Considering the significant increase in interest in using machine learning (ML) methods for Earth Orientation Parameters (EOP) forecasting, we would like to organize a special internal sub-campaign as part of the post-operational phase of the Second Earth Orientation Parameters Prediction Comparison Campaign ‒ EOP PML. The purpose of EOP PML is to analyse ML approaches in EOP forecasting in more detail. To objectively evaluate the ML-based forecasting methods themselves without analysing the impact of input data selection, we propose that willing participants submit their EOP forecasts by adhering to even stricter rules than those imposed for the 2nd EOP PCC.
The forecasts submitted by participants of EOP PML will be included in the analyses routinely conducted by the campaign office as part of the 2nd EOP PCC. This includes routine analysis and comparison with predictions provided by the regular campaign contributors. These ML-based predictions that will follow the stricter rules will therefore contribute to the overall database of predictions and will be analysed according to the current rules (to compare with existing methods). Additionally, they will also be analysed separately (to evaluate the method itself). Thanks to this, on the one hand, we will be able to continue all 2nd EOP PCC activities with increased number of submissions and, on the other hand, will be capable of evaluating the ML methods in a more detailed and objective manner.
To take part in the EOP PML, one needs to register using the registration form, which is the same as for the regular 2nd EOP PCC.
Registration for the EOP PML is now open, and the first submission of predictions in the EOP PML is scheduled for Wednesday, November 6, 2024 (before 20:00 UTC). However, any new candidate is welcomed at any time. If you are not ready to adjust your forecasts to the new rules by this time, you can also submit your predictions at a later date (the only requirement is to submit forecasts every Wednesday before 20:00 UTC).
Detailed information about the EOP PML and instructions can be found in this document.
