Abstract: Accurate and reliable precious metal futures price forecasting is extremely crucial for investment decision-making and government gold reserves. In this paper, an adaptive model named VMD-Res.-EEMD-ELM is documented for predicting precious metal futures price, which combines the advantages of secondary decomposition and extreme learning machines. Variational modal decomposition (VMD) is selected as the main decomposition technique to generate a sequence of modal components (VMFi) and residual sequence (Res.). The ensemble empirical mode decomposition (EEMD) is used to perform secondary decomposition of the residual sequence. And then each component is put into the extreme learning machine (ELM) with good generalization ability to generate the outputs which will be superimposed to form the final prediction result. The proposed model not only makes full use of the advantages of the secondary decomposition technology, but also solves the problem that the traditional variational modal decomposition hybrid prediction model does not consider the influence of residuals. The new model is tested on the historical data of two daily return rate sequence - gold and silver futures price, which are collected from Choice financial terminal (the main financial data aggregator in China, an equivalent of Bloomberg). Empirical research exhibits that the hybrid model proposed in this paper can fully capture the characteristics of the daily return rate sequence of gold and silver futures price with the excellent performance which achieves directional accuracy(DSTAT) of 83.33% and 93.33%, and MAE of 0.15 and 0.11 respectively. Meanwhile, by comparison, the prediction accuracy of the proposed model is significantly higher than other existing models.

Key words: time series; decomposition and integration; multi-modal integration prediction; machine learning；hybrid model