Financial Time Series Prediction Based on EEMD-RNN

Abstract

This paper proposes to construct a nonlinear financial time series prediction model with recurrent neural networks (RNN) and ensemble empirical mode decomposition (EEMD). And the effectiveness of the model is explored using the Hang Seng Composite Index (HSCI) as an example. The empirical results show that the combined prediction based on EEMD-RNN is much better than those based on other approaches, such as RNN and BP neural networks. The combined model can also effectively predict the future trend of the HSCI based on the reconstructed series, and can achieve prediction accuracy of 82.86%.