Kaijian He

Oil Price Forecasting with an EMD-Based Multiscale Neural Network Learning Paradigm

In this study, a multiscale neural network learning paradigm based on empirical mode decomposition (EMD) is proposed for crude oil price prediction. In this learning paradigm, the original price series are first decomposed into various independent intrinsic mode components (IMCs) with a range of frequency scales. Then the internal correlation structures of different IMCs are explored by neural network model. With the neural network weights, some important IMCs are selected as final neural network inputs and some unimportant IMCs that are of little use in the mapping of input to output are discarded. Finally, the selected IMCs are input into another neural network model for prediction purpose. For verification, the proposed multiscale neural network learning paradigm is applied to a typical crude oil price -- West Texas Intermediate (WTI) crude oil spot price prediction.

Estimating VaR in crude oil market: A novel multi-scale non-linear ensemble approach incorporating wavelet analysis and neural network

Facing the complicated non-linear nature of risk evolutions, current risk measurement approaches offer insufficient explanatory power and limited performance. Thus this paper proposes wavelet decomposed non-linear ensemble value at risk (WDNEVaR), a novel semi-parametric paradigm, incorporating both, wavelet analysis and artificial neural network technique to further improve the modeling accuracy and reliability. Wavelet analysis is utilized to capture the multi-scale data characteristics across scales while artificial neural network technique is utilized to reduce estimation biases following non-linear ensemble algorithms. Experiment results in three major markets suggest that the proposed WDNEVaR is superior to more traditional approaches as it provides value at risk (VaR) estimates at higher reliability and accuracy.

A hybrid slantlet denoising least squares support vector regression model for exchange rate prediction

Abstract Despite the active exploration of linear and nonlinear modeling of exchange rates, there is no consensus on the optimal forecasting model other than the traditional random walk and ARMA benchmark models in the literature. Given the increasing recognition of heterogeneous market structure, this paper proposes an alternative Slantlet denoising based hybrid methodology that attempts to incorporate the linear and nonlinear data features. The recently emerging Slantlet analysis is introduced to separate the linear data features as it constructs filters with varying lengths at different scales and has more appealing time localization features than the normal wavelet analysis. Meanwhile, the Least Squares Support Vector Regression (LSSVR) is used to model and correct for the empirical errors nonlinear in nature. As empirical studies were conducted in the Euro exchange rate market, the performance of the proposed algorithm was compared with those of benchmark models including random walk, ARMA and LSSVR models. The proposed algorithm outperforms the benchmark models. More importantly the proposed methodology explores and offers deeper insights as to the underlying data generating process.

Ensemble forecasting of Value at Risk via Multi Resolution Analysis based methodology in metals markets

Subject to shocks worldwide, the metals markets in the era of structural changes and globalization have seen a very competitive and volatile market environment. Proper risk measurement and management in the metals markets are of critical value to the investors belonging to different parts of the economy due to their unique role as important industry inputs to the manufacturing process. Although traditional risk management methodologies have worked in the past, we are now facing the challenge of rapidly changing market conditions. Markets now demand the methodologies that estimate more reliable and accurate VaRs. This paper proposes a Multi Resolution Analysis (MRA) based nonlinear ensemble methodology for Value at Risk Estimates (MRNEVaR). The MRA using wavelet analysis is introduced to analyze the dynamic risk evolution at a finer time scale domain and provide insights into different aspects of the underlying risk evolution. The nonlinear ensemble approach using the artificial neural network technique is introduced to determine the optimal ensemble weights and stabilize the forecasts. Performances of the proposed MRNEVaR and more traditional ARMA-GARCH VaR are evaluated and compared during empirical studies in three major metals markets using Kupiec backtesting and Diebold-Mariano test procedures. Experiment results confirm that VaR estimates produced by MRNEVaR provide superior forecasts that are significantly more reliable and accurate than traditional methods.

Estimating Real Estate Value-at-Risk Using Wavelet Denoising and Time Series Model

As the real estate market develops rapidly and is increasingly securitized, it has become an important investment asset in the portfolio design. Thus the measurement of its market risk exposure has attracted attentions from academics and industries due to its peculiar behavior and unique characteristics such as heteroscedasticity and multi scale heterogeneity in its risk and noise evolution etc. This paper proposes the wavelet denoising ARMA-GARCH approach for measuring the market risk level in the real estate sector. The multi scale heterogeneous noise level is determined in the level dependent manner in wavelet analysis. The autocorrelation and heteroscedasticity characteristics for both data and noises are modeled in the ARMA-GARCH framework. Experiment results in Chinese real estate market suggest that the proposed methodology achieves the superior performance by improving the reliability of VaR estimated upon those from traditional ARMA-GARCH approach.

A Wavelet Based Multi Scale VaR Model for Agricultural Market

Participants in the agricultural industries are subject to significant market risks due to long production lags. Traditional methodology analyzes the risk evolution following a time invariant approach. However, this paper analyzes and proposes wavelet analysis to track risk evolution in a time variant fashion. A wavelet-econometric hybrid model is further proposed for VaR estimates. The proposed wavelet decomposed VaR (WDVaR) is ex-ante in nature and is capable of estimating risks that are multi-scale structured. Empirical studies in major agricultural markets are conducted for both the hybrid ARMA-GARCH VaR and the proposed WDVaR. Experiment results confirm significant performance improvement. Besides, incorporation of time variant risks tracking capability offers additional flexibility for adaptability of the proposed hybrid algorithm to different market environments. WDVaR can be tailored to specific market characteristics to capture unique investment styles, time horizons, etc.

Estimation of Value-at-Risk for Exchange Risk Via Kernel Based Nonlinear Ensembled Multi Scale Model

Risk level in the exchange rate market is dynamically evolving with complicated structures. To further refine the analysis process and achieve more accurate measurement, this paper proposes a novel kernel based nonlinear ensembled multi scale Value at Risk methodology for evaluating the risk level in the exchange rate market. In the proposed algorithm, wavelet analysis is introduced to analyze the multi scale heterogeneous risk structures across different time scales. The Principle Component Analysis is used to extract principle components from the redundant forecast matrixes. Then the support vector regression technique is integrated into the modeling process to nonlinearly ensemble forecast matrixes and produce more stable and accurate results. Taking Euro market as a typical test case, empirical studies employing the proposed algorithm shows the superior performance than benchmark ARMA-GARCH and realized volatility based approaches.

Modeling VaR in Crude Oil Market: A Multi Scale Nonlinear Ensemble Approach Incorporating Wavelet Analysis and ANN

Price fluctuations in the crude oil markets worldwide have attracted significant attentions from both, industries and academics, due to their profound impact on businesses and governments. Proper measurement and management of risks due to unexpected price movements in the markets has been crucial from both, operational and strategic perspectives. However, risk measurements from current approaches offer insufficient explanatory power and performance due to the complicated non-linear nature of risk evolutions. This paper adopts a VaR approach to measure risks and proposes multi-scale non-linear ensemble approaches to model the risk evolutions in WTI crude oil market. The proposed WDNEVaR follows a semi-parametric paradigm, incorporating both, wavelet analysis and artificial neural network techniques. Experiment results from empirical studies suggest that the proposed WDNEVaR is superior to traditional approaches. It provides VaR estimates of higher reliability and accuracy. It also brings significantly more flexibility during the modeling attempts.

Forecasting Crude Oil Price with Multiscale Denoising Ensemble Model

Crude oil price becomes more volatile and sensitive to increasingly diversified influencing factors with higher level of deregulations worldwide. Current methodologies are being challenged as they have been constrained by traditional approaches assuming homogeneous time horizons and investment strategies. Approximations they provided over the long term time horizon no longer satisfy the accuracy requirement at shorter term and more microlevels. This paper proposes a novel crude oil price forecasting model based on the wavelet denoising ARMA models ensemble by least square support vector regression with the reduced forecasting matrix dimensions by independent component analysis. The proposed methodology combines the multi resolution analysis and nonlinear ensemble framework. The wavelet denoising based algorithm is introduced to separate and extract the underlying data components with distinct features, corresponding to investors with different investment scales, which are modeled with time series models of different specifications and parameters. Then least square support vector regression is introduced to nonlinearly ensemble results based on different wavelet families to further reduce the estimation biases and improve the forecasting generalizability. Empirical studies show the significant performance improvement when the proposed model is tested against the bench-mark models.

Crude Oil Price Prediction Using Slantlet Denoising Based Hybrid Models

The accurate prediction of crude oil price movement has always been the central issue with profound implications across different levels of the economy. This study conducts empirical investigations into the characteristics of crude oil market and proposes a novel Slantlet denoising based hybrid methodology for the prediction of its movement. The proposed algorithm models the underlying data characteristics in a more refined manner, integrating linear models such as ARMA and nonlinear models such as Support Vector Regression. Empirical studies confirm the superiority of the proposed Slantlet based hybrid models against benchmark alternatives. The performance improvement is attributed to the finer separation of complicated factors influencing the crude oil behaviors into linear and nonlinear components in the multi scale domain, which improves the goodness of fit and reduces the overfitting issue.