Supervised Machine Learning for the Automatic Classification of Triggers from ASIM/MXGS on board the ISS

During ASIM operations from June 2018 until the end of 2019, 486 TGFs were observed. For this task, the ASDC (ASIM Science Data Center) dealt with numerous triggers from the instrument (5000 per week). The relocation of the instrument from EPF SDX to EPF SDN (Starboard Deck Nadir) of the Columbus ISS Module on January 10, 2022, demonstrated that the MXGS location capabilities could be used not only for TGF location but also for imaging GRB events, as its Field of View in the SDN port encompasses both Earth and space. It's worth noting that only a few of the ASIM triggers correspond to TGF events. There is a screening process employing a series of algorithms to detect and discard false positives (triggers that are not TGFs). Nevertheless, the ASIM archive retains all data from every trigger. Due to the extended operational time, there is currently a sufficiently large database that enables us to present the initial results here using novel machine learning methods, such as kernel methods or neural networks, for the automatic categorization of both present and future events. Furthermore, our interest goes beyond mere classification, as we are currently investigating whether various explainability methods applied to these techniques can assist in identifying the relevant features of the signal for such classification. The aim of this work is to provide a tool to quantify new physical processes that could be the cause of instrument triggers and to examine whether there is a connection with the Earth-Space global circuit.