Comparative Evaluation of Hybrid Machine Learning-ARIMA Models for Geomagnetic Storm Prediction Using the Kp Index

This dataset contains the processed daily time series used for training and evaluating Comparative Evaluation of Hybrid Machine Learning-ARIMA Models for Geomagnetic Storm Prediction Using the Kp Index The dataset includes geomagnetic and solar parameters compiled from publicly available sources (NOAA Space Weather Prediction Center and GFZ Potsdam). Variables include Kp index components, Ap index components, daily Ap average, Sunspot Number (SN), F10.7 solar radio flux (observed and adjusted), Bartels Solar Rotation number (BSR), and derived features used in preprocessing. The Kp index was used both as a continuous target variable and as a binary storm classification variable (Storm = 1 if Kp > 4; 0 otherwise). Temporal resolution: DailyTemporal coverage: 2010–2024 All preprocessing and modeling procedures are described in the associated manuscript submitted to Earth and Space Science (Manuscript ID: 2026EA005111).