Foreign Exchange Prediction using CEEMDAN and Improved FA-LSTM

Abstract

In Foreign Exchange (Forex) Prediction with high accuracy it becomes a challenge because time series data has chaotic characteristics, uncertainty, and complexity. To improve the accuracy of the forex prices prediction, prediction models are proposed which Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Improved Firefly Algorithm-Long Short Term Memory (IFA-LSTM). In this model the preprocessing data using the CEEMDAN to decomposed into IMF sequence and residual sequence. LSTM prediction models are established for all each characteristic series from CEEMDAN deposition. IFA is applied to optimize neural network structure to improve the performance of the model prediction accuracy. We compare our proposed models with LSTM and CEEMDAN-LSTM models, the experimental results show that the proposed models performs better in the prediction of forex time series.