Nonlinear Forecasting of Noisy Financial Data

Soofi, Abdol S.; Cao, Liangyue

doi:10.1007/978-1-4615-0931-8_22

Cited by 15 publications

(12 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the SSA technique relies on the Singular Value Decomposition (SVD) approach for noise reduction, which is regarded as a more effective noise reduction tool in comparison to standard filtering techniques which decompose series in different frequencies (Soofi and Cao, 2002;Ortu et al, 2013). Furthermore, unlike local methods, such as linear filtering or wavelets, or even the HW, the SSA exploits the trajectory matrix computed using all parts of a time series (Alexandrov, 2009).…”

Section: Methodology and Iv-ssa-hw Modelmentioning

confidence: 99%

Forecasting global stock market implied volatility indices

Degiannakis

Filis

Hassani

2018

Journal of Empirical Finance

View full text Add to dashboard Cite

This study compares parametric and non-parametric techniques in terms of their forecasting power on implied volatility indices. We extend our comparisons using combined and model-averaging models. The forecasting models are applied on eight implied volatility indices of the most important stock market indices. We provide evidence that the non-parametric models of Singular Spectrum Analysis combined with Holt-Winters (SSA-HW) exhibit statistically superior predictive ability for the one and ten trading days ahead forecasting horizon. By contrast, the model-averaged forecasts based on both parametric (Autoregressive Integrated model) and non-parametric models (SSA-HW) are able to provide improved forecasts, particularly for the ten trading days ahead forecasting horizon. For robustness purposes, we build two trading strategies based on the aforementioned forecasts, which further confirm that the SSA-HW and the ARI-SSA-HW are able to generate significantly higher net daily returns in the out-of-sample period.

show abstract

Section: Methodology and Iv-ssa-hw Modelmentioning

confidence: 99%

Forecasting global stock market implied volatility indices

Degiannakis

Filis

Hassani

2018

Journal of Empirical Finance

View full text Add to dashboard Cite

show abstract

“…This strategy resembles the filtered historical simulation (FHS) in that Studentized observations are used to obtain q ε (α) (see Barone-Adesi et al 1998or McNeil & Frey, 2000. Third, b v tjn , t ¼ 1, 2, …,n can be used as the base to predict v n + h with standard prediction techniques, avoiding in that way the distorting effect of the added noise (see for example Soofi & Cao, 2002). More details are given in Section 4.…”

Section: Var In Sv Modelsmentioning

confidence: 99%

“…First, noting that the optimal predictor of v n + h in (2) coincides with the optimal predictor of y n + h À μ, traditional forecasting techniques can be implemented in the centered sample of observables y 1 , … , y n to predict y n + h À μ and use this prediction as the forecast of v n + h . However, the added noise may cause significant differences between the empirical prediction and v n + h (see, for example, Soofi & Cao, 2002). To avoid that potentially distorting effect of the added noise, the second approach consists of predicting v n + h by applying the forecasting strategies on the estimated v t , t ¼ 1,…, n obtained by the application of some signal extraction technique.…”

Section: Out-of-sample Prediction: Forecasting Techniquesmentioning

confidence: 99%

“…Then exponential smoothing (ETS) and neural network (NN) as implemented in the package forecast in R are also implemented. The application of these four techniques on the series of observables y * t and on the signal extracted using SSA b v ssa tjn permits us to assess if the added noise exerts a negative effect in the prediction of the signal as advocated by Soofi and Cao (2002) or if, on the contrary, the error generated in the extraction of the signal is of such a magnitude that using y * t is a better option. In order to distinguish notationally both strategies, a hat is used to denote predictions that have been obtained by applying the corresponding technique to b v ssa tjn , whereas a tilde is used if it has been applied to the series of observables y * t .…”

Section: Out-of-sample Prediction: Forecasting Techniquesmentioning

confidence: 99%

See 1 more Smart Citation

Singular spectrum analysis for value at risk in stochastic volatility models

Arteche

García‐Enríquez

2021

Journal of Forecasting

View full text Add to dashboard Cite

Estimation of the value at risk (VaR) requires prediction of the future volatility. Whereas this is a simple task in ARCH and related models, it becomes much more complicated in stochastic volatility (SV) processes where the volatility is a function of a latent variable that is not observable. In-sample (present and past values) and out-of-sample (future values) predictions of that unobservable variable are thus necessary. This paper proposes singular spectrum analysis (SSA), which is a fully nonparametric technique that can be used for both purposes. A combination of traditional forecasting techniques and SSA is also considered to estimate the VaR. Their performance is assessed in an extensive Monte Carlo and with an application to a daily series of S&P500 returns.

show abstract

“…However, not much work has been done [4]. In this study, we employ two nonlinear noise reduction methods to alleviate these problems.…”

Section: Introductionmentioning

confidence: 99%

Application of Neural Networks for Foreign Exchange Rates Forecasting with Noise Reduction

Huang

Lai

Wang

2007

Computational Science – ICCS 2007

View full text Add to dashboard Cite

Abstract. Predictive models are generally fitted directly from the original noisy data. It is well known that noise can seriously limit the prediction performance on time series. In this study, we apply the nonlinear noise reduction methods to the problem of foreign exchange rates forecasting with neural networks (NNs). The experiment results show that the nonlinear noise reduction methods can improve the prediction performance of NNs. Based on the modified DieboldMariano test, the improvement is not statistically significant in most cases. We may need more effective nonlinear noise reduction methods to improve prediction performance further. On the other hand, it indicates that NNs are particularly well appropriate to find underlying relationship in the environment characterized by complex, noisy, irrelevant or partial information. We also find that the nonlinear noise reduction methods work more effectively when the foreign exchange rates are more volatile.

show abstract

Nonlinear Forecasting of Noisy Financial Data

Cited by 15 publications

References 11 publications

Forecasting global stock market implied volatility indices

Forecasting global stock market implied volatility indices

Singular spectrum analysis for value at risk in stochastic volatility models

Application of Neural Networks for Foreign Exchange Rates Forecasting with Noise Reduction

Contact Info

Product

Resources

About