Table 1 Summary of benchmarks used in this work.
From: Hybrid evolutionary-gradient training improves long-term time series forecasting
Dataset | Domain | Interval | Brief description |
|---|---|---|---|
ETTm140 | Energy | 1 min | Electricity transformer measurements including load and oil temperature, with data spanning July 2016 to July 2018 |
ETTm240 | Energy | 1 min | Same source as ETTm1 with a different transformer stream; supports long horizon forecasting at minute level |
ETTh140 | Energy | 1 h | Hourly transformer data for long horizon forecasting |
ETTh240 | Energy | 1 h | Another hourly stream from the same domain; evaluates robustness across forecasting horizons |
Electricity41 | Energy | 1 h | Hourly electricity consumption of 321 customers from 2012 to 2014 |
Exchange42 | Economy | 1 day | Daily exchange rates of eight countries from 1990 to 2016 |
Traffic43 | Transportation | 1 h | Road occupancy measurements from 862 sensors on SF Bay Area highways, covering January 2015 to December 2016 |
Weather44 | Weather | 10 min | 21 meteorological variables recorded at 10 min intervals during 2020 at the MPI-BGC weather station |
