A Novel Non-Ferrous Metals Price Forecast Model Based on LSTM and Multivariate Mode Decomposition

Non-ferrous metals are important bulk commodities and play a significant part in the development of society. Their price forecast is great reference value for investors policymakers. However, developing robust model tricky due to price’s drastic fluctuations. In this work, novel fusion based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Singular Spectrum Analysis (SSA), Long Short-Term Memory (LSTM) constructed non-ferrous forecast. Considering complexity their change, dual-stage signal preprocessing which combines CEEMDAN SSA utilized. Firstly, we use algorithm decompose original nonlinear sequence into multiple Intrinsic Functions (IMFs) residual. Secondly, component maximum sample entropy decomposed by SSA; so-called Multivariate (MMD). A series experimental results show that proposed MMD-LSTM method more stable than other seven benchmark models, providing reasonable scheme metals.