Code for our CIKM'22 paper Spatial-Temporal Identity: A Simple yet Effective Baseline for Multivariate Time Series Forecasting.

Code for our SIGKDD'24 paper GinAR: An End-To-End Multivariate Time Series Forecasting Model Suitable for Variable Missing

A Fair and Scalable Time Series Forecasting Benchmark and Toolkit.

Dynamic frequency domain graph convolution for traffic prediction

Code for our Information Fusion paper MGSFformer: A Multi-Granularity Spatiotemporal Fusion Transformer for air quality prediction

Code for our SIGKDD'22 paper Pre-training-Enhanced Spatial-Temporal Graph Neural Network For Multivariate Time Series Forecasting.

Code for our CIKM'23 paper DSformer: A Double Sampling Transformer for Multivariate Time Series Long-term Prediction