Hybrid time series prediction model based on SSA-VMD-TCN

Abstract

In order to improve the prediction accuracy of non-stationary time series, this paper proposes a deep learning hybrid model SSA-VMD-TCN based on sparrow search algorithm (SSA), variational mode decomposition (VMD) and sequential convolution network (TCN). The model achieves better prediction effect by reducing the complexity of nonlinear sequence. The sSA-VMD-TCN model first uses VMD to effectively decompose the original sequence into a certain number of intrinsic modal components (IMF) and residual components. Meanwhile, SSA algorithm is used to optimize the input parameters ofTCN prediction model, and then the models are modeled on each IMF. Finally, the results of each sequence test set are added as the final result. This shows that the model is an effective time series forecasting model.