This version uses the latest heasoft updates (as of August 2021). The model is optimizd to work in the energy range 0.5-10 keV (unlike the previous version). It uses 40 MKF parameters, including those from the geomagnetic data. It is a classification model that samples the MKF parameters every 1 second and classifies the background data into 35 basis spectra. The root-mean-squared performance in the background estimates is 0.21 counts/s vs 4.04 in the 3C50 model. If the 1% outlier backgorund observations are discarded, the performance is 0.15 counts/s (vs 1.25 for the 3C50 model).
This version requires a geomagnetic data files that can be downloaded from https://heasarc.gsfc.nasa.gov/FTP/caldb/data/gen/pcf/geomag (see discussion here).. Please download the 6 files: dst_kyoto.fits, f107_petincton.fits, geomag.tar.gz, kp_noaa.fits, kp_potsdam.fits, solarphi_oulu.fits into some directory and then provide the location of that directory as input to the nicerBgML.py script.