search

decisionintelligence / TFB

TFB: Towards Comprehensive and Fair Benchmarking of Time Series Forecasting Methods (PVLDB 2024) https://www.vldb.org/pvldb/vol17/p2363-hu.pdf
MIT License
211 stars 20 forks source link

About result #13

Open Sylphid172 opened 3 weeks ago

Sylphid172 commented 3 weeks ago

We know that there are over 8000 univariate datasets, and the results of the paper are presented with different attributes (such as year). What I want to ask is how different datasets are counted into this result when calculating the result. I didn't understand after reading the source code

qiu69 commented 3 weeks ago

Thank you for your attention and questions！ For each category of dataset (year), we use a different config file to run it (fixed_forecast_config_year. json), and then when we calculate the results, we group them according to their characteristics (seasonal, trend. You can find them in FORECAST_META.csv）， and then we report the mean of all time series results within each group.