Forecasting Crude Oil Price Using Secondary Decomposition‐Reconstruction‐Ensemble Model Based on Variational Mode Decomposition
提出一种基于变分模态分解的二次分解-重构-集成模型，结合TTAO优化BiLSTM，预测WTI原油现货价格，误差逐步降低。
ABSTRACT The fluctuating crude oil price affects producers, consumers, investors, policy‐making, and economic stability. This paper forecasts the spot price of West Texas Intermediate (WTI) crude oil using weekly data from 1991 to 2024, considering factors from the US crude oil market, financial markets, and economic policies. We present a new secondary decomposition‐reconstruction‐ensemble model based on variational mode decomposition (VMD). Triangulation topology aggregation optimizer (TTAO) algorithm is first utilized to optimize the VMD and BiLSTM for sequence decomposition and prediction. The proposed model reconstructs sequences based on the permutation entropy (PE) of subsequences after primary decomposition and conducts a secondary decomposition on the high‐frequency reconstructed sequence. The model predicts subsequences and reconstructed sequences using TTAO‐BiLSTM and integrates results via LSTM. Prediction errors decrease sequentially across univariate BiLSTM, multivariate BiLSTM, single decomposition‐ensemble, single decomposition‐reconstruction‐ensemble, and the proposed secondary decomposition‐reconstruction‐ensemble models. TTAO outperforms adaptive moment estimation (Adam) in optimizing BiLSTM within all models.