Applying Machine Learning Tools to the Identification of Foreshock Transient Events

Abstract:

Our previous research attempted to establish the relationship between foreshock transient events and transients in the ionosphere observed with ground magnetometers. This earlier work relied on foreshock transient event lists that were generated by a visual survey of the THEMIS data near the bowshock/foreshock. Our aim is to extend our earlier work, and the overall understanding of foreshock transients, by employing machine learning tools to identify foreshock transient events. Successful application of these tools would allow use to survey much more data. We first present results of automated classification of THEMIS data into the three primary regions of solar wind, magnetosheath, and magnetosphere. We then present our initial results of training an SVM classifier using the human generated event list and applying it to a more extensive data set.