<ul><li>A new framework called AEFIN is proposed for non-stationary time series forecasting.</li><li>AEFIN enhances information sharing between stable and unstable components using a cross-attention mechanism.</li><li>The framework combines Fourier analysis networks with MLP to explore seasonal patterns and trend characteristics.</li><li>AEFIN outperforms common models in forecasting accuracy under non-stationary data conditions, showcasing improved mean square error and mean absolute error.</li></ul>

Non-Stationary Time Series Forecasting Based on Fourier Analysis and Cross Attention Mechanism

Discover more