The prediction of crude oil price remains a challenging issue due to its complicated data generating process. Aside from the long perceived nonlinear data feature issue, recent empirical evidence suggests that the mixture of data characteristics in the time scale domain is another important data feature to be incorporated in the modeling process. This paper proposes a novel Morphological Component Analysis based hybrid methodology for modeling the multi scale heterogeneous data generating process. Empirical studies in the marker crude oil market show the significant performance improvement of the proposed algorithm, against benchmark models. The superior performance of the proposed model is attributed to the separation of the underlying distinct data features and the identification of appropriate model specifications for them. Meanwhile, the proposed methodology offers additional insights into the underlying data generating process and their economic viability.