Traditional Autoregressive Research Articles

Day-ahead spatiotemporal solar irradiation forecasting using frequency-based hybrid principal component analysis and neural network

Owing to a shortage of fossil fuels, environmental pollution and the greenhouse effect, renewable energy generation has become important in a modern smart grid. However, the characteristics of renewable power generation are volatile and uncertain. This work proposes a new day-ahead spatiotemporal forecasting method for solar irradiation to ensure the efficient operation of power systems. First, the frequency features of a solar irradiation time-series are extracted by discrete Fourier transform (DFT). Principal component analysis (PCA) is then used to identify the crucial frequency features, which are input to an Elman-based neural network to carry out subsequent 24-hour (day ahead) solar irradiation forecasting. Historical weather data from five weather stations that are near the target location are used. Comparative studies of traditional Autoregressive Integrated Moving Average Model (ARIMA), PCA-Back-Propagation (BP)-based neural network, the persistence method, DFT-PCA-BP and the proposed DFT-PCA-Elman method reveal that the proposed method is the most accurate in day-ahead forecasting using spatiotemporal data.

ARPILC: An Approach for Short-Term Prediction of Freeway Entrance Flow

In modern intelligent transportation systems, short-term prediction of freeway entrance flow plays an important role in providing travelers timely traffic conditions. Short-term prediction approaches are expected to be computation-efficient, operation-convenient, and able to forecast with high accuracy. Nevertheless, few existing prediction approaches satisfy all the requirements at the same time. For addressing this issue, we propose an autoregressive (AR) model parameterized by iterative learning control (ILC) (called ARPILC) to make short-term predictions of freeway entrance flows. Different from traditional AR models which have constant parameters, the proposed AR method has time-dependent parameters which are optimized by the ILC method with historical dataset. In this manner, the ARPILC model can describe time-dependent characteristics, especially nonlinear characteristics, of freeway entrance flows much better. The ARPILC model predicts short-term entrance flows as weighted sums of the outputs of several AR models with time-dependent parameters. We validate the ARPILC model with the toll collection data during July 2012 in Jiangsu province, China, and compare it with several existing prediction approaches. The results show that the ARPILC model can predict freeway entrance flows with a high accuracy and has better prediction performance than existing approaches.

Hybrid algorithm for short-term forecasting of PM2.5 in China

In recent years, the forecasting of particles with a diameter of 2.5 μm or less (PM2.5) has been a popular research topic, and involves multiple sources of pollution, making it difficult to determine all of the contributing meteorological and environmental factors. When only the PM2.5 concentration time series is considered without other exogenous information, accurate forecasting is important and should be efficient. To address this problem, this paper proposes a hybrid algorithm consisting of multiple models to improve prediction accuracy. The innovation of the proposed hybrid algorithm is to decompose the original single one-dimensional (1D) PM2.5 data into multi-dimensional information which effectively mines information hidden in the 1D data. Then, it uses traditional prediction methods to forecast each sequence and to reconstruct its forecasting results to obtain the final forecasting results. Three hybrid models, Wavelet-ANN, Wavelet-ARIMA and Wavelet-SVM, are developed to forecast the 2016 PM2.5 trends in 5 Cities in China. The results showed that: (1) Hybrid models (Wavelet-ANN, Wavelet-ARIMA and Wavelet-SVM) can forecast short-term PM2.5 concentrations in China. Compared with the traditional Autoregressive Integrated Moving Average (ARIMA), Artificial Neural Network (ANN) and Support Vector Machine (SVM) models, hybrid models can significantly improve prediction accuracy. (2) The Wavelet-ARIMA model has higher accuracy with respect to predicting PM2.5 concentrations. In particular, it can more accurately capture the mutational points of PM2.5 concentrations, which can provide effective information support for generating warnings about atmospheric pollution. The hybrid algorithm proposed in this paper can be effectively applied to the short-term forecasting of PM2.5 concentrations and can significantly improve the accuracy of prediction.

Using the AR–SVR–CPSO hybrid model to forecast vibration signals in a high-speed train transmission system

Accurate vibration time series modeling can mine the internal law of data and provide valuable references for reliability assessment. To improve the prediction accuracy, this study proposes a hybrid model – called the AR–SVR–CPSO hybrid model – that combines the auto regression (AR) and support vector regression (SVR) models, with the weights optimized by the chaotic particle swarm optimization (CPSO) algorithm. First, the auto regression model with the difference method is employed to model the vibration time series. Second, the support vector regression model with the phase space reconstruction is constructed for predicting the vibration time series once more. Finally, the predictions of the AR and SVR models are weighted and summed together, with the weights being optimized by the CPSO. In addition, the data collected from the reliability test platform of high-speed train transmission systems and the “NASA prognostics data repository” are used to validate the hybrid model. The experimental results demonstrate that the hybrid model proposed in this study outperforms the traditional AR and SVR models.

Mobile Demand Forecasting via Deep Graph-Sequence Spatiotemporal Modeling in Cellular Networks

The demand forecasting plays a crucial role in the predictive physical and virtualized network management in cellular networks, which can effectively reduce both the capital and operational expenditures by fully exploiting the network infrastructure. In this paper, we study the per-cell demand forecasting in cellular networks. The success of demand forecasting relies on the effective modeling of both the spatial and temporal aspects of the per-cell demand time series. However, the main challenge of the spatial relevancy modeling in the per-cell demand forecasting is the irregular spatial distribution of cells in a network, where applying grid-based models (e.g., convolutional neural networks) would lead to degradation of spatial granularity. In this paper, we propose to model the spatial relevancy among cells by a dependency graph based on spatial distances among cells without the loss of spatial granularity. Such spatial distance-based graph modeling is confirmed by the spatiotemporal analysis via semivariogram, which suggests that the relevancy between any two cells declines as their spatial distance increases. Hence, the graph convolutional networks and long short-term memory (LSTM) from deep learning are employed to model the spatial and temporal aspects, respectively. In addition, the deep graph-sequence model, graph convolutional LSTM, is further employed to simultaneously characterize both the spatial and temporal aspects of mobile demand forecasting. Experiments demonstrate that our proposed graph-sequence demand forecasting model could achieve a superior forecasting performance compared with the other two proposed models as well as the traditional auto regression integrated moving average time series model.

Isophote-Constrained Autoregressive Model With Adaptive Window Extension for Image Interpolation

The autoregressive (AR) model is widely used in image interpolations. Traditional AR models consider utilizing the dependence between pixels to model the image signal. However, they ignore the valuable patch-level information for image modeling. In this paper, we propose to integrate both the pixel-level and patch-level information to depict the relationship between high-resolution and low-resolution pixels and obtain better image interpolation results. In particular, we propose an isophote-constrained AR (ICAR) model to perform AR-flavored interpolation within an identified joint stable region and further develop an AR interpolation with an adaptive window extension. Considering the smoothness along the isophote curve, the ICAR model searches only several successive similar patches along the isophote curve over a large region to construct an adaptive window. These overlapped patches, representing the patch-level structure similarity, are used to construct a joint AR model. To better characterize the piecewise stationarity and determine whether a pixel is suitable for AR estimation, we further propose pixel-level and patch-level similarity metrics and embed them into the ICAR model, introducing a weighted ICAR model. Comprehensive experiments demonstrate that our method can effectively reconstruct the edge structures and suppress jaggy or ringing artifacts. In the objective quality evaluation, our method achieves the best results in terms of both peak signal-to-noise ratio and structural similarity for both simple size doubling (two times) and for arbitrary scale enlargements.

Neural network copula portfolio optimization for exchange traded funds

This paper attempts to investigate if adopting accurate forecasts from Neural Network (NN) models can lead to statistical and economically significant benefits in portfolio management decisions. In order to achieve that, three NNs, namely the Multi-Layer Perceptron, Recurrent Neural Network and the Psi Sigma Network (PSN), are applied to the task of forecasting the daily returns of three Exchange Traded Funds (ETFs). The statistical and trading performance of the NNs is benchmarked with the traditional Autoregressive Moving Average models. Next, a novel dynamic asymmetric copula model (NNC) is introduced in order to capture the dependence structure across ETF returns. Based on the above, weekly re-balanced portfolios are obtained and compared using the traditional mean–variance and the mean–CVaR portfolio optimization approach. In terms of the results, PSN outperforms all models in statistical and trading terms. Additionally, the asymmetric skewed t copula statistically outperforms symmetric copulas when it comes to modelling ETF returns dependence. The proposed NNC model leads to significant improvements in the portfolio optimization process, while forecasting covariance accounting for asymmetric dependence between the ETFs also improves the performance of obtained portfolios.

Structural health monitoring data reconstruction of a concrete cable-stayed bridge based on wavelet multi-resolution analysis and support vector machine

The accuracy and integrity of stress data acquired by bridge heath monitoring system is of significant importance for bridge safety assessment. However, the missing and abnormal data are inevitably existed in a realistic monitoring system. This paper presents a data reconstruction approach for bridge heath monitoring based on the wavelet multi-resolution analysis and support vector machine (SVM). The proposed method has been applied for data imputation based on the recorded data by the structural health monitoring (SHM) system instrumented on a prestressed concrete cable-stayed bridge. The effectiveness and accuracy of the proposed wavelet-based SVM prediction method is examined by comparing with the traditional autoregression moving average (ARMA) method and SVM prediction method without wavelet multi-resolution analysis in accordance with the prediction errors. The data reconstruction analysis based on 5-day and 1-day continuous stress history data with obvious preternatural signals is performed to examine the effect of sample size on the accuracy of data reconstruction. The results indicate that the proposed data reconstruction approach based on wavelet multi-resolution analysis and SVM is an effective tool for missing data imputation or preternatural signal replacement, which can serve as a solid foundation for the purpose of accurately evaluating the safety of bridge structures.

Autoregressive forests for multivariate time series modeling

Multivariate Time Series (MTS) modeling has received significant attention in the last decade because of the complex nature of the data. Efficient representations are required to deal with the high dimensionality due to the increase in the number of variables and duration of the time series in different applications. For example, model-based approaches such as Hidden Markov Models (HMM) or autoregressive (AR) models focus on finding a model to represent the series with the model parameters to handle this problem. Both HMM and AR models are known to be very successful in the representation of the time series however most of the HMM approaches assume independence and traditional AR models consider linear dependence between the variables of MTS. As most of the real systems exhibit nonlinear relations, traditional approaches fail to represent the time series. To handle these problems, we propose an autoregressive tree-based ensemble approach that can model the nonlinear behavior embedded in the time series with the help of tree-based learning. Multivariate autoregressive forest, namely mv-ARF, is a nonparametric vector autoregression approach which provides an easy and efficient representation that scales well with large datasets. An error-based representation based on the learned models is the basis of the proposed approach. This is very similar to time series kernels used for multivariate time series classification problems. We test mv-ARF on MTS classification problems and show that mv-ARF provides fast and competitive results on benchmark datasets from several domains. Furthermore, mv-ARF provides a research direction for vector autoregressive models that breaks from the linear dependency models to potentially foster other promising nonlinear approaches.

Alternative Models for Forecasting Macroeconomic Indicators in Kazakhstan

This paper indicates special aspects of using vector auto-regression models to forecast rates of basic macroeconomic indicators in short term. In particular, traditional vector auto-regression model, Bayesian vector auto-regression model and factor augmented vector auto-regression model are shown. For parameter estimation of these models the author uses time series of Kazakhstani macroeconomic indicators between 1996 and 2015 quarterly. In virtue of mean-root-square error prediction the conclusion of optimal model is going to be chosen.

Effects of Stock Indices of Developed and Emerging Markets on Economic Activity in Colombia: a FAVAR Approach

Analizamos el efecto de las condiciones financieras de economias emergentes y desarrolladas sobre el comportamiento de la economia colombiana. Para ello, estimamos indices de condiciones financieras para mercados accionarios de paises desarrollados y emergentes usando analisis de componentes principales. Posteriormente, estos indices se incluyeron como regresores en un modelo tradicional de vectores auto-regresivos (VAR); es decir, estimamos un modelo de vectores auto-regresivos aumentado con factores (FAVAR) que incluye un factor no observable y una variable observable de la actividad economica real con el fin de evaluar el efecto de las condiciones financieras internacionales sobre las variables macroeconomicas colombianas. De esta manera, encontramos que las condiciones accionarias de los paises emergentes tienen un efecto positivo sobre el comportamiento macroeconomico de Colombia cuando se lleva a cabo el analisis con datos mensuales.

The Effect of Commuting Costs to Employment Centers on Urban Property Values: A Spatial Analysis in Bogott, Colombia

We study the effect that changes of commuting costs to intra urban employment centers may have on property prices in Bogota, Colombia. To explore the relationship between prices and communing costs, we use a difference in difference spatial model. Parametric and non-parametric estimations suggest a negative and significant impact of commuting costs on property values for dwellings far away from employment centers. The estimated price elasticity with respect to commuting costs for housing located in distant zones, indicates that a 1% increase in fuel price, reduces property values in 1.48%. Non-parametric Geographically Weighted Regression (GWR) showed to outperform traditional Spatial Autoregressive (SAR, SEM) models.

An EMD-recursive ARIMA method to predict wind speed for railway strong wind warning system

To protect running trains against the strong crosswind along Chinese Qinghai–Tibet railway, a strong wind warning system is developed. As one of the most important technologies of the developed system, a new short-term wind speed forecasting method is proposed by adopting the Empirical Mode Decomposition (EMD) and the improved Recursive Autoregressive Integrated Moving Average (RARIMA) model. The proposed forecasting method consists of three computational steps as: (a) use the EMD method to decompose the original wind speed data into a group of wind speed sub-layers; (b) build the forecasting models for all the decomposed sub-layers by utilizing the RARIMA algorithm; (c) employ the built RARIMA models to predict the wind speed in the sub-layers; and (d) summarize the predicted results of the wind speed sub-layers to get the final forecasting results for the original wind speed. Since the wind speed forecasting method is proposed for the real-time warning system, the forecasting accuracy and the time performance of the forecasting computation are both considered. Two experiments show that: (a) the proposed method has better forecasting performance than the traditional Autoregressive Integrated Moving Average (ARIMA) model, the Persistent Random Walk Model (PRWM) and the Back Propagation (BP) neural networks; and (b) the proposed method has satisfactory performance in both of the accuracy and the time performance.

Modeling the price of trends of teak wood using statistical and artificial neural network techniques

Modeling the trends and patterns in financial data is of great interest to the business community to support the decision-making process. In this study, the historical trends in the real prices of teak wood were described using spline models and the time periods for which the rate of change in real prices differed were identified. The possible reasons for this phenomenon such as impact of forest legislations and other factors have been explained. In forecasting teak wood prices, Artificial Neural Network (ANN) was compared with the traditional Auto Regressive Integrated Moving Average (ARIMA) model. The Mean Absolute Percentage Error (MAPE) was lesser in the case of ANN than the ARIMA model. Further, the turning points were more closely predicted by ANN. It appeared that forecast by ANN was heavily dependent on the previous value(s) immediate to the forecasting year. The study concluded that the next year price forecasts by univariate ARIMA and ANN models may be far from actual prices due to unanticipated factors.

An integrated wavelet-support vector machine for groundwater level prediction in Visakhapatnam, India

Accurate and reliable prediction of the groundwater level variation is significant and essential in water resources management of a basin. The situation is complicated by the fact that the variation of groundwater level is highly nonlinear in nature because of interdependencies and uncertainties in the hydro-geological process. Models such as Artificial Neural Networks (ANN) and Support Vector Machine (SVM) have proved to be effective in modeling virtually any nonlinear function with a greater degree of accuracy. In recent times, combining several techniques to form a hybrid tool to improve the accuracy of prediction has become a common practice for various applications. This integrated method increases the efficiency of the model by combining the unique features of the constituent models to capture different patterns in the data. In the present study, an attempt is made to predict monthly groundwater level fluctuations using integrated wavelet and support vector machine modeling. The discrete wavelet transform with two coefficients (db2 wavelet) is adopted for decomposing the input data into wavelet series. These series are further used as input variables in different combinations for Support Vector Regression (SVR) model to forecast groundwater level fluctuations. The monthly data of precipitation, maximum temperature, mean temperature and groundwater depth for the period 2001–2012 are used as the input variables. The proposed Wavelet-Support Vector Regression (WA-SVR) model is applied to predict the groundwater level variations for three observation wells in the city of Visakhapatnam, India. The performance of the WA-SVR model is compared with SVR, ANN and also with the traditional Auto Regressive Integrated Moving Average (ARIMA) models. Results indicate that WA-SVR model gives better accuracy in predicting groundwater levels in the study area when compared to other models.

The effects of monetary policy using structural factor analysis

The traditional Vector Autoregression (VAR) method is widely used to trace out the effects of monetary policy innovations on the economy. However, this method suffers from the curse of dimensionality, so that in practice VARs are estimated on a limited number of variables, leading to a potential missing information problem. In this article we use the method of structural factor analysis to evaluate the effects of monetary policy on key macroeconomic variables in a data rich environment. This methodology allows us to extract information on monetary policy and its impact on the economy from a much larger data set than is possible with the traditional VAR method. We propose two structural factor models. One is the Structural Factor Augmented Vector Autoregressive (SFAVAR) model and the other is the Structural Factor Vector Autoregressive (SFVAR) model. Compared to the traditional VAR, both models incorporate information from hundreds of data series, series that can be and are monitored by the central bank in setting policy. Moreover, the factors used are structurally meaningful, a feature that adds to the understanding of the ‘black box’ of the monetary transmission mechanism. Both models generate qualitatively reasonable impulse response functions. For the SFVAR model, both the price puzzle and the liquidity puzzle are eliminated.

Technology Forecasting Via Published Patent Applications and Patent Grants

The objective of this research is to study and establish the relationships between published patent applications and patent grants for exploring the technology development trend on a specific technology since more and more patents have their applications published before they are granted. Two modeling algorithms based on the patent grant/publish ratio as well as one long-term modeling algorithm based on the average publish-to-grant lag, were developed accordingly. The relationships between patent grants and published patent applications were constructed through two case studies on Magnetic Random Access Memory and Organic Light-Emitting Diode technologies and corresponding forecasts were then conducted. Comparing to the traditional time-series Autoregressive Integrated Moving Average method, the predicting power of the modeling algorithms based on the patent grant/publish ratio was satisfactory. On the other hand, the modeling algorithm based on the characteristic of average publish-to-grant lag has shown superior predicting power. Results from these two applications help us to validate the proposed methods and appropriate tools for forecasting the patent grants.

Monetary Policy Analysis Based on Lasso-Assisted Vector Autoregression (Lavar)

Measuring the quantitative effects of monetary policy on the economy has been playing a central role in promoting economic growth and stability. However, in the presence of numerous macroeconomic variables, traditional vector autoregression (VAR) could only accommodate a few data series, and thus may ignore the information set which is actually observed by central banks and financial market participants. In this paper, we propose a novel VAR model with the aid of new developments in high-dimensional statistical inference. Our approach could handle hundreds of observed data series simultaneously, and increase the prediction accuracy as well as the robustness of monetary policy analysis in a data-rich environment. It has been shown that our model outperforms factor-augmented VAR in terms of in-sample-fit and out-of-sample forecasts. Moreover, impulse responses are observed for all macroeconomic variables, where “price puzzle”, a commonly observed empirical anomaly, is resolved.

Conditional Heteroscedasticity in Streamflow Process: Paradox or Reality?

The various physical mechanisms governing the dynamics of streamflow processes act on a seemingly wide range of temporal and spatial scales; almost all the mechanisms involved present some degree of nonlinearity. Against the backdrop of these issues, in this paper, attempt was made to critically look at the subject of Autoregressive Conditional Heteroscedasticity (ARCH) or volatility of streamflow processes, a form of nonlinear phenomena. Towards this end, streamflow data (both daily and monthly) of the River Benue, Nigeria were used for the study. Results obtained from the analyses indicate that the existence of conditional heteroscedasticity in streamflow processes is no paradox. Too, ARCH effect is caused by seasonal variation in the variance for monthly flows and could partly explain same in the daily streamflow. It was also evident that the traditional seasonal Autoregressive Moving Average (ARMA) models are inadequate in describing ARCH effect in daily streamflow process though, robust for monthly streamflow; and can be removed if proper deseasonalisation pre-processing was done. Considering the findings, the potential for a hybrid Autoregressive Moving Average (ARMA) and Generalised Autoregressive Conditional Heteroscedasticity (GARCH)type models should be further explored and probably embraced for modelling daily streamflow regime in view of the relevance of statistical modelling in hydrology.

Basket trading under co-integration with the logistic mixture autoregressive model

In this paper, we propose a co-integration model with a logistic mixture auto-regressive equilibrium error (co-integrated LMAR), in which the equilibrium relationship among cumulative returns of different financial assets is modelled by a logistic mixture autoregressive time series model. The traditional autoregression (AR) based unit root test (ADF test), used in testing co-integration, cannot give a sound explanation when a time series passes the ADF test. However, its largest root in the AR polynomial is extremely close to, but less than, one, which is most likely the result of a mixture of random-walk and mean-reverting processes in the time series data. With this background, we put an LMAR model into the co-integration framework to identify baskets that have a large spread but are still well co-integrated. A sufficient condition for the stationarity of the LMAR model is given and proved using a Markovian approach. A two-step estimating procedure, combining least-squares estimation and the Expectation-Maximization (EM) algorithm, is given. The Bayesian information criterion (BIC) is used in model selection. The co-integrated LMAR model is applied to basket trading, which is a widely used tool for arbitrage. We use simulation to assess the model in basket trading strategies with the statistical arbitrage feature in equity markets. Data from several sectors of the Hong Kong Hang Seng Index are used in a simulation study on basket trading. Empirical results show that a portfolio using the co-integrated LMAR model has a higher return than portfolios selected by traditional methods. Although the volatility in the return increases, the Sharpe ratio also increases in most cases. This risk–return profile can be explained by the shorter converging period in the co-integrated LMAR model and the larger volatility in the ‘mean-reverting’ regime.