pages with no inclusion of login forms in most of the present modern solutions,which deals with the detection of phishing.Contrarily,the URLs in both classes from the login page due,considering the representation of a real case scenario and the demonstration for obtaining the rate of false-positive with the existing approaches during the legal login pages provides the test having URLs.In addition,some model reduces the accuracy rather than training the base model and testing the latest URLs.In addition,a feature analysis is performed on the present phishing domains to identify various approaches to using the phishers in the campaign.A new dataset called the MUPD dataset is used for evaluation.Lastly,a prediction model,the Dense forward-backwards Long Short Term Memory(LSTM)model(d−FBLSTM),is presented for combining the forward and backward propagation of LSMT to obtain the accuracy of 98.5%on the initiated login URL dataset.
文摘The prediction of mild cognitive impairment or Alzheimer’s disease(AD)has gained the attention of huge researchers as the disease occurrence is increasing,and there is a need for earlier prediction.Regrettably,due to the highdimensionality nature of neural data and the least available samples,modelling an efficient computer diagnostic system is highly solicited.Learning approaches,specifically deep learning approaches,are essential in disease prediction.Deep Learning(DL)approaches are successfully demonstrated for their higher-level performance in various fields like medical imaging.A novel 3D-Convolutional Neural Network(3D-CNN)architecture is proposed to predict AD with Magnetic resonance imaging(MRI)data.The proposed model predicts the AD occurrence while the existing approaches lack prediction accuracy and perform binary classification.The proposed prediction model is validated using the Alzheimer’s disease Neuro-Imaging Initiative(ADNI)data.The outcomes demonstrate that the anticipated model attains superior prediction accuracy and works better than the brain-image dataset’s general approaches.The predicted model reduces the human effort during the prediction process and makes it easier to diagnose it intelligently as the feature learning is adaptive.Keras’experimentation is carried out,and the model’s superiority is compared with various advanced approaches for multi-level classification.The proposed model gives better prediction accuracy,precision,recall,and F-measure than other systems like Long Short Term Memory-Recurrent Neural Networks(LSTM-RNN),Stacked Autoencoder with Deep Neural Networks(SAE-DNN),Deep Convolutional Neural Networks(D-CNN),Two Dimensional Convolutional Neural Networks(2D-CNN),Inception-V4,ResNet,and Two Dimensional Convolutional Neural Networks(3D-CNN).
基金Forest and Wood Products Australia Limited(FWPA)through project PNC465-1718:Advanced real-time measurements at harvest to increase value recovery and also supported by Beijing Forestry University through the special fund for characteristic development under the program of Building World-class University and Disciplines.
文摘A new model for predicting the total tree height for harvested stems from cut-to-length(CTL)harvester data was constructed for Pinus radiata(D.Don)following a conceptual analysis of relative stem profi les,comparisons of candidate models forms and extensive selections of predictor variables.Stem profi les of more than 3000 trees in a taper data set were each processed 6 times through simulated log cutting to generate the data required for this purpose.The CTL simulations not only mimicked but also covered the full range of cutting patterns of nearly 0.45×106 stems harvested during both thinning and harvesting operations.The single-equation model was estimated through the multipleequation generalized method of moments estimator to obtain effi cient and consistent parameter estimates in the presence of error correlation and heteroscedasticity that were inherent to the systematic structure of the data.The predictive performances of our new model in its linear and nonlinear form were evaluated through a leave-one-tree-out cross validation process and compared against that of the only such existing model.The evaluations and comparisons were made through benchmarking statistics both globally over the entire data space and locally within specifi c subdivisions of the data space.These statistics indicated that the nonlinear form of our model was the best and its linear form ranked second.The prediction accuracy of our nonlinear model improved when the total log length represented more than 20%of the total tree height.The poorer performance of the existing model was partly attributed to the high degree of multicollinearity among its predictor variables,which led to highly variable and unstable parameter estimates.Our new model will facilitate and widen the utilization of harvester data far beyond the current limited use for monitoring and reporting log productions in P.radiata plantations.It will also facilitate the estimation of bark thickness and help make harvester data a potential source of taper data to reduce the intensity and cost of the conventional destructive taper sampling in the fi eld.Although developed for P.radiata,the mathematical form of our new model will be applicable to other tree species for which CTL harvester data are routinely captured during thinning and harvesting operations.
基金National Natural Science Foundation of China(4157508241530531+1 种基金41605048)Special Scientific Research Project for Public Interest(GYHY201306021)
文摘In order to study the spatial structure and dynamical mechanism of extreme precipitation in East Asia, a corresponding climate network is constructed by employing the method of event synchronization. It is found that the area of East Asian summer extreme precipitation can be separated into two regions: one with high area-weighted connectivity receiving heavy precipitation mostly during the active phase of the East Asian Summer Monsoon(EASM),and another one with low area-weighted connectivity receiving heavy precipitation during both the active and the retreating phase of the EASM. Besides, a new way for the prediction of extreme precipitation is also developed by constructing a directed climate networks. The simulation accuracy in East Asia is 58% with a 0-day lead, and the prediction accuracy is 21% and average 12% with a 1-day and an n-day(2≤n≤10) lead, respectively. Compared to the normal EASM year, the prediction accuracy is low in weak years and high in strong years, which is relevant to the differences of correlations and extreme precipitation rates in different EASM situations. Recognizing and indentifying these effects is good for understanding and predicting extreme precipitation in East Asia.
基金Supported by National Natural Science Foundation of China (11401418)。
文摘Grey prediction is vital in statistical prediction with wide applications.However,most grey prediction methods focus on annual predictions of the monotonic time series instead of the seasonal time series.The paper uses the extended model of the grey GM(1,1)model to predict the seasonal time series.Some improvements have been made in two aspects to improve the prediction accuracy of the model.1)We introduce seasonal multiple factors to transform the original time series,which improves the adaptability of the seasonal data to the model.The transformed series conforms to the law presented by the model.2)The seasonal data are in superimposed sine and cosine fluctuations with tendencies.Therefore,the paper extends the grey action quantity of the traditional GM(1,1)model.The newly extended grey model is called the GM(1,1,exp×sin,exp×cos)model,which is provided with the parameter optimization methods and time response equations.According to the proposed modeling method,we establish a GM(1,1,exp×sin,exp×cos)model for China's quarterly gross domestic product(GDP)with high accuracy.
文摘An optimized data-matching machine learning algorithm is developed to provide high-prediction accuracy of total burned areas for specific wildfire incidents.It is applied to a well-studied forest-fire dataset from Portugal Montesinho Natural Park considering 13 input variables.The total burned area distribution of the 517 burn events in that dataset is highly positively skewed.The model is transparent and avoids regressions and hidden layers.This increases its detailed datamining capabilities.It matches the highest burned-area prediction accuracy achieved for this datasetwith a wide range of traditionalmachine learning algorithms.The two-stage prediction process provides informative feature selection that establishes the relative influences of the input variables on burned-area predictions.Optimizing with mean absolute error(MAE)and root mean square error(RMSE)as separate objective functions provides complementary information with which to data mine each total burnedarea incident.Such insight offers potential agricultural,ecological,environmental and forestry benefits by improving the understanding of the key influences associated with each burn event.Data mining the differential trends of cumulative absolute error and squared error also provides detailed insight with which to determine the suitability of each optimized solution to accurately predict burned-areas events of specific types.Such prediction accuracy and insight leads to confidence in how each prediction is derived.It provides knowledge to make appropriate responses and mitigate specific burn incidents,as they occur.Such informed responses should lead to short-term and long-term multi-faceted benefits by helping to prevent certain types of burn incidents being repeated or spread.
基金Supported by the National Natural Science Foundation of China(41405006 and 91224004)Meteorological Key Technology Integration and Application Program(CMAGJ2015M85)+2 种基金National Key Technology Research and Development Program(2015BAK10B03)China Meteorological Administration Special Public Welfare Research Fund(GYHY201506002)Basic Research Fund of the Chinese Academy of Meteorological Sciences(2014R016 and 2015Z003)
文摘The inverse of expected error variance is utilized to determine weights of individual ensemble members based on the THORPEX(The Observing System Research and Predictability Experiment) Interactive Grand Global Ensemble(TIGGE) forecast datasets. The weights of all ensemble members are thus calculated for summer 2012, with the NCEP final operational global analysis(FNL) data as the truth. Based on the weights of all ensemble members, the variable weighted ensemble mean(VWEM) of temperature of summer 2013 is derived and compared with that from the simple equally weighted ensemble mean. The results show that VWEM has lower root-mean-square error(RMSE) as well as absolute error, and has improved the temperature prediction accuracy. The improvements are quite notable over the Tibetan Plateau and its surrounding areas; specifically, a relative improvement rate of RMSE of more than 24% in 2-m temperature is demonstrated. Moreover, the improvement rates vary slightly with the prediction lead-time(24–96 h). It is suggested that the VWEM approach be employed in operational ensemble prediction to provide guidance for weather forecasting and climate prediction.
文摘Aiming at the problems of low prediction accuracy and weak generalization ability of current reliability prediction models,this paper proposes a hybrid multi-layer heterogeneous particle swarm optimization algorithm(HMHPSO)that can simultaneously optimize the structure and parameters of the GRU neural network.It first introduced a multi-layer heteromass particle swarm optimization(MHPSO)algorithm,which sets the population topology as a hierarchical structure and introduces the concept of attractors,so as to improve the update formula of particle speed,and enhance the information interaction ability between particles,increase the diversity of the groups,thereby improving the optimization ability of the algorithm.Then the HMHPSO used the quantum particle swarm optimization(QPSO)algorithm to determine the structure of the GRU,that is,the number of hidden nodes.Experimental results show that the algorithm can generate GRU neural networks with high generalization performance and low architecture complexity,and has better prediction accuracy in software reliability prediction.
