The Research Of Complexity Testing Methods And The Applications In Energy Markets

Time series data analysis and forecasting are important issues within research fields of data mining and management science. Capturing the data characteristics of time series data is the foundation for further research. Complexity as one of the most important measurements for analyzing time series data, it covers or is at least closely related to different data characteristics within nonlinear system theory, meanwhile, complexity can reflect the property of the research object. Recently, with the deterioration of the environment and climate, how to efficiently analyze the characteristics of time series and offer corresponding management plan becomes a hot issue. Based on a series of complexity indices, this paper analyses the property of some energy markets.Firstly, according to the existing complexity testing methods, this paper proposes a framework about complexity testing, which covers the self-similarity (long term persistence) of times series, the property of attractor in the phase space and the disorder state of the data dynamics. Based on the framework, the multi-fractal detrended fluctuation analysis, correlation dimension, and sample entropy are respectively conducted, then the entropy weight method is then used to generate a new integrated index. Meanwhile, the proposed measure is applied to three typical energy markets analyses, including natural gas, crude oil and carbon market. The empirical results find that natural gas market and crude oil market (two comparatively mature and competitive markets) are much more efficient than carbon market (an emerging market). Compared with natural gas market, the crude oil market is tested at a relatively lower level of market efficiency due to the sensitiveness to some extra factors. Secondly, considering the impact of time scales on complexity, this paper further proposed the complexity exploration based on multi-scale analysis. In the proposed approach, ensemble empirical mode decomposition (EEMD), is first implemented to decompose the time series data into a set of independent meaningful components with different time scales for further analyses. Second, fuzzy entropy, is applied to estimate the complexity characteristics of both the whole clean energy system and the different hidden features. The nuclear energy and hydropower markets in China and USA are analyzed via the proposed method, and we find the complexity levels of the clean energy markets in USA are significantly higher than those of China.Finally, we further analyze the cross-correlation between two time series derived from the complexity testing of one time series. The recent multifractal asymmetric detrended cross-correlation analysis method is adopted to investigate the cross-correlation between the WTI crude oil and financial markets (US dollar index and gold markets), besides, the time varying multifractal asymmetric cross-correlations under different time scales are considered. The empirical results indicate that both the cross-correlations between WTI crude oil market and US dollar index market and the cross-correlations between WTI crude oil market and gold market are asymmetric and multifractal, and the cross-correlations between the WTI crude oil market and the US dollar index market are more persistent whether the trending of the US dollar index market is either positive or negative.