Informer_Casual_LSTM: Causal Characteristics and Non-Stationary Time Series Prediction
Yuxia Yang, Yilin Dong · 2024
Many existing forecasting models tend to work well when forecasting short-sequence time series. However, when working with long sequence time series, the performance suff ers significantly. Recently, there has been more intense research in this direction, and Informer is currently the most efficient predicting model. Informer's main drawback is that it does not allow for incremental learning. This paper proposes an improved multi-scenario time series forecasting model-Informer CasualConv_GRU, which captures causal relationships at different time scales through multi-scale causal convolution and gated recurrent units (GRU) to address strong nonlinearity and high noise in the data. Validation on datasets from the energy and financial sectors shows that the proposed model performs excellently in short-term and mid-term forecasting.