MULTIFRACTAL ANALYSIS WITH DETRENDING WEIGHTED AVERAGE ALGORITHM OF HISTORICAL VOLATILITY
Abstract
In this paper, we develop the multifractal detrending weighted average algorithm of historical volatility (MF-DHV) for one-dimensional multifractal measure based on the classical multifractal detrended fluctuation analysis (MF-DFA). In the calculation process of getting a local trend for MF-DHV, historical volatility is taken to develop an moving average algorithm, which is different from the simple moving average function in multifractal detrended moving average (MF-DMA). We assess the performance of three methods such as MF-DFA, MF-DMA, and MF-DHV based on the p-model multiplicative cascading constructed time series. The computational results show that all the estimated generalized Hurst exponent H(q), the scaling exponent τ(q), and the singularity spectrum f(α) of MF-DHV are in good agreement with the theoretical values. In addition, we also calculate the standard deviations of Herr and τerr for three methods, and the lowest errors in MF-DHV provides the most accurate estimates. To avoid the accidental selection of parameters, we change the total length of the generated multifractal simulation data and p-value, respectively. It is found that in all the cases, the MF-DHV outperforms the other two methods.
