Identifying Cassini's Magnetospheric Location Using Magnetospheric Imaging Instrument (MIMI) Data and Machine Learning

We analyzed data from Cassini's Magnetospheric Imaging Instrument (MIMI) and Magnetometer (MAG) and attempted to identify the region of Saturn's magnetosphere that Cassini was in at a given time using machine learning. MIMI data are from the Charge-Energy-Mass Spectrometer (CHEMS) instrument and the Low-Energy Magnetospheric Measurement System (LEMMS). We trained on data where the region is known based on a previous analysis of Cassini Plasma Spectrometer (CAPS) plasma data. Three magnetospheric regions are considered: Magnetosphere, Magnetosheath, and Solar Wind. MIMI particle intensities, magnetic field values, and spacecraft position are used as input attributes, and the output is the CAPS-based region, which is available from 2004 to 2012. We then use the trained classifier to identify Cassini's magnetospheric regions for times after 2012, when CAPS data is no longer available. Training accuracy is evaluated by testing the classifier performance on a time range of known regions that the classifier has never seen. Preliminary results indicate a 68% accuracy on such test data. Other techniques are being tested that may increase this performance. We present the data and algorithms used, and will describe the latest results, including the magnetospheric regions post-2012 identified by the algorithm.