nasa-ibm-ai4science/surya-bench-flare-forecasting

该数据集为太阳耀斑预测提供了基于滚动窗口方法的二进制标签。这些标签是根据每个窗口内的最大耀斑强度和累积耀斑强度生成的。数据集覆盖了从2010年5月13日到2024年12月31日的时间段，并分为训练集、验证集和测试集。数据集支持两种任务：二元分类和回归任务。每个记录都包括时间步长、24小时窗口内的最高耀斑强度、累积指数以及对应的二进制标签。

This dataset provides binary labels for solar flare forecasting using a rolling window approach. The labels are generated based on the maximum flare intensity and the cumulative flare intensity observed within each window. The dataset covers the period from May 13, 2010, to December 31, 2024, and is divided into training, validation, and test sets. It supports two tasks: binary classification and regression. Each record includes the timestep, the highest flare intensity in the 24-hour window, the cumulative index, and the corresponding binary labels.