Search

A reliable and efficient forecasting system can be used to warn the general public against the increasing PM2.5 concentration. This paper proposes a novel AdaBoost‐ensemble technique based on a hybrid data preprocessing‐analysis strategy, with the following contributions: (i) a new decomposition strategy is proposed based on the hybrid data preprocessing‐analysis strategy, which combines the merits of two popular decomposition algorithms and has been proven to be a promising decomposition strategy; (ii) the long short‐term memory (LSTM), as a powerful deep learning forecasting algorithm, is applied to individually forecast the decomposed components, which can effectively capture the long‐short patterns of complex time series; and (iii) a novel AdaBoost‐LSTM ensemble technique is then developed to integrate the individual forecasting results into the final forecasting results, which provides significant improvement to the forecasting performance. To evaluate the proposed model, a comprehensive and scientific assessment system with several evaluation criteria, comparison models, and experiments is designed. The experimental results indicate that our developed hybrid model considerably surpasses the compared models in terms of forecasting precision and statistical testing and that its excellent forecasting performance can guide in developing effective control measures to decrease environmental contamination and prevent the health issues caused by a high PM2.5 concentration. [ABSTRACT FROM AUTHOR]

Practitioners and policy makers rely on accurate crude oil forecasting to avoid price risks and grasp investment opportunities, but the core of existing predictive models for such prices is based on point-valued inputs and outputs, which may suffer from informational loss of volatility. This paper addresses this issue by proposing a modified threshold autoregressive interval-valued models with interval-valued factors (MTARIX), as extended by Sun et al. [Threshold autoregressive models for interval-valued time series. J. Econom., 2018, 206, 414–446], to analyze and forecast interval-valued crude oil prices. In contrast to point-valued data methods, MTARIX models simultaneously capture nonlinear features in price trend and volatility, and this informational gain can produce more accurate forecasts. Several interval-valued factors and point-valued threshold variables are analyzed, including supply and demand, speculation, stock market, monetary market, technical factor, and search query data. Empirical results suggest that MTARIX models with appropriate threshold variables outperform other competing forecast models (ACIX, CR-SETARX, ARX, and VARX). The findings indicate that oil price range information is more valuable than oil price level information in forecasting crude oil prices. [ABSTRACT FROM AUTHOR]

Crude oil prices are of vital importance for market participants and governments to make energy policies and decisions. In this paper, we apply a newly proposed autoregressive conditional interval (ACI) model to forecast crude oil prices. Compared with the existing point-based forecasting models, the interval-based ACI model can capture the dynamics of oil prices in both level and range of variation in a unified framework. Rich information contained in interval-valued observations can be simultaneously utilized, thus enhancing parameter estimation efficiency and model forecasting accuracy. In forecasting the monthly West Texas Intermediate (WTI) crude oil prices, we document that the ACI models outperform the popular point-based time series models. In particular, ACI models deliver better forecasts than univariate ARMA models and the vector error correction model (VECM). The gain of ACI models is found in out-of-sample monthly price interval forecasts as well as forecasts for point-valued highs, lows, and ranges. Compared with GARCH and conditional autoregressive range (CARR) models, ACI models are also superior in volatility (conditional variance) forecasts of oil prices. A trading strategy that makes use of the monthly high and low forecasts is further developed. This trading strategy generally yields more profitable trading returns under the ACI models than the point-based VECM. [ABSTRACT FROM AUTHOR]

Forecasting air passenger flows is receiving increasing attention, especially due to its intrinsic difficulties and wide applications. Total passengers are used as a proxy for air transport demand. However, the time series of air passenger flows usually has complicated behavior with high volatility and irregularity. This paper proposes a MIV-based nonlinear vector auto-regression neural network (NVARNN) approach to forecast air passenger flows. In the proposed MIV-NVARNN learning approach, (1) a method of mean impact value (MIV) based on neural network is used for identifying and extracting input variables; (2) NVARNN is firstly proposed to deal with the irregularity and volatility of the time series of air passenger flows. To illustrate and verify the effectiveness of the proposed approach, we tested its directional and level forecasting accuracy using the time series of Beijing International Airport's passenger flows. The results of out-of-sample forecasting performance show that the proposed MIV-NVARNN approach consistently outperforms single models and other hybrid approaches in terms of level forecasting accuracy, directional forecasting accuracy and robustness analysis. • A new hybrid approach is proposed for air passenger flow forecasting. • Two steps are involved: variable selection and forecast modelling. • Nonlinear vector auto-regression neural network is firstly developed to forecast air passenger flow. • Empirical results statistically verify the forecasting performance of our new hybrid approach. [ABSTRACT FROM AUTHOR]

We propose a model-free test for strict stationarity. The idea is to estimate a nonparametric time-varying characteristic function and compare it with the empirical characteristic function based on the whole sample. We also propose several derivative tests to check time-invariant moments, weak stationarity, and pth order stationarity. Monte Carlo studies demonstrate excellent power of our tests. We apply our tests to various macroeconomic time series and find overwhelming evidence against strict and weak stationarity for both level and first-differenced series. This suggests that the conventional time series econometric modeling strategies may have room to be improved by accommodating these time-varying features. [ABSTRACT FROM AUTHOR]

We propose an efficient numerical integration-based nonparametric entropy estimator for serial dependence and show that the new entropy estimator has a smaller asymptotic variance than Hong and White’s (2005) sample average-based estimator. This delivers an asymptotically more efficient test for serial dependence. In particular, the uniform kernel gives the smallest asymptotic variance for the numerical integration-based entropy estimator over a class of positive kernel functions. Moreover, the naive bootstrap can be used to obtain accurate inferences for our test, whereas it is not applicable to Hong and White’s (2005) sample averaging approach. A simulation study confirms the merits of our approach. [ABSTRACT FROM AUTHOR]

This paper proposes two types of dummy variables for an interval regression model to assess the impact of economic shocks/crises on an interval time series (ITS), e.g. daily intervals of energy prices. We present different economic interpretations of the two types of dummy variables for an interval regression model. Particularly, we discuss how they measure the direction and magnitudes of the change of an ITS caused by an economic crisis, and develop the corresponding hypothesis tests. A main advantage of the proposed ITS modelling approach over traditional point-based methods is that it can assess the change in both the trend and volatility of an asset price process simultaneously. This is due to the informational gain of an ITS sample over a point-valued sample, e.g. closing prices, since an interval observation contains both the trend and variation information of a price process in a given period. Using the proposed interval framework, we focus on the impact of the subprime mortgage crisis in the commodity market as a case study based on the ITS of monthly crude oil future price data. Empirical results suggest a strong evidence that the subprime crisis has lowered the level/trend and increased the volatility of crude oil prices. We also show that the trend of crude oil future prices moves towards an equilibrium state driven by the variation of the price process in last period, and the speculation index, as a proxy of crude oil market liquidity, is significant in explaining the dynamics of crude oil prices. Both findings provide quantitative evidence for theoretical results in the previous literature. [ABSTRACT FROM AUTHOR]

Abstract: In this study, two hybrid approaches based on seasonal decomposition and least squares support vector regression (LSSVR) model are proposed for short-term forecasting of air passenger. In the formulation of the proposed hybrid approaches, the air passenger time series is first decomposed into three components: trend-cycle component, seasonal factor and irregular component. Then the LSSVR model is used to predict the components independently and these prediction results of the components are combined as an aggregated output. Empirical analysis shows that the proposed hybrid approaches are better than other time series models, indicating that they are promising tools to predict complex time series with high volatility and irregularity. [Copyright &y& Elsevier]

For time series, the problem that we often encounter is how to extract the patterns hidden in the real world data for forecasting its future values. A single linear or nonlinear model is inadequate in modeling and forecasting the time series, because most of them usually contain both linear and nonlinear patterns. This study constructs a hybrid forecasting model that combines autoregressive integrated moving average (ARIMA) with Elman artificial neural network (ANN) for short-term forecasting of time series. The proposed approach considers the linear and nonlinear patterns in the real data simultaneously so that it can mine more precise characteristics to describe the time series better. Finally, the forecasting results of the hybrid model are adjusted with the knowledge from text mining and expert system. The empirical results on the container throughput forecast of Tianjin Port show that the forecasts by the hybrid model are superior to those of ARIMA model and Elman network. [ABSTRACT FROM AUTHOR]

In this paper, a hybrid model integrating wavelet decomposition and least squares support machines (LSSVM) is proposed for crude oil price forecasting. In this model, the Haar à trous wavelet transform is first selected to decompose an original time series into several sub-series with different scales. Then the LSSVM is used to predict each sub-series. Subsequently, the final oil price forecast is obtained by reconstructing the results of the sub-series forecasts. The experimental results show that the integrated model, based on multi-scale wavelet decomposition, outperforms the traditional single-scale models. Furthermore, the proposed hybrid model is the best among all the models compared in this study. To fully integrate the advantages of several models, a combined forecasting model is presented. The study shows that the combined forecasting model is clearly better than any individual model for crude oil price forecasting. [ABSTRACT FROM AUTHOR]

In this study, a novel decomposition-ensemble methodology is proposed for container throughput forecasting. Firstly, the sample data of container throughput at ports are decomposed into several components. Secondly, the time series of the various components are thoroughly investigated to accurately capture the data characteristics. Then, an individual forecasting model is selected for each component based on the data characteristic analysis (DCA). Finally, the forecasting results are combined as an aggregated output. An empirical analysis is implemented for illustration and verification purposes. Our results suggest that proposed hybrid models can achieve better performance than other methods. [ABSTRACT FROM AUTHOR]

Nowadays information technology advances allow the collecting and storage of large complex datasets in many areas. Modeling and forecasting interval-valued time series (ITS) has drawn much attention over the last two decades because interval-valued observations contain more information than point-valued observations over the same period and remove undesirable noises in high-frequency data. However, most work mainly focuses on modeling a linear univariate ITS or bivariate point process. This paper proposes nonparametric regression models for interval-valued time series with imposing constraints, e.g., monotonicity. This setting with a monotonic constraint is consistent with the existing literature, which focuses on incorporating valuable empirical information in modeling and forecasts. Two constraint estimators are developed and asymptotic properties are established. Monte Carlo simulation is conducted to show the finite sample performance. An empirical application to equity premium documents that the proposed model yields a better forecast performance than some popular models in the literature. • First applies the parsimonious nonparametric spirit to interval time series. • Provide an algorithm obtaining the constraint interval estimators with bagging. • Derive the consistency and limit distribution of the proposed estimators. • Outperforms existing methods in the interval predictive model for stock returns. [ABSTRACT FROM AUTHOR]

To enhance the accuracy of interval carbon price forecasting, this study proposes a secondary decomposition-ensemble framework. Firstly, the bivariate empirical mode decomposition (BEMD) is applied for primary decomposition of the original interval-valued time series (ITS). Next, the multi-scale permutation entropy (MPE) is introduced to measure the unpredictability of each decomposed component ITS, and the multivariate variational mode decomposition (MVMD) is employed to implement secondary decomposition of the component ITS with the highest complexity. Then, a sparrow search algorithm-enhanced interval multi-layer perceptron (SSA-iMLP) is developed for forecasting each component ITS. Finally, all forecasts of component ITSs are aggregated into ITS forecasts of carbon prices. Using carbon price ITS data from Hubei and Guangdong Emission Exchanges in China, empirical analysis is conducted. The results show that our proposed model has higher predictive accuracy and stronger robustness than benchmark models, indicating that the framework is promising for ITS forecasting in complex scenarios. • Bivariate empirical mode decomposition is used for primary decomposition. • Multi-scale permutation entropy is introduced to measure the unpredictability. • A secondary decomposition is performed to the component ITS with the highest complexity. • An interval multi-layer perception model with sparrow search algorithm is developed. • Our proposed model exhibits superiority of predictive accuracy. [ABSTRACT FROM AUTHOR]

Short-term load forecasting has evolved into an important aspect of power system in safe operation and rational dispatching. However, given the load series' instability and volatility, this is a challenging task. To this end, this study proposes a dynamic decomposition-reconstruction-ensemble approach by cleverly and dynamically combining two proven and effective techniques (i.e., the reconstruction techniques and the secondary decomposition techniques). In fact, by introducing the decomposition-reconstruction process based on the dynamic classification, filtering, and giving the criteria for determining the components that need to be decomposed again, our proposed model improves the decomposition-ensemble forecasting framework. Our proposed model makes full use of decomposition techniques, complexity analysis, reconstruction techniques, secondary decomposition techniques, and a neural network optimized by an automatic hyperparameter optimization algorithm. Besides, we compared our proposed model with state-of-the-art models including five models with reconstruction techniques and two models with secondary decomposition techniques. The experiment results demonstrate the superiority of our proposed dynamic decomposition-reconstruction technique in terms of forecasting accuracy, precise direction, equality, stability, correlation, comprehensive accuracy, and statistical tests. To conclude, our proposed model has the potential to be a useful tool for short-term load forecasting. • A dynamic decomposition-reconstruction strategy is proposed for time series forecasting. • Improve the existing decomposition-ensemble forecasting framework. • Three sets of real load data were utilized to evaluate the forecasting performance. • Experiments verify the effectiveness of dynamic decomposition-reconstruction strategy. [ABSTRACT FROM AUTHOR]

• Strategies of previous tourism literature are summarized. • A meticulous review and comparison of five strategies is conducted. • The performance of five forecasting strategies is explored and compared. • The DIRMO strategy is most practical among all these strategies. • Empirical findings provide a reference paradigm for future research. Previous researches have proposed five strategies to deal with complex multi-step ahead forecasting tasks. However, these strategies have not received much attention in the field of tourism research and the performance of them is still unknown. Accordingly, we summarize the strategies used in multi-step ahead tourism demand forecasting articles and produce a comparative analysis of five strategies. By employing nine tourist arrival time series in Hong Kong, the study explores the performance of different strategy-driven forecasting approaches in tourism demand. The empirical results show that Direct and DIRMO (s = 2) strategies perform better than other strategies in terms of forecasting accuracy, but there is no significant difference between them, excepting the latter is lower than the former in computational complexity. [ABSTRACT FROM AUTHOR]

Despite the considerable deployed resources, current policing efforts are failing to stop crimes before they start, and therefore, also failing to adequately protect lives and property. To promote the intelligent transformation from reactive to proactive policing, this study proposed a hierarchical crime prediction framework. First, the temporal dependency in the frequency domain was decomposed and a network constructed to capture the spatial relationships within the sub-frequencies. Human mobility in a city was then utilized to characterize the dynamic relationships within the network. Using the proposed framework, this study examined the crime distribution evolution in Chicago to holistically predict the short-term crimes in the different communities. The framework was found to have high predictive accuracy and significant potential in promoting proactive policing. It was concluded that: (1) as the crime distribution evolution comes from the spatial relationship changes, these dynamic relationships are critical in explaining and characterizing the evolution; and (2) the social interactions constructed using the human activity data can characterize the dynamic crime distribution relationships. • Development of a model to forecast short-term crime in a community network. • Decomposing time series and capturing spatial patterns within decompositions. • Utilizing taxis to proxy mobility within and between communities. • Explaining how crime distribution change within and across communities. [ABSTRACT FROM AUTHOR]

• A decomposition–ensemble model is proposed for interval-valued load forecasting. • Multivariate multiscale permutation entropy technique is utilized to perform complexity analysis on decomposed interval-valued components for capturing internal features and reconstructing the components. • An automatic Bayesian optimization algorithm based on the Tree-structured Parzen Estimator algorithm is used in hyperparameter optimization. • Empirical results verify that the proposed approach outperforms other benchmarks under study. Short-term load forecasting is crucial for power demand-side management and the planning of the power system. Considering the necessity of interval-valued time series modeling and forecasting for the power system, this study proposes an interval decomposition-reconstruction-ensemble learning approach to forecast interval-valued load, in terms of the concept of "divide and conquer". First, bivariate empirical mode decomposition is applied to decompose the original interval-valued data into a finite number of bivariate modal components for extracting and identifying the fluctuation characteristics of data. Second, based on the complexity analysis of each bivariate modal component by multivariate multiscale permutation entropy, the components were reconstructed for capturing inner factors and reduce the accumulation of estimation errors. Third, long short-term memory is utilized to synchronously forecast the upper and the lower bounds of each bivariate component and optimized by the Bayesian optimization algorithm. Finally, generating the aggregated interval-valued output by ensemble the forecasting results of the upper and lower bounds of each component severally. The electric load of five states in Australia is used for verification, and the empirical results show that the forecasting accuracy of our proposed learning approach is significantly superior to single models and the decomposition-ensemble models without reconstruction. This indicates that our proposed learning approach appears to be a promising alternative for interval load forecasting. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, a novel hybrid ensemble learning paradigm integrating ensemble empirical mode decomposition (EEMD) and least squares support vector regression (LSSVR) is proposed for nuclear energy consumption forecasting, based on the principle of “decomposition and ensemble”. This hybrid ensemble learning paradigm is formulated specifically to address difficulties in modeling nuclear energy consumption, which has inherently high volatility, complexity and irregularity. In the proposed hybrid ensemble learning paradigm, EEMD, as a competitive decomposition method, is first applied to decompose original data of nuclear energy consumption (i.e. a difficult task) into a number of independent intrinsic mode functions (IMFs) of original data (i.e. some relatively easy subtasks). Then LSSVR, as a powerful forecasting tool, is implemented to predict all extracted IMFs independently. Finally, these predicted IMFs are aggregated into an ensemble result as final prediction, using another LSSVR. For illustration and verification purposes, the proposed learning paradigm is used to predict nuclear energy consumption in China. Empirical results demonstrate that the novel hybrid ensemble learning paradigm can outperform some other popular forecasting models in both level prediction and directional forecasting, indicating that it is a promising tool to predict complex time series with high volatility and irregularity. [Copyright &y& Elsevier]