Network autoregression and factor model are effective methods for modeling network time series data.In this study,we propose a network autoregression model with a factor structure that incorporates a latent group stru...Network autoregression and factor model are effective methods for modeling network time series data.In this study,we propose a network autoregression model with a factor structure that incorporates a latent group structure to address nodal heterogeneity within the network.An iterative algorithm is employed to minimize a least-squares objective function,allowing for simultaneous estimation of both the parameters and the group structure.To determine the unknown number of groups and factors,a PIC criterion is introduced.Additionally,statistical inference of the estimated parameters is presented.To assess the validity of the proposed estimation and inference procedures,we conduct extensive numerical studies.We also demonstrate the utility of our model using a stock dataset obtained from the Chinese A-Share stock market.展开更多
Forecasting solar irradiance is a critical task in the renewable energy sector, as it provides essential information regarding the potential energy production from solar panels. This study aims to utilize the Vector A...Forecasting solar irradiance is a critical task in the renewable energy sector, as it provides essential information regarding the potential energy production from solar panels. This study aims to utilize the Vector Autoregression (VAR) model to forecast solar irradiance levels and weather characteristics in the San Francisco Bay Area. The results demonstrate a correlation between predicted and actual solar irradiance, indicating the effectiveness of the VAR model for this task. However, the model may not be sufficient for this region due to the requirement of additional weather features to reduce disparities between predictions and actual observations. Additionally, the current lag order in the model is relatively low, limiting its ability to capture all relevant information from past observations. As a result, the model’s forecasting capability is limited to short-term horizons, with a maximum horizon of four hours.展开更多
-In this paper, monthly mean SST data in a large area are used. After the spacial average of the data is carried out and the secular monthly means are substracted, a time series (Jan. 1951-Dec. 1985) of SST anomalies ...-In this paper, monthly mean SST data in a large area are used. After the spacial average of the data is carried out and the secular monthly means are substracted, a time series (Jan. 1951-Dec. 1985) of SST anomalies of the cold tongue water area in the eastern tropical Pacific Ocean is obtained. On the basis of the time series, an autoregression model, a self-exciting threshold autoregression model and an open loop autoregression model are developed respectively. The interannual variations are simulated by means of those models. The simulation results show that all the three models have made very good hindcasting for the nine El Nino events since 1951. In order to test the reliability of the open loop threshold model, extrapolated forecast was made for the period of Jan. 1986-Feb. 1987. It can be seen from the forecasting that the model could forecast well the beginning and strengthening stages of the recent El Nino event (1986-1987). Correlation coefficients of the estimations to observations are respectively 0. 84, 0. 88 and 0. 89. It is obvious that all the models work well and the open loop threshold one is the best. So the open loop threshold autoregression model is a useful tool for monitoring the SSTinterannual variation of the cold tongue water area in the Eastern Equatorial Pacific Ocean and for estimating the El Nino strength.展开更多
Weather forecasting is crucial to both the demand and supply sides of electricity systems. Temperature has a great effect on the demand side. Moreover, solar and wind are very promising renewable energy sources and ar...Weather forecasting is crucial to both the demand and supply sides of electricity systems. Temperature has a great effect on the demand side. Moreover, solar and wind are very promising renewable energy sources and are, thus, important on the supply side. In this paper, a large vector autoregression(VAR) model is built to forecast three important weather variables for 61 cities around the United States. The three variables at all locations are modeled as response variables. Lag terms are used to capture the relationship between observations in adjacent periods and daily and annual seasonality are modeled to consider the correlation between the same periods in adjacent days and years. We estimate the VAR model with16 years of hourly historical data and use two additional years of data for out-of-sample validation. Forecasts of up to six-hours-ahead are generated with good forecasting performance based on mean absolute error, root mean square error, relative root mean square error, and skill scores. Our VAR model gives forecasts with skill scoresthat are more than double the skill scores of other forecasting models in the literature. Our model also provides forecasts that outperform persistence forecasts by between6% and 80% in terms of mean absolute error. Our results show that the proposed time series approach is appropriate for very short-term forecasting of hourly solar radiation,temperature, and wind speed.展开更多
This study proposes two metrics using the nearest neighbors method to improve the accuracy of time-series forecasting. These two metrics can be treated as a hybrid forecasting approach to combine linear and non-linear...This study proposes two metrics using the nearest neighbors method to improve the accuracy of time-series forecasting. These two metrics can be treated as a hybrid forecasting approach to combine linear and non-linear forecasting techniques. One metric redefines the distance in k-nearest neighbors based on the coefficients of autoregression (AR) in time series. Meanwhile, an improvement to Kulesh's adaptive metrics in the nearest neighbors is also presented. To evaluate the performance of the two proposed metrics, three types of time-series data, namely deterministic synthetic data, chaotic time-series data and real time-series data, are predicted. Experimental results show the superiority of the proposed AR-enhanced k-nearest neighbors methods to the traditional k-nearest neighbors metric and Kulesh's adaptive metrics.展开更多
In this paper,we consider the limit distribution of the error density function estima-tor in the rst-order autoregressive models with negatively associated and positively associated random errors.Under mild regularity...In this paper,we consider the limit distribution of the error density function estima-tor in the rst-order autoregressive models with negatively associated and positively associated random errors.Under mild regularity assumptions,some asymptotic normality results of the residual density estimator are obtained when the autoregressive models are stationary process and explosive process.In order to illustrate these results,some simulations such as con dence intervals and mean integrated square errors are provided in this paper.It shows that the residual density estimator can replace the density\estimator"which contains errors.展开更多
Long-term time series forecasting stands as a crucial research domain within the realm of automated machine learning(AutoML).At present,forecasting,whether rooted in machine learning or statistical learning,typically ...Long-term time series forecasting stands as a crucial research domain within the realm of automated machine learning(AutoML).At present,forecasting,whether rooted in machine learning or statistical learning,typically relies on expert input and necessitates substantial manual involvement.This manual effort spans model development,feature engineering,hyper-parameter tuning,and the intricate construction of time series models.The complexity of these tasks renders complete automation unfeasible,as they inherently demand human intervention at multiple junctures.To surmount these challenges,this article proposes leveraging Long Short-Term Memory,which is the variant of Recurrent Neural Networks,harnessing memory cells and gating mechanisms to facilitate long-term time series prediction.However,forecasting accuracy by particular neural network and traditional models can degrade significantly,when addressing long-term time-series tasks.Therefore,our research demonstrates that this innovative approach outperforms the traditional Autoregressive Integrated Moving Average(ARIMA)method in forecasting long-term univariate time series.ARIMA is a high-quality and competitive model in time series prediction,and yet it requires significant preprocessing efforts.Using multiple accuracy metrics,we have evaluated both ARIMA and proposed method on the simulated time-series data and real data in both short and long term.Furthermore,our findings indicate its superiority over alternative network architectures,including Fully Connected Neural Networks,Convolutional Neural Networks,and Nonpooling Convolutional Neural Networks.Our AutoML approach enables non-professional to attain highly accurate and effective time series forecasting,and can be widely applied to various domains,particularly in business and finance.展开更多
China has resolved its overall regional poverty in 2020 by attaining moderate societal prosperity.The country has entered a new development stage designed to achieve its second centenary goal.However,ecological fragil...China has resolved its overall regional poverty in 2020 by attaining moderate societal prosperity.The country has entered a new development stage designed to achieve its second centenary goal.However,ecological fragility and risk susceptibility have increased the risk of returning to ecological poverty.In this paper,the Liupan Mountain Region of China was used as a case study,and the counties were used as the scale to reveal the spatiotempora differentiation and influcing factors of the risk of returning to poverty in study area.The indicator data for returning to ecological poverty from 2011-2020 were collected and summarized in three dimensions:ecological,economic and social.The autoregressive integrated moving average model(ARIMA)time series and exponential smoothing method(ES)were used to predict the multidimensional indicators of returning to ecological poverty for 61 counties(districts)in the Liupan Mountain Region for 2021-2030.The back propagation neural network(BPNN)and geographic information system(GIS)were used to generate the spatial distribution and time variation for the index of the risk of returning to ecological poverty(RREP index).The results show that 1)ecological factors were the main factors in the risk of returning to ecological poverty in Liupan Mountain Region.2)The RREP index for the 61 counties(districts)exhibited a downward trend from 2021-2030.The RREP index declined more in medium-and high-risk areas than in low-risk areas.From 2021 to 2025,the RREP index exhibited a slight downward trend.From 2026 to2030,the RREP index was expected to decline faster,especially from 2029-2030.3)Based on the RREP index,it can be roughly divided into three types,namely,the high-risk areas,the medium-risk areas,and the low-risk areas.The natural resource conditions in lowrisk areas of returning to ecological poverty,were better than those in medium-and high-risk areas.展开更多
Sentence classification is the process of categorizing a sentence based on the context of the sentence.Sentence categorization requires more semantic highlights than other tasks,such as dependence parsing,which requir...Sentence classification is the process of categorizing a sentence based on the context of the sentence.Sentence categorization requires more semantic highlights than other tasks,such as dependence parsing,which requires more syntactic elements.Most existing strategies focus on the general semantics of a conversation without involving the context of the sentence,recognizing the progress and comparing impacts.An ensemble pre-trained language model was taken up here to classify the conversation sentences from the conversation corpus.The conversational sentences are classified into four categories:information,question,directive,and commission.These classification label sequences are for analyzing the conversation progress and predicting the pecking order of the conversation.Ensemble of Bidirectional Encoder for Representation of Transformer(BERT),Robustly Optimized BERT pretraining Approach(RoBERTa),Generative Pre-Trained Transformer(GPT),DistilBERT and Generalized Autoregressive Pretraining for Language Understanding(XLNet)models are trained on conversation corpus with hyperparameters.Hyperparameter tuning approach is carried out for better performance on sentence classification.This Ensemble of Pre-trained Language Models with a Hyperparameter Tuning(EPLM-HT)system is trained on an annotated conversation dataset.The proposed approach outperformed compared to the base BERT,GPT,DistilBERT and XLNet transformer models.The proposed ensemble model with the fine-tuned parameters achieved an F1_score of 0.88.展开更多
This paper studies the autoregression models of order one, in a general time series setting that allows for weakly dependent innovations. Let {Xt} be a linear process defined by Xt =∑k=0^∞ψ kεt-k, where {ψk, k ≥...This paper studies the autoregression models of order one, in a general time series setting that allows for weakly dependent innovations. Let {Xt} be a linear process defined by Xt =∑k=0^∞ψ kεt-k, where {ψk, k ≥ 0} is a sequence of real numbers and {εk, k = 0, ±1, ±2,...} is a sequence of random variables. Two results are proved in this paper. In the first result, assuming that {εk, k ≥ 1} is a sequence of asymptotically linear negative quadrant dependent （ALNQD） random variables, the authors find the limiting distributions of the least squares estimator and the associated regression t statistic. It is interesting that the limiting distributions are similar to the one found in earlier work under the assumption of i.i.d, innovations. In the second result the authors prove that the least squares estimator is not a strong consistency estimator of the autoregressive parameter a when {εk, k ≥ 1} is a sequence of negatively associated （NA） random variables, and ψ0 = 1, ψk = 0, k ≥ 1.展开更多
China＇s astonishing economic growth implies a necessity to understand its inflation. The present paper employs threshold nonrecursive structural vector autoregression analysis to explore the asymmetric effects of mac...China＇s astonishing economic growth implies a necessity to understand its inflation. The present paper employs threshold nonrecursive structural vector autoregression analysis to explore the asymmetric effects of macro-variables on inflation in low and high inflation regimes. The empirical evidence demonstrates, first, that the reactions of inflation to various shocks are inflation-regime-dependent and asymmetric. Second, monetary policy influences China ＂s high inflation and adjusting the domestic interest rate in China may be an effective way to control inflation in a high inflation regime, but not in a low inflation regime. In a high inflation regime, a high inflation rate may cause the macro-policy authorities to increase the domestic interest rate, in an attempt to stabilize high inflation. Third, contrary to expectations, the world oil price is not a strong cost-push factor in a low inflation regime. Oil price increases may increase inflation in a high inflation regime, but there is no such obvious effect in a low inflation regime. Finally, China ＂s nominal effective exchange rate influences inflation in both low and high inflation regimes. A nominal effeetive exchange rate appreciation might be effective in controlling domestic inflation in both regimes.展开更多
Through extension of canonical correlation to the analysis of meteorological element fields (MEF), a concept from combination of canonical autocorrelation with canonical autoregression (CAR) is developed for short-ter...Through extension of canonical correlation to the analysis of meteorological element fields (MEF), a concept from combination of canonical autocorrelation with canonical autoregression (CAR) is developed for short-term climatic prediction of MEFs with a formulated scheme. Experi- mental results suggest that the scheme is of encouraging usefulness to a weak persistence MEF, i.e., rainfall field and, in particular, to a strong persistance one like a SST field.展开更多
Since 2002,the People's Bank of China has frequently used both quantity-based direct monetary instruments and price-based indirect monetary instruments to promote economic growth and stabilize price level.Specific...Since 2002,the People's Bank of China has frequently used both quantity-based direct monetary instruments and price-based indirect monetary instruments to promote economic growth and stabilize price level.Specifically,this study estimates 13 three-variable factor-augmented vector autoregression (FAVAR) models to explore how two types of monetary instruments affect China's economy and price level.Overall,we find that monetary policy has positive effects on China's economy and price level.Second,this study clearly states that the effectiveness of China's monetary policy on the economy has depended on China's quantity-based direct monetary instruments since 2002.Third,the effectiveness of quantity-based direct monetary instruments on China's economy and price level is dependent on the significant and positive effects of quantity-based direct monetary instruments after the 2008 financial crisis.Fourth,the significant and positive effects of price-based indirect monetary instruments on China's economy and price level before 2008 cannot fundamentally change their current insignificant effects on China's economy and price level.展开更多
The stock market is a vital component of the broader financial system,with its dynamics closely linked to economic growth.The challenges associated with analyzing and forecasting stock prices have persisted since the ...The stock market is a vital component of the broader financial system,with its dynamics closely linked to economic growth.The challenges associated with analyzing and forecasting stock prices have persisted since the inception of financial markets.By examining historical transaction data,latent opportunities for profit can be uncovered,providing valuable insights for both institutional and individual investors to make more informed decisions.This study focuses on analyzing historical transaction data from four banks to predict closing price trends.Various models,including decision trees,random forests,and Long Short-Term Memory(LSTM)networks,are employed to forecast stock price movements.Historical stock transaction data serves as the input for training these models,which are then used to predict upward or downward stock price trends.The study’s empirical results indicate that these methods are effective to a degree in predicting stock price movements.The LSTM-based deep neural network model,in particular,demonstrates a commendable level of predictive accuracy.This conclusion is reached following a thorough evaluation of model performance,highlighting the potential of LSTM models in stock market forecasting.The findings offer significant implications for advancing financial forecasting approaches,thereby improving the decision-making capabilities of investors and financial institutions.展开更多
In the parameter tracking of time-varying systems, the ordinary method is weighted least squares with the rectangular window or the exponential window. In this paper we propose a new kind of sliding window called the ...In the parameter tracking of time-varying systems, the ordinary method is weighted least squares with the rectangular window or the exponential window. In this paper we propose a new kind of sliding window called the multiple exponential window, and then use it to fit time-varying Gaussian vector autoregressive models. The asymptotic bias and covariance of the estimator of the parameter for time-invariant models are also derived. Simulation results show that the multiple exponential windows have better parameter tracking effect than rectangular windows and exponential ones.展开更多
Although the popular multi-fidelity surrogate models,stochastic collocation and nonlinear autoregression have been applied successfully to multiple benchmark problems in different areas of science and engineering,they...Although the popular multi-fidelity surrogate models,stochastic collocation and nonlinear autoregression have been applied successfully to multiple benchmark problems in different areas of science and engineering,they have certain limitations.We propose a uniform Bayesian framework that connects these two methods allowing us to combine the strengths of both.To this end,we introduce Greedy-NAR,a nonlinear Bayesian autoregressive model that can handle complex between-fidelity correlations and involves a sequential construction that allows for significant improvements in performance given a limited computational budget.The proposed enhanced nonlinear autoregressive method is applied to three benchmark problems that are typical of energy applications,namely molecular dynamics and computational fluid dynamics.The results indicate an increase in both prediction stability and accuracy when compared to those of the standard multi-fidelity autoregression implementations.The results also reveal the advantages over the stochastic collocation approach in terms of accuracy and computational cost.Generally speaking,the proposed enhancement provides a straightforward and easily implemented approach for boosting the accuracy and efficiency of concatenated structure multi-fidelity simulation methods,e.g.,the nonlinear autoregressive model,with a negligible additional computational cost.展开更多
BACKGROUND Hepatitis B(HB)and hepatitis C(HC)place the largest burden in China,and a goal of eliminating them as a major public health threat by 2030 has been set.Making more informed and accurate forecasts of their s...BACKGROUND Hepatitis B(HB)and hepatitis C(HC)place the largest burden in China,and a goal of eliminating them as a major public health threat by 2030 has been set.Making more informed and accurate forecasts of their spread is essential for developing effective strategies,heightening the requirement for early warning to deal with such a major public health threat.AIM To monitor HB and HC epidemics by the design of a paradigmatic seasonal autoregressive fractionally integrated moving average(SARFIMA)for projections into 2030,and to compare the effectiveness with the seasonal autoregressive integrated moving average(SARIMA).METHODS Monthly HB and HC incidence cases in China were obtained from January 2004 to June 2023.Descriptive analysis and the Hodrick-Prescott method were employed to identify trends and seasonality.Two periods(from January 2004 to June 2022 and from January 2004 to December 2015,respectively)were used as the training sets to develop both models,while the remaining periods served as the test sets to evaluate the forecasting accuracy.RESULTS There were incidents of 23400874 HB cases and 3590867 HC cases from January 2004 to June 2023.Overall,HB remained steady[average annual percentage change(AAPC)=0.44,95%confidence interval(95%CI):-0.94-1.84]while HC was increasing(AAPC=8.91,95%CI:6.98-10.88),and both had a peak in March and a trough in February.In the 12-step-ahead HB forecast,the mean absolute deviation(15211.94),root mean square error(18762.94),mean absolute percentage error(0.17),mean error rate(0.15),and root mean square percentage error(0.25)under the best SARFIMA(3,0,0)(0,0.449,2)12 were smaller than those under the best SARIMA(3,0,0)(0,1,2)12(16867.71,20775.12,0.19,0.17,and 0.27,respectively).Similar results were also observed for the 90-step-ahead HB,12-step-ahead HC,and 90-step-ahead HC forecasts.The predicted HB incidents totaled 9865400(95%CI:7508093-12222709)cases and HC totaled 1659485(95%CI:856681-2462290)cases during 2023-2030.CONCLUSION Under current interventions,China faces enormous challenges to eliminate HB and HC epidemics by 2030,and effective strategies must be reinforced.The integration of SARFIMA into public health for the management of HB and HC epidemics can potentially result in more informed and efficient interventions,surpassing the capabilities of SARIMA.展开更多
The study aims to investigate the financial technology(FinTech)factors influencing Chinese banking performance.Financial expectations and global realities may be changed by FinTech’s multidimensional scope,which is l...The study aims to investigate the financial technology(FinTech)factors influencing Chinese banking performance.Financial expectations and global realities may be changed by FinTech’s multidimensional scope,which is lacking in the traditional financial sector.The use of technology to automate financial services is becoming more important for economic organizations and industries because the digital age has seen a period of transition in terms of consumers and personalization.The future of FinTech will be shaped by technologies like the Internet of Things,blockchain,and artificial intelligence.The involvement of these platforms in financial services is a major concern for global business growth.FinTech is becoming more popular with customers because of such benefits.FinTech has driven a fundamental change within the financial services industry,placing the client at the center of everything.Protection has become a primary focus since data are a component of FinTech transactions.The task of consolidating research reports for consensus is very manual,as there is no standardized format.Although existing research has proposed certain methods,they have certain drawbacks in FinTech payment systems(including cryptocurrencies),credit markets(including peer-to-peer lending),and insurance systems.This paper implements blockchainbased financial technology for the banking sector to overcome these transition issues.In this study,we have proposed an adaptive neuro-fuzzy-based K-nearest neighbors’algorithm.The chaotic improved foraging optimization algorithm is used to optimize the proposed method.The rolling window autoregressive lag modeling approach analyzes FinTech growth.The proposed algorithm is compared with existing approaches to demonstrate its efficiency.The findings showed that it achieved 91%accuracy,90%privacy,96%robustness,and 25%cyber-risk performance.Compared with traditional approaches,the recommended strategy will be more convenient,safe,and effective in the transition period.展开更多
Construction Industry operates relying on various key economic indicators.One of these indicators is material prices.On the other hand,cost is a key concern in all operations of the construction industry.In the uncert...Construction Industry operates relying on various key economic indicators.One of these indicators is material prices.On the other hand,cost is a key concern in all operations of the construction industry.In the uncertain conditions,reliable cost forecasts become an important source of information.Material cost is one of the key components of the overall cost of construction.In addition,cost overrun is a common problem in the construction industry,where nine out of ten construction projects face cost overrun.In order to carry out a successful cost management strategy and prevent cost overruns,it is very important to find reliable methods for the estimation of construction material prices.Material prices have a time dependent nature.In order to increase the foreseeability of the costs of construction materials,this study focuses on estimation of construction material indices through time series analysis.Two different types of analysis are implemented for estimation of the future values of construction material indices.The first method implemented was Autoregressive Integrated Moving Average(ARIMA),which is known to be successful in estimation of time series having a linear nature.The second method implementedwas Non-LinearAutoregressive Neural Network(NARNET)which is known to be successful in modeling and estimating of series with non-linear components.The results have shown that depending on the nature of the series,both these methods can successfully and accurately estimate the future values of the indices.In addition,we found out that Optimal NARNET architectures which provide better accuracy in estimation of the series can be identified/discovered as result of grid search on NARNET hyperparameters.展开更多
基金Supported by National Natural Science Foundation of China(72222009,71991472)。
文摘Network autoregression and factor model are effective methods for modeling network time series data.In this study,we propose a network autoregression model with a factor structure that incorporates a latent group structure to address nodal heterogeneity within the network.An iterative algorithm is employed to minimize a least-squares objective function,allowing for simultaneous estimation of both the parameters and the group structure.To determine the unknown number of groups and factors,a PIC criterion is introduced.Additionally,statistical inference of the estimated parameters is presented.To assess the validity of the proposed estimation and inference procedures,we conduct extensive numerical studies.We also demonstrate the utility of our model using a stock dataset obtained from the Chinese A-Share stock market.
文摘Forecasting solar irradiance is a critical task in the renewable energy sector, as it provides essential information regarding the potential energy production from solar panels. This study aims to utilize the Vector Autoregression (VAR) model to forecast solar irradiance levels and weather characteristics in the San Francisco Bay Area. The results demonstrate a correlation between predicted and actual solar irradiance, indicating the effectiveness of the VAR model for this task. However, the model may not be sufficient for this region due to the requirement of additional weather features to reduce disparities between predictions and actual observations. Additionally, the current lag order in the model is relatively low, limiting its ability to capture all relevant information from past observations. As a result, the model’s forecasting capability is limited to short-term horizons, with a maximum horizon of four hours.
文摘-In this paper, monthly mean SST data in a large area are used. After the spacial average of the data is carried out and the secular monthly means are substracted, a time series (Jan. 1951-Dec. 1985) of SST anomalies of the cold tongue water area in the eastern tropical Pacific Ocean is obtained. On the basis of the time series, an autoregression model, a self-exciting threshold autoregression model and an open loop autoregression model are developed respectively. The interannual variations are simulated by means of those models. The simulation results show that all the three models have made very good hindcasting for the nine El Nino events since 1951. In order to test the reliability of the open loop threshold model, extrapolated forecast was made for the period of Jan. 1986-Feb. 1987. It can be seen from the forecasting that the model could forecast well the beginning and strengthening stages of the recent El Nino event (1986-1987). Correlation coefficients of the estimations to observations are respectively 0. 84, 0. 88 and 0. 89. It is obvious that all the models work well and the open loop threshold one is the best. So the open loop threshold autoregression model is a useful tool for monitoring the SSTinterannual variation of the cold tongue water area in the Eastern Equatorial Pacific Ocean and for estimating the El Nino strength.
基金supported by the National Science Foundation (No: 1029337)supported by an allocation of computing time from the Ohio Supercomputer Center
文摘Weather forecasting is crucial to both the demand and supply sides of electricity systems. Temperature has a great effect on the demand side. Moreover, solar and wind are very promising renewable energy sources and are, thus, important on the supply side. In this paper, a large vector autoregression(VAR) model is built to forecast three important weather variables for 61 cities around the United States. The three variables at all locations are modeled as response variables. Lag terms are used to capture the relationship between observations in adjacent periods and daily and annual seasonality are modeled to consider the correlation between the same periods in adjacent days and years. We estimate the VAR model with16 years of hourly historical data and use two additional years of data for out-of-sample validation. Forecasts of up to six-hours-ahead are generated with good forecasting performance based on mean absolute error, root mean square error, relative root mean square error, and skill scores. Our VAR model gives forecasts with skill scoresthat are more than double the skill scores of other forecasting models in the literature. Our model also provides forecasts that outperform persistence forecasts by between6% and 80% in terms of mean absolute error. Our results show that the proposed time series approach is appropriate for very short-term forecasting of hourly solar radiation,temperature, and wind speed.
基金the National Natural Science Foundation of China(No.61203337)the Natural Science Foundation of Shanghai(No.12ZR1440200)
文摘This study proposes two metrics using the nearest neighbors method to improve the accuracy of time-series forecasting. These two metrics can be treated as a hybrid forecasting approach to combine linear and non-linear forecasting techniques. One metric redefines the distance in k-nearest neighbors based on the coefficients of autoregression (AR) in time series. Meanwhile, an improvement to Kulesh's adaptive metrics in the nearest neighbors is also presented. To evaluate the performance of the two proposed metrics, three types of time-series data, namely deterministic synthetic data, chaotic time-series data and real time-series data, are predicted. Experimental results show the superiority of the proposed AR-enhanced k-nearest neighbors methods to the traditional k-nearest neighbors metric and Kulesh's adaptive metrics.
基金supported by the National Natural Science Foundation of China(12131015,12071422)。
文摘In this paper,we consider the limit distribution of the error density function estima-tor in the rst-order autoregressive models with negatively associated and positively associated random errors.Under mild regularity assumptions,some asymptotic normality results of the residual density estimator are obtained when the autoregressive models are stationary process and explosive process.In order to illustrate these results,some simulations such as con dence intervals and mean integrated square errors are provided in this paper.It shows that the residual density estimator can replace the density\estimator"which contains errors.
文摘Long-term time series forecasting stands as a crucial research domain within the realm of automated machine learning(AutoML).At present,forecasting,whether rooted in machine learning or statistical learning,typically relies on expert input and necessitates substantial manual involvement.This manual effort spans model development,feature engineering,hyper-parameter tuning,and the intricate construction of time series models.The complexity of these tasks renders complete automation unfeasible,as they inherently demand human intervention at multiple junctures.To surmount these challenges,this article proposes leveraging Long Short-Term Memory,which is the variant of Recurrent Neural Networks,harnessing memory cells and gating mechanisms to facilitate long-term time series prediction.However,forecasting accuracy by particular neural network and traditional models can degrade significantly,when addressing long-term time-series tasks.Therefore,our research demonstrates that this innovative approach outperforms the traditional Autoregressive Integrated Moving Average(ARIMA)method in forecasting long-term univariate time series.ARIMA is a high-quality and competitive model in time series prediction,and yet it requires significant preprocessing efforts.Using multiple accuracy metrics,we have evaluated both ARIMA and proposed method on the simulated time-series data and real data in both short and long term.Furthermore,our findings indicate its superiority over alternative network architectures,including Fully Connected Neural Networks,Convolutional Neural Networks,and Nonpooling Convolutional Neural Networks.Our AutoML approach enables non-professional to attain highly accurate and effective time series forecasting,and can be widely applied to various domains,particularly in business and finance.
基金Under the auspices of National Natural Science Foundation of China(No.42071230)。
文摘China has resolved its overall regional poverty in 2020 by attaining moderate societal prosperity.The country has entered a new development stage designed to achieve its second centenary goal.However,ecological fragility and risk susceptibility have increased the risk of returning to ecological poverty.In this paper,the Liupan Mountain Region of China was used as a case study,and the counties were used as the scale to reveal the spatiotempora differentiation and influcing factors of the risk of returning to poverty in study area.The indicator data for returning to ecological poverty from 2011-2020 were collected and summarized in three dimensions:ecological,economic and social.The autoregressive integrated moving average model(ARIMA)time series and exponential smoothing method(ES)were used to predict the multidimensional indicators of returning to ecological poverty for 61 counties(districts)in the Liupan Mountain Region for 2021-2030.The back propagation neural network(BPNN)and geographic information system(GIS)were used to generate the spatial distribution and time variation for the index of the risk of returning to ecological poverty(RREP index).The results show that 1)ecological factors were the main factors in the risk of returning to ecological poverty in Liupan Mountain Region.2)The RREP index for the 61 counties(districts)exhibited a downward trend from 2021-2030.The RREP index declined more in medium-and high-risk areas than in low-risk areas.From 2021 to 2025,the RREP index exhibited a slight downward trend.From 2026 to2030,the RREP index was expected to decline faster,especially from 2029-2030.3)Based on the RREP index,it can be roughly divided into three types,namely,the high-risk areas,the medium-risk areas,and the low-risk areas.The natural resource conditions in lowrisk areas of returning to ecological poverty,were better than those in medium-and high-risk areas.
文摘Sentence classification is the process of categorizing a sentence based on the context of the sentence.Sentence categorization requires more semantic highlights than other tasks,such as dependence parsing,which requires more syntactic elements.Most existing strategies focus on the general semantics of a conversation without involving the context of the sentence,recognizing the progress and comparing impacts.An ensemble pre-trained language model was taken up here to classify the conversation sentences from the conversation corpus.The conversational sentences are classified into four categories:information,question,directive,and commission.These classification label sequences are for analyzing the conversation progress and predicting the pecking order of the conversation.Ensemble of Bidirectional Encoder for Representation of Transformer(BERT),Robustly Optimized BERT pretraining Approach(RoBERTa),Generative Pre-Trained Transformer(GPT),DistilBERT and Generalized Autoregressive Pretraining for Language Understanding(XLNet)models are trained on conversation corpus with hyperparameters.Hyperparameter tuning approach is carried out for better performance on sentence classification.This Ensemble of Pre-trained Language Models with a Hyperparameter Tuning(EPLM-HT)system is trained on an annotated conversation dataset.The proposed approach outperformed compared to the base BERT,GPT,DistilBERT and XLNet transformer models.The proposed ensemble model with the fine-tuned parameters achieved an F1_score of 0.88.
基金supported by the National Natural Science Foundation of China under Grant Nos.10971081 and 11001104985 Project of Jilin University
文摘This paper studies the autoregression models of order one, in a general time series setting that allows for weakly dependent innovations. Let {Xt} be a linear process defined by Xt =∑k=0^∞ψ kεt-k, where {ψk, k ≥ 0} is a sequence of real numbers and {εk, k = 0, ±1, ±2,...} is a sequence of random variables. Two results are proved in this paper. In the first result, assuming that {εk, k ≥ 1} is a sequence of asymptotically linear negative quadrant dependent （ALNQD） random variables, the authors find the limiting distributions of the least squares estimator and the associated regression t statistic. It is interesting that the limiting distributions are similar to the one found in earlier work under the assumption of i.i.d, innovations. In the second result the authors prove that the least squares estimator is not a strong consistency estimator of the autoregressive parameter a when {εk, k ≥ 1} is a sequence of negatively associated （NA） random variables, and ψ0 = 1, ψk = 0, k ≥ 1.
文摘China＇s astonishing economic growth implies a necessity to understand its inflation. The present paper employs threshold nonrecursive structural vector autoregression analysis to explore the asymmetric effects of macro-variables on inflation in low and high inflation regimes. The empirical evidence demonstrates, first, that the reactions of inflation to various shocks are inflation-regime-dependent and asymmetric. Second, monetary policy influences China ＂s high inflation and adjusting the domestic interest rate in China may be an effective way to control inflation in a high inflation regime, but not in a low inflation regime. In a high inflation regime, a high inflation rate may cause the macro-policy authorities to increase the domestic interest rate, in an attempt to stabilize high inflation. Third, contrary to expectations, the world oil price is not a strong cost-push factor in a low inflation regime. Oil price increases may increase inflation in a high inflation regime, but there is no such obvious effect in a low inflation regime. Finally, China ＂s nominal effective exchange rate influences inflation in both low and high inflation regimes. A nominal effeetive exchange rate appreciation might be effective in controlling domestic inflation in both regimes.
文摘Through extension of canonical correlation to the analysis of meteorological element fields (MEF), a concept from combination of canonical autocorrelation with canonical autoregression (CAR) is developed for short-term climatic prediction of MEFs with a formulated scheme. Experi- mental results suggest that the scheme is of encouraging usefulness to a weak persistence MEF, i.e., rainfall field and, in particular, to a strong persistance one like a SST field.
基金The anthors thank the support from Tianjin Philosophy and Social Science Planning Project(No.TJYYQN19-004)Project of National and Regional Research Center,Ministry of Education(No.ZX20170183)National Natural Science Foundation Youth Project(No.71803089).
文摘Since 2002,the People's Bank of China has frequently used both quantity-based direct monetary instruments and price-based indirect monetary instruments to promote economic growth and stabilize price level.Specifically,this study estimates 13 three-variable factor-augmented vector autoregression (FAVAR) models to explore how two types of monetary instruments affect China's economy and price level.Overall,we find that monetary policy has positive effects on China's economy and price level.Second,this study clearly states that the effectiveness of China's monetary policy on the economy has depended on China's quantity-based direct monetary instruments since 2002.Third,the effectiveness of quantity-based direct monetary instruments on China's economy and price level is dependent on the significant and positive effects of quantity-based direct monetary instruments after the 2008 financial crisis.Fourth,the significant and positive effects of price-based indirect monetary instruments on China's economy and price level before 2008 cannot fundamentally change their current insignificant effects on China's economy and price level.
文摘The stock market is a vital component of the broader financial system,with its dynamics closely linked to economic growth.The challenges associated with analyzing and forecasting stock prices have persisted since the inception of financial markets.By examining historical transaction data,latent opportunities for profit can be uncovered,providing valuable insights for both institutional and individual investors to make more informed decisions.This study focuses on analyzing historical transaction data from four banks to predict closing price trends.Various models,including decision trees,random forests,and Long Short-Term Memory(LSTM)networks,are employed to forecast stock price movements.Historical stock transaction data serves as the input for training these models,which are then used to predict upward or downward stock price trends.The study’s empirical results indicate that these methods are effective to a degree in predicting stock price movements.The LSTM-based deep neural network model,in particular,demonstrates a commendable level of predictive accuracy.This conclusion is reached following a thorough evaluation of model performance,highlighting the potential of LSTM models in stock market forecasting.The findings offer significant implications for advancing financial forecasting approaches,thereby improving the decision-making capabilities of investors and financial institutions.
文摘In the parameter tracking of time-varying systems, the ordinary method is weighted least squares with the rectangular window or the exponential window. In this paper we propose a new kind of sliding window called the multiple exponential window, and then use it to fit time-varying Gaussian vector autoregressive models. The asymptotic bias and covariance of the estimator of the parameter for time-invariant models are also derived. Simulation results show that the multiple exponential windows have better parameter tracking effect than rectangular windows and exponential ones.
基金This work has been supported by DARPA TRADES Award HR0011-17-2-0016.
文摘Although the popular multi-fidelity surrogate models,stochastic collocation and nonlinear autoregression have been applied successfully to multiple benchmark problems in different areas of science and engineering,they have certain limitations.We propose a uniform Bayesian framework that connects these two methods allowing us to combine the strengths of both.To this end,we introduce Greedy-NAR,a nonlinear Bayesian autoregressive model that can handle complex between-fidelity correlations and involves a sequential construction that allows for significant improvements in performance given a limited computational budget.The proposed enhanced nonlinear autoregressive method is applied to three benchmark problems that are typical of energy applications,namely molecular dynamics and computational fluid dynamics.The results indicate an increase in both prediction stability and accuracy when compared to those of the standard multi-fidelity autoregression implementations.The results also reveal the advantages over the stochastic collocation approach in terms of accuracy and computational cost.Generally speaking,the proposed enhancement provides a straightforward and easily implemented approach for boosting the accuracy and efficiency of concatenated structure multi-fidelity simulation methods,e.g.,the nonlinear autoregressive model,with a negligible additional computational cost.
基金Supported by the Key Scientific Research Project of Universities in Henan Province,No.21A330004Natural Science Foundation in Henan Province,No.222300420265.
文摘BACKGROUND Hepatitis B(HB)and hepatitis C(HC)place the largest burden in China,and a goal of eliminating them as a major public health threat by 2030 has been set.Making more informed and accurate forecasts of their spread is essential for developing effective strategies,heightening the requirement for early warning to deal with such a major public health threat.AIM To monitor HB and HC epidemics by the design of a paradigmatic seasonal autoregressive fractionally integrated moving average(SARFIMA)for projections into 2030,and to compare the effectiveness with the seasonal autoregressive integrated moving average(SARIMA).METHODS Monthly HB and HC incidence cases in China were obtained from January 2004 to June 2023.Descriptive analysis and the Hodrick-Prescott method were employed to identify trends and seasonality.Two periods(from January 2004 to June 2022 and from January 2004 to December 2015,respectively)were used as the training sets to develop both models,while the remaining periods served as the test sets to evaluate the forecasting accuracy.RESULTS There were incidents of 23400874 HB cases and 3590867 HC cases from January 2004 to June 2023.Overall,HB remained steady[average annual percentage change(AAPC)=0.44,95%confidence interval(95%CI):-0.94-1.84]while HC was increasing(AAPC=8.91,95%CI:6.98-10.88),and both had a peak in March and a trough in February.In the 12-step-ahead HB forecast,the mean absolute deviation(15211.94),root mean square error(18762.94),mean absolute percentage error(0.17),mean error rate(0.15),and root mean square percentage error(0.25)under the best SARFIMA(3,0,0)(0,0.449,2)12 were smaller than those under the best SARIMA(3,0,0)(0,1,2)12(16867.71,20775.12,0.19,0.17,and 0.27,respectively).Similar results were also observed for the 90-step-ahead HB,12-step-ahead HC,and 90-step-ahead HC forecasts.The predicted HB incidents totaled 9865400(95%CI:7508093-12222709)cases and HC totaled 1659485(95%CI:856681-2462290)cases during 2023-2030.CONCLUSION Under current interventions,China faces enormous challenges to eliminate HB and HC epidemics by 2030,and effective strategies must be reinforced.The integration of SARFIMA into public health for the management of HB and HC epidemics can potentially result in more informed and efficient interventions,surpassing the capabilities of SARIMA.
基金from funding agencies in the public,commercial,or not-for-profit sectors.
文摘The study aims to investigate the financial technology(FinTech)factors influencing Chinese banking performance.Financial expectations and global realities may be changed by FinTech’s multidimensional scope,which is lacking in the traditional financial sector.The use of technology to automate financial services is becoming more important for economic organizations and industries because the digital age has seen a period of transition in terms of consumers and personalization.The future of FinTech will be shaped by technologies like the Internet of Things,blockchain,and artificial intelligence.The involvement of these platforms in financial services is a major concern for global business growth.FinTech is becoming more popular with customers because of such benefits.FinTech has driven a fundamental change within the financial services industry,placing the client at the center of everything.Protection has become a primary focus since data are a component of FinTech transactions.The task of consolidating research reports for consensus is very manual,as there is no standardized format.Although existing research has proposed certain methods,they have certain drawbacks in FinTech payment systems(including cryptocurrencies),credit markets(including peer-to-peer lending),and insurance systems.This paper implements blockchainbased financial technology for the banking sector to overcome these transition issues.In this study,we have proposed an adaptive neuro-fuzzy-based K-nearest neighbors’algorithm.The chaotic improved foraging optimization algorithm is used to optimize the proposed method.The rolling window autoregressive lag modeling approach analyzes FinTech growth.The proposed algorithm is compared with existing approaches to demonstrate its efficiency.The findings showed that it achieved 91%accuracy,90%privacy,96%robustness,and 25%cyber-risk performance.Compared with traditional approaches,the recommended strategy will be more convenient,safe,and effective in the transition period.
基金supported by the Energy Cloud R&D Program through the National Research Foundation of Korea(NRF)funded by the Ministry of Science,ICT(2019M3F2A1073164)MSGSU BAP(2021-25).
文摘Construction Industry operates relying on various key economic indicators.One of these indicators is material prices.On the other hand,cost is a key concern in all operations of the construction industry.In the uncertain conditions,reliable cost forecasts become an important source of information.Material cost is one of the key components of the overall cost of construction.In addition,cost overrun is a common problem in the construction industry,where nine out of ten construction projects face cost overrun.In order to carry out a successful cost management strategy and prevent cost overruns,it is very important to find reliable methods for the estimation of construction material prices.Material prices have a time dependent nature.In order to increase the foreseeability of the costs of construction materials,this study focuses on estimation of construction material indices through time series analysis.Two different types of analysis are implemented for estimation of the future values of construction material indices.The first method implemented was Autoregressive Integrated Moving Average(ARIMA),which is known to be successful in estimation of time series having a linear nature.The second method implementedwas Non-LinearAutoregressive Neural Network(NARNET)which is known to be successful in modeling and estimating of series with non-linear components.The results have shown that depending on the nature of the series,both these methods can successfully and accurately estimate the future values of the indices.In addition,we found out that Optimal NARNET architectures which provide better accuracy in estimation of the series can be identified/discovered as result of grid search on NARNET hyperparameters.
