Crude oil price analysis and forecasting using wavelet decomposed ensemble model

To improve the forecasting accuracy of crude oil price with deeper understanding of the market microstructure, this paper proposes a wavelet decomposed ensemble model. The proposed model follows the Heterogeneous Market Hypothesis that assumes the unstationarity and dynamic changing nature of the underlying market structure and introduces the wavelet analysis to analyze the dynamic underlying Data Generating Process at finer time scale domain. The simple averaging based ensemble model is introduced to reduce the estimation bias resulting from the use of different wavelet families by deriving market consensus view. The ensemble members are selected dynamically based on their in-sample performance among forecast matrices based on different wavelet families. Results from empirical studies show the superior performance of the proposed algorithm against the benchmark models, in terms of both level and directional predictive accuracy. The proposed model can effectively extract and model the time varying heterogeneous market microstructure, whose accurate characterization results in further improvement in market analysis and predictability.