Reducing overestimating and underestimating volatility via the augmented blending-ARCH model

TL;DR

This paper introduces augmented and blending ARCH models to improve volatility forecasting in financial time series, addressing issues of overestimation and underestimation inherent in SVR-GARCH models.

Contribution

The paper proposes novel augmented and blending ARCH models that enhance peak and trough prediction accuracy in volatility forecasting.

Findings

Improved volatility prediction accuracy on SH300 and S&P500 datasets.

Enhanced detection of peaks and troughs in financial time series.

Models outperform traditional SVR-GARCH in empirical tests.

Abstract

SVR-GARCH model tends to "backward eavesdrop" when forecasting the financial time series volatility in which case it tends to simply produce the prediction by deviating the previous volatility. Though the SVR-GARCH model has achieved good performance in terms of various performance measurements, trading opportunities, peak or trough behaviors in the time series are all hampered by underestimating or overestimating the volatility. We propose a blending ARCH (BARCH) and an augmented BARCH (aBARCH) model to overcome this kind of problem and make the prediction towards better peak or trough behaviors. The method is illustrated using real data sets including SH300 and S&P500. The empirical results obtained suggest that the augmented and blending models improve the volatility forecasting ability.