On performance assessment of machine learning- based GNSS ionospheric delay correction model based on space weather predictors in immediate positioning environment

Availability of advanced open-source statistical computing environments, such as R and Python as well as GUI-based derivatives of those, combined with the abundance of space weather observations openyl available attract a growing number of researchers to develop predictive and explanatory machine learning-based models. Experience show the insufficient attention is given to model performance assessment, and selection of feasible models based on their performance. Here we argue the more consideration should be given to machine learning-based model performance assessment, propose the statistics-based methodology for space weather-regarded model performance assessment, and demonstrate its deployment in the case of machine learning-based GNSS ionospheric correction model based on space weather predictors in the immediate positioning environment of a GNSS receiver.