Skip to content

Byte7/Solar-Flares-Prediction-RHESSI-Mission

Repository files navigation

Solar-Flares-Prediction-RHESSI-Mission

The initial steps taken in the preprocessing are as follows -

  1. The dataset was taken from NASA RHESSI Data repository accessed through a web form.
  2. The initial format of files was FITS. This was converted to CSV by using tools provided here.
  3. The many csv files generated from the FITS files were concatenated to make the current dataset in the repository. This script was used for the same.

Results

Our Top 3 models' results for prediction of a energy range of a solar flare based on attributes of the flare are -

  1. Gradient Boosting Classifier - 87 % accuracy
  2. Random Forests Classifier - 86 % accuracy
  3. Decision Tree Calssifier - 82 % accuracy

A detailed blog post is available here explaining how I went around the data and prediction of flare's energy.