Enhancing Solar Flare Image Prediction Using Autoencoders

EasyChair Preprint 11778

Abstract

The study introduces an innovative framework for enhancing solar flare image prediction through deep learning - autoencoding, addressing limitations of traditional machine learning models. Solar flares impact space weather and Earth's technology, requiring improved prediction models. The research integrates an auto-encoder and advanced techniques like binary thresholding, Hessian matrix eigenvalue calculation, and Canny edge detection for feature extraction and shape analysis. Motivated by societal and economic impacts, it aims to mitigate disruptions caused by solar flares. Recent incidents underscore the urgency for reliable predictive methods. The study combines image processing and machine learning, utilizing an Autoencoder with convolutional and transpose convolutional layers, contributing to feature representation and understanding in solar flare analysis.

Keyphrases: Autoencoding, Binary Thresholding, Canny Edge Detection, Hessian matrix, solar flare

@booklet{EasyChair:11778,
  author    = {Alfred Mawuli Dogbatse and Adrian Beard and Masud Rana Rashel and Akm Kamrul Islam},
  title     = {Enhancing Solar Flare Image Prediction Using Autoencoders},
  howpublished = {EasyChair Preprint 11778},
  year      = {EasyChair, 2024}}