Assessing the Multiple Imputation Approach for Univariate Time Series Data of Geomagnetic Disturbance Event in Solar Cycle 24

Space weather data frequently contains gaps and missing information due to factors such as the uneven distribution of monitoring equipment, interruptions in transmission, equipment malfunctions, and the inherently dynamic nature of space weather events. These data gaps pose challenges in constructing comprehensive models, understanding the underlying physics, and making well-informed decisions to mitigate potential risks. This study introduces an innovative approach that employs the Known Sub-Sequence Algorithm (KSSA) to impute missing data in univariate time series space weather and geomagnetic data. The KSSA, a machine learning algorithm, provides nine distinct imputation methods that can be assessed for each dataset. The dataset used encompasses univariate space weather time series data, including parameters like solar wind speed, solar wind dynamic pressure, heliospheric magnetic field, symmetrical H index, and low-latitude geomagnetic data. The dataset comprises a total of 40,320 data points collected at a one-minute frequency. Performance evaluation employs the root mean square error (RMSE). The study’s findings underscore the superiority of the Autoregressive Integrated Moving Average (ARIMA) method across various parameters and iterations, followed by the Stineman Interpolation (STIT) method. The incorporation of KSSA and the comparative assessment of imputation methods yield valuable insights into effectively addressing missing data challenges inherent in the analysis of space weather and geomagnetic data.