An Efficient Feature-Optimized Data Mining Model for Geomagnetic Storm Forecasting Using Solar Wind and Interplanetary Magnetic Field Parameters

In the study a geomagnetic storm forecasting data mining model is created with the help of key solar wind and interplanetary magnetic field (IMF) parameters that are optimized based on their features. The suggested model makes use of high-resolution solar wind plasma and IMF measurements retrieved in the OMNI database, and the disturbance storm time (Dst) index is used as the main measure of the strength of geomagnetic storms. A strategy of feature optimization is adopted to determine the most geoeffective parameters to predict the computational efficiency and reliability of predictive power, such as solar wind speed, proton density, total vertical magnetic field strength and southward component of IMF (Bz). Weakly correlated and redundant features are automatically removed in order to simplify the model without decreasing forecast accuracy. An optimized feature subset is then applied to the data to form a predictive model based on data mining that has the potential to learn the nonlinear relationships between the upstream solar wind conditions and geomagnetic response. The performance of the models is measured by standard statistical measures and compared to non-optimized baseline models. These findings indicate that the proposed methodology has a better forecasting accuracy at a much lower cost of computation, thus it can be used in near real-time space weather forecasting. The results indicate that feature selection is also vital in space weather prediction and that the majority of the geomagnetic disturbances are due to IMF orientation and solar wind dynamics. The suggested model offers a stable and effective framework of the operational geomagnetic storm forecasting and leads to the creation of well-grounded data-driven space weather forecasting systems.