GestaltCogTeam / BasicTS

A Fair and Scalable Time Series Forecasting Benchmark and Toolkit.
Apache License 2.0
626 stars 109 forks source link

some issue about model performance #100

Closed chenywu closed 2 months ago

chenywu commented 10 months ago

The models trained in this repository perform much better than the models trained in other repositories. For example, Graph Wavenet in this repository has a mae of 14.65 on the PEMS08 dataset, while Graph Wavenet in another repository has a mae of 18.28 on the PEMS08 dataset. what have you done to improve the accuracy?

zezhishao commented 10 months ago

You can refer to our paper to find the reasons for inconsistent performance. Exploring Progress in Multivariate Time Series Forecasting: Comprehensive Benchmarking and Heterogeneity Analysis.