Speaker
Description
The ongoing study aims to predict geomagnetic disturbances at low latitudes with machine learning. For this, however, a treatment of missing data (gaps) in satellites and ground stations is necessary. In order to solve this problem, the Python library heliogap was developed, which automates the extraction, cleaning, and statistical gap analysis of solar wind parameters from OMNIweb/ACE and ground magnetic field data from the Embrace MagNet network.The library features a self-healing cache architecture and a mathematical engine optimized for the processing of millions of data points, being capable of isolating continuous blocks of data. It evaluates multiple interpolation methods, such as Akima, nearest, and quadratic, in simulated data gaps, generating standardized error matrices (RMSE, MAE, WMAPE).Preliminary results provided a statistical diagnosis of the gap distributions. By establishing a fully automated data curation pipeline, heliogap creates the base for the future training of predictive AI algorithms and for the subsequent deployment of a real-time monitoring platform.