XLinear：增强频率的多层感知器与交叉滤波器用于稳健的长程预测

Time series forecasting is widely utilized across various domains. Multi-layer perceptron (MLP)-based forecasters have shown greater robustness to noise compared to Transformer-based approaches. However, MLPs struggle with capturing complex features, which limits their ability to model long-range dependencies. To address this issue, XLinear is proposed as an MLP-based forecaster for long-range forecasting. The model first decomposes the time series into trend and seasonal components. For the trend component, which contains long-range characteristics, Enhanced Frequency Attention (EFA) is designed to capture long-term dependencies by leveraging frequency-domain operations. Additionally, a CrossFilter Block is introduced to handle the correlations in the seasonal component, thereby enhancing the accuracy and robustness of the forecasts.