Irina Zhelavskaya

ST Solar-Terrestrial Sciences

The 2018 Outstanding Student Poster and PICO (OSPP) Award is awarded to Irina Zhelavskaya for the poster/PICO entitled:

Systematic analysis of machine learning techniques for Kp prediction in the framework of the H2020 project ‘SWAMI’ (Zhelavskaya, I.; Shprits, Y.; Vasile, R.; Stolle, C.; Matzka, J.)

Click here to download the poster/PICO file.

Irina Zhelavskaya is a PhD student at the University of Potsdam, Germany, and a research assistant at the GFZ German Research Centre for Geosciences in Potsdam in Section 2.8 “Magnetospheric physics” supervised by Prof. Yuri Shprits. Her PhD project is supported by the fellowship from Geo.X, the research network for Geosciences in Berlin and Potsdam. Her research lies at the intersection of space physics and machine learning with a particular focus on modeling the dynamics of the plasmasphere and prediction of the geomagnetic Kp index.
The awarded poster presents a systematic comparison of different machine learning techniques on the task of predicting the Kp index up to 9 hours ahead, and how methods for optimal inputs selection can be applied to this task. The study shows that Neural Networks have the optimal performance on this task, and that Random Forest can be used to select a reduced set of optimal inputs to the predictive model without a significant reduction in performance.