A Combinatorial Approach to Piecewise Linear Time Series Analysis

Over recent years, several nonlinear time series models have been proposed in the literature. One model that has found a large number of successful applications is the threshold autoregressive model (TAR). The TAR model is a piecewise linear process whose central idea is to change the parameters of a linear autoregressive model according to the value of an observable variable, called the threshold variable. If this variable is a lagged value of the time series, the model is called a self-exciting threshold autoregressive (SETAR) model. In this article, we propose a heuristic to estimate a more general SETAR model, where the thresholds are multivariate. We formulate the task of finding multivariate thresholds as a combinatorial optimization problem. We develop an algorithm based on a greedy randomized adaptive search procedure (GRASP) to solve the problem. GRASP is an iterative randomized sampling technique that has been shown to quickly produce good quality solutions for a wide variety of optimization problems. The proposed model performs well on both simulated and real data.     

Multi-step prediction of time series with random missing data

Multi-step prediction is still an open challenge in time series prediction. Moreover, practical observations are often incomplete because of sensor failure or outliers causing missing data. Therefore, it is very important to carry out research on multi-step prediction of time series with random missing data. Based on nonlinear filters and multilayer perceptron artificial neural networks (ANNs), one novel approach for multi-step prediction of time series with random missing data is proposed in the study. With the basis of original nonlinear filters which do not consider the missing data, first we obtain the generalized nonlinear filters by using a sequence of independent Bernoulli random variables to model random interruptions. Then the multi-step prediction model of time series with random missing data, which can be fit for the online training of generalized nonlinear filters, is established by using the ANN’s weights to present the state vector and the ANN’s outputs to present the observation equation. The performance between the original nonlinear filters based ANN model for multi-step prediction of time series with missing data and the generalized nonlinear filters based ANN model for multi-step prediction of time series with missing data is compared. Numerical results have demonstrated that the generalized nonlinear filters based ANN are proportionally superior to the original nonlinear filters based ANN for multi-step prediction of time series with missing data.     

Forecasting time series with multiple seasonal patterns

A new approach is proposed for forecasting a time series with multiple seasonal patterns. A state space model is developed for the series using the innovations approach which enables us to develop explicit models for both additive and multiplicative seasonality. Parameter estimates may be obtained using methods from exponential smoothing. The proposed model is used to examine hourly and daily patterns in hourly data for both utility loads and traffic flows. Our formulation provides a model for several existing seasonal methods and also provides new options, which result in superior forecasting performance over a range of prediction horizons. In particular, seasonal components can be updated more frequently than once during a seasonal cycle. The approach is likely to be useful in a wide range of applications involving both high and low frequency data, and it handles missing values in a straightforward manner.     

Extended and Unscented Kalman filtering based feedforward neural networks for time series prediction

With the ability to deal with high non-linearity, artificial neural networks (ANNs) and support vector machines (SVMs) have been widely studied and successfully applied to time series prediction. However, good fitting results of ANNs and SVMs to nonlinear models do not guarantee an equally good prediction performance. One main reason is that their dynamics and properties are changing with time, and another key problem is the inherent noise of the fitting data. Nonlinear filtering methods have some advantages such as handling additive noises and following the movement of a system when the underlying model is evolving through time. The present paper investigates time series prediction algorithms by using a combination of nonlinear filtering approaches and the feedforward neural network (FNN). The nonlinear filtering model is established by using the FNN’s weights to present state equation and the FNN’s output to present the observation equation, and the input vector to the FNN is composed of the predicted signal with given length, then the extended Kalman filtering (EKF) and Unscented Kalman filtering (UKF) are used to online train the FNN. Time series prediction results are presented by the predicted observation value of nonlinear filtering approaches. To evaluate the proposed methods, the developed techniques are applied to the predictions of one simulated Mackey-Glass chaotic time series and one real monthly mean water levels time series. Generally, the prediction accuracy of the UKF-based FNN is better than the EKF-based FNN when the model is highly nonlinear. However, comparing from prediction accuracy and computational effort based on the prediction model proposed in our study, we draw the conclusion that the EKF-based FNN is superior to the UKF-based FNN for the theoretical Mackey-Glass time series prediction and the real monthly mean water levels time series prediction.     

Forecasting model with asymmetric market response and its application to pricing of consumer package goods

This paper presents a dynamic forecasting model that accommodates asymmetric market responses to marketing mix variable—price promotion—by threshold models. As a threshold variable to generate a mechanism for different market responses, we use the counterpart to the concept of a price threshold applied to a representative consumer in a store. A Bayesian approach is taken for statistical modelling because of advantages that it offers over estimation and forecasting. The proposed model incorporates the lagged effects of a price variable. Thereby, myriad pricing strategies can be implemented in the time horizon. Their effectiveness can be evaluated using the predictive density. We intend to improve the forecasting performance over conventional linear time series models. Furthermore, we discuss efficient dynamic pricing in a store using strategic simulations under some scenarios suggested by an estimated structure of the models. Empirical studies illustrate the superior forecasting performance of our model against conventional linear models in terms of the root mean square error of the forecasts. Useful information for dynamic pricing is derived from its structural parameter estimates. This paper develops a dynamic forecasting model that accommodates asymmetric market responses to marketing mix variable—price promotion—by the threshold models. Copyright © 2005 John Wiley & Sons, Ltd.     

Identification of a nonlinear circuit

The identification of a differential equation underlying a measured time series is a prerequisite for numerous types of applications. In the validation of a proposed parameterized model one often faces the dilemma that it is hard to decide whether possible discrepancies between the measured time series and the simulated model output are caused by an inappropriate model or by wrongly specified parameters in a correct type of model. We propose a combination of parametric modelling based on Bock's multiple shooting algorithm and nonparametric modelling based on optimal transformations as a strategy to test proposed models and if rejected suggest and test new ones. We exemplify this strategy on an experimental time series from a nonlinear chaotically oscillating circuit where we finally obtain an extremely accurate reconstruction of the observed attractor.     

Forecasting nonlinear time series with a hybrid methodology

In recent years, artificial neural networks (ANNs) have been used for forecasting in time series in the literature. Although it is possible to model both linear and nonlinear structures in time series by using ANNs, they are not able to handle both structures equally well. Therefore, the hybrid methodology combining ARIMA and ANN models have been used in the literature. In this study, a new hybrid approach combining Elman’s Recurrent Neural Networks (ERNN) and ARIMA models is proposed. The proposed hybrid approach is applied to Canadian Lynx data and it is found that the proposed approach has the best forecasting accuracy.     

Stress testing network reconstruction via graphical causal model

An resilience optimal evaluation of financial portfolios implies having plausible hypotheses about the multiple interconnections between the macroeconomic variables and the risk parameters. In this article, we propose a graphical model for the reconstruction of the causal structure that links the multiple macroeconomic variables and the assessed risk parameters, it is this structure that we call stress testing network. In this model, the relationships between the macroeconomic variables and the risk parameter define a “relational graph” among their time‐series, where related time‐series are connected by an edge. Our proposal is based on the temporal causal models, but unlike, we incorporate specific conditions in the structure which correspond to intrinsic characteristics this type of networks. Using the proposed model and given the high‐dimensional nature of the problem, we used regularization methods to efficiently detect causality in the time‐series and reconstruct the underlying causal structure. In addition, we illustrate the use of model in credit risk data of a portfolio. Finally, we discuss its uses and practical benefits in stress testing.     

On the construction of a nonlinear recursive predictor

In this paper, we present a novel approach for constructing a nonlinear recursive predictor. Given a limited time series data set, our goal is to develop a predictor that is capable of providing reliable long-term forecasting. The approach is based on the use of an artificial neural network and we propose a combination of network architecture, training algorithm, and special procedures for scaling and initializing the weight coefficients. For time series arising from nonlinear dynamical systems, the power of the proposed predictor has been successfully demonstrated by testing on data sets obtained from numerical simulations and actual experiments.     

一种改进的基于指数平滑神经网络模型的时间序列预测方法

基于指数平滑模型与误差反传神经网络法提出了一个改进的时间序列预测方法.将神经网络模型移植入指数加权滑动平均模型中,充分考虑了时间序列的部分线性性和非线性性对预测结果的影响,是传统的混合模型的一个更合理的改进.最后通过对上证指数时间序列的实证分析,以预测均方误差为检验标准,对五种常用的时间序列预测模型进行了预测精度的比较,而且经验证所提出的改进的时间序列预测模型相对来说具有更小的预测均方误差.     

Time series forecasting with neural network ensembles: an application for exchange rate prediction

This paper investigates the use of neural network combining methods to improve time series forecasting performance of the traditional single keep-the-best (KTB) model. The ensemble methods are applied to the difficult problem of exchange rate forecasting. Two general approaches to combining neural networks are proposed and examined in predicting the exchange rate between the British pound and US dollar. Specifically, we propose to use systematic and serial partitioning methods to build neural network ensembles for time series forecasting. It is found that the basic ensemble approach created with non-varying network architectures trained using different initial random weights is not effective in improving the accuracy of prediction while ensemble models consisting of different neural network structures can consistently outperform predictions of the single ‘best’ network. Results also show that neural ensembles based on different partitions of the data are more effective than those developed with the full training data in out-of-sample forecasting. Moreover, reducing correlation among forecasts made by the ensemble members by utilizing data partitioning techniques is the key to success for the neural ensemble models. Although our ensemble methods show considerable advantages over the traditional KTB approach, they do not have significant improvement compared to the widely used random walk model in exchange rate forecasting.     

Evolutionary learning of fuzzy grey cognitive maps for the forecasting of multivariate,interval-valued time series

Time series are built as a result of real-valued observations ordered in time; however, in some cases, the values of the observed variables change significantly, and those changes do not produce useful information. Therefore, within defined periods of time, only those bounds in which the variables change are considered. The temporal sequence of vectors with the interval-valued elements is called a ‘multivariate interval-valued time series.’ In this paper, the problem of forecasting such data is addressed. It is proposed to use fuzzy grey cognitive maps (FGCMs) as a nonlinear predictive model. Using interval arithmetic, an evolutionary algorithm for learning FGCMs is developed, and it is shown how the new algorithm can be applied to learn FGCMs on the basis of historical time series data. Experiments with real meteorological data provided evidence that, for properly-adjusted learning and prediction horizons, the proposed approach can be used effectively to the forecasting of multivariate, interval-valued time series. The domain-specific interpretability of the FGCM-based model that was obtained also is confirmed.     

Piecewise finite series solution of nonlinear initial value differential problem

In this paper, we apply a piecewise finite series as a hybrid analytical-numerical technique for solving some nonlinear systems of ordinary differential equations. The finite series is generated by using the Adomian decomposition method, which is an analytical method that gives the solution based on a power series and has been successfully used in a wide range of problems in applied mathematics. We study the influence of the step size and the truncation order of the piecewise finite series Adomian (PFSA) method on the accuracy of the solutions when applied to nonlinear ODEs. Numerical comparisons between the PFSA method with different time steps and truncation orders against Runge-Kutta type methods are presented. Based on the numerical results we propose a low value truncation order approach with small time step size. The numerical results show that the PFSA method is accurate and easy to implement with the proposed approach.     

Fuzzy regression analysis using neural networks

In this paper, we propose simple but powerful methods for fuzzy regression analysis using neural networks. Since neural networks have high capability as an approximator of nonlinear mappings, the proposed methods can be applied to more complex systems than the existing LP based methods. First we propose learning algorithms of neural networks for determining a nonlinear interval model from the given input-output patterns. A nonlinear interval model whose outputs approximately include all the given patterns can be determined by two neural networks. Next we show two methods for deriving nonlinear fuzzy models from the interval model determined by the proposed algorithms. Nonlinear fuzzy models whose h-level sets approximately include all the given patterns can be derived. Last we show an application of the proposed methods to a real problem.     

基于C-STGARCH模型的大庆原油现货价格波动研究

引用Dueker等(2011)提出的同期门槛平滑转换广义自回归条件异方差(C-STGARCH)模型对我国大庆原油现货价格的波动状态进行了实证分析,以求对大庆原油现货价格的波动有一个新的、更深刻的量度.研究显示:第一,大庆原油现货价格的波动是不稳定的,并且存在显著的非对称和非线性现象;第二,CSTGARCH模型能很好地刻画大庆原油现货价格波动的这些现象,并且发现油价的波动以3.738%为门槛点存在高波动区和低波动区两种状态,低波动区的波动持续性比高波动区强,然而,对平滑转换持续性的影响方面,高波动区要略大于低波动区.     

Bayesian Semiparametric Regression Analysis of Multicategorical Time-Space Data

We present a unified semiparametric Bayesian approach based on Markov random field priors for analyzing the dependence of multicategorical response variables on time, space and further covariates. The general model extends dynamic, or state space, models for categorical time series and longitudinal data by including spatial effects as well as nonlinear effects of metrical covariates in flexible semiparametric form. Trend and seasonal components, different types of covariates and spatial effects are all treated within the same general framework by assigning appropriate priors with different forms and degrees of smoothness. Inference is fully Bayesian and uses MCMC techniques for posterior analysis. The approach in this paper is based on latent semiparametric utility models and is particularly useful for probit models. The methods are illustrated by applications to unemployment data and a forest damage survey.     

Improving demand forecasting accuracy using nonlinear programming software

We address the problem of forecasting real time series with a proportion of zero values and a great variability among the nonzero values. In order to calculate forecasts for a time series, the model coefficients must be estimated. The appropriate choice of values for the smoothing parameters in exponential smoothing methods relies on the minimization of the fitting errors of historical data. We adapt the generalized Holt–Winters formulation so that it can consider the starting values of the local components of level, trend and seasonality as decision variables of the nonlinear programming problem associated with this forecasting procedure. A spreadsheet model is used to solve the problems of optimization efficiently. We show that our approach produces accurate forecasts with little data per product.     

A partially linearized sigma point filter for latent state estimation in nonlinear time series models

A new technique for the latent state estimation of a wide class of nonlinear time series models is proposed. In particular, we develop a partially linearized sigma point filter in which random samples of possible state values are generated at the prediction step using an exact moment-matching algorithm and then a linear programming based procedure is used in the update step of the state estimation. The effectiveness of the new filtering procedure is assessed via a simulation example that deals with a highly nonlinear, multivariate time series representing an interest rate process.     

稀疏VAR在股票收益率研究的应用

向量自回归模型(VAR)广泛应用在对时间相依的多元时间序列建模中,但在高维数据建模中,自回归的系数膨胀可能导致噪音估计、不稳定的预测、解释上的困难等问题。在实际应用中,序列的真实模型往往具有稀疏性,因此运用稀疏VAR模型对高维时间序列进行建模,不仅可以解决高维数据带来的上述困难,也有利于寻找高维数据内在的真实模型。本文以10家公司的股票收益率为研究对象,采用3种不同的稀疏估计方法,不但分析了股票收益率之间的动态关系,而且通过实证分析展示了稀疏估计的优势。     

Multi-population mortality models: A factor copula approach

Modeling mortality co-movements for multiple populations have significant implications for mortality/longevity risk management. A few two-population mortality models have been proposed to date. They are typically based on the assumption that the forecasted mortality experiences of two or more related populations converge in the long run. This assumption might be justified by the long-term mortality co-integration and thus be applicable to longevity risk modeling. However, it seems too strong to model the short-term mortality dependence. In this paper, we propose a two-stage procedure based on the time series analysis and a factor copula approach to model mortality dependence for multiple populations. In the first stage, we filter the mortality dynamics of each population using an ARMA–GARCH process with heavy-tailed innovations. In the second stage, we model the residual risk using a one-factor copula model that is widely applicable to high dimension data and very flexible in terms of model specification. We then illustrate how to use our mortality model and the maximum entropy approach for mortality risk pricing and hedging. Our model generates par spreads that are very close to the actual spreads of the Vita III mortality bond. We also propose a longevity trend bond and demonstrate how to use this bond to hedge residual longevity risk of an insurer with both annuity and life books of business.