Scalable Transformer for High Dimensional Multivariate Time Series Forecasting

This paper was accepted by CIKM 2024. We provide the open-source code here.

Running pipeline

Relation sparsity

Run ./datasets/crime/top-k-train corr-compute.py to get the correlation, modeling with the accelerated computation - DeepGraph.

Run the main file

Config the parameters and run ./run.py to train and evaluate the model.