Using Cloud Based and Random Forest to Predict the Frequency of the Solar Flare

Abstract

Solar flares are intense bursts of radiation from the sun that can significantly impact space weather, disrupting satellite communications, GPS systems, and even terrestrial power grids. In this study, this paper explore the use of machine learning for predicting the frequency of solar flares by leveraging the UCI Solar Flare Dataset. A Random Forest Regressor model is employed, with hyperparameter tuning performed via GridSearchCV to enhance predictive accuracy. The preprocessing pipeline includes label encoding for categorical attributes such as Zurich class, spot size, and spot distribution, alongside the removal of redundant or low-variance features. Our model achieves competitive performance, with a mean squared error (MSE) and a mean absolute error (MAE). To support scalable data processing and model training, the system was deployed on a cloud-based platform. The results highlight the promise of machine learning techniques in advancing space weather forecasting capabilities, offering potential benefits for early-warning systems and infrastructure resilience.