Abstract:
After centuries of acceptance of geocentrism, the heliocentric model and cosmic pluralism strengthened the idea of multiple worlds. Currently, thousands of these objects have already been confirmed and a considerable fraction of them was detected by the Kepler mission, through the transit method. This method makes it possible to analyze the light curves of observed stars to search for temporary decrease in the stellar flux caused by the passage of celestial bodies in front of the stellar disks. However, it is not enough just to detect these bodies, it is also important to characterize them. The planetary parameters directly depend on the atmospheric parameters of their host stars. Based on this, the Kepler Input Catalog (KIC) was created, which is a compilation of other databases, gathering information about the stars in the Kepler mission’s field of view. However, the precision of the KIC parameters has been discussed, since they were not obtained homogeneously. This work aims to recharacterize stars from the Kepler mission, combining Machine Learning and photometric information from the 12 optical filter system of the Javalambre Photometric Local Universe Survey (J-PLUS) and from 4 filters from the Wide-field Infrared Survey Explorer (WISE), in order to generate stellar parameter prediction algorithms (Tef, log g and [Fe/H]). In addition to the 16 magnitudes, their combinations in pairs (colors) create another 120 input columns, with useful information for the algorithm (totaling 136 columns of magnitude information per object, the features). In addition, each object in the sample used for training must have its physical parameters (Tef, log g and [Fe/H]) well defined in the literature. For our predictors, we retrieve these values from a crossmatch between LAMOST, WISE and J-PLUS, with about 100.000 stars. After careful optimization of the models generated by the algorithm, we obtained good predictions for these three atmospheric parameters. Their precisions were ~70 K for Tef, ~0,08 dex for log g, and ~0,10 dex for [Fe/H], in the test sample. We applied the models to 44.483 stars from the Kepler mission, also observed by J-PLUS/WISE. Using the predicted parameters, we were able to estimate the bolometric correction (with a precision of ~0,022 mag), and derive luminosities, radii and masses of these objects. This information can be very useful for the characterization of transiting exoplanets, eventually detected around these stars.