Analysis of Asia Pacific stock markets with a novel multiscale model

Stock price prediction is considered a challenging task in the field of financial time series prediction. In recent years, the application of new data mining techniques, including empirical mode decomposition (EMD), to financial time series prediction has attracted increasing attention. Unfortunately, EMD has two major shortcomings when applied to this task: (1) EMD has been traditionally applied to very long time series, and is subject to a long incubation period precluding its real-time application. (2) After the application of EMD, large volumes of data are produced, and some form of dimensionality reduction is still required. In order to solve these problems and improve EMD’s performance in time series prediction, this paper proposes a hybrid model combining EMD, principal component analysis (PCA) and BP neural network (BPNN). This novel hybrid model is based on concepts of decomposition and information fusion. In order to evaluate its forecasting performance, the proposed model was compared with other four typical models, with prediction metrics demonstrating its superiority, including in terms of directional symmetry (DS).