Automatic Classification of Sunspot Groups for Space Weather Analysis

The sun is the unlimited energy source for life on the earth. However, besides as the energy
source, the sun also gives disruptions to the universe around the earth and also to the life on
the earth. Sources of the disruptions from the sun are flares and Coronal Mass
Ejection/CME. Both of those disruptions in general come from group of sunspots. With the
growing of dependency of human life with modern technology, either facility on the surface of
the earth or in universe around the earth, the disruptions from the sun should be anticipated.
In order to know the complexity level of sunspot groups and their activity, Modified-Zurich
sunspot classification is used. Image of sunspots can be taken using the Michelson Doppler
Imager instrument (MDI) Continuum / SOHO (Solar and Heliospheric Observatory).
This research was conducted on the automatic classification of sunspot group that can be
used to analyze the space weather conditions and provide information to the public. There
are two stages to classify sunspot groups namely feature extraction and pattern recognition.
For feature extraction, we used digital image processing to get features of sunspot group, and
for pattern recognition, we used artificial neural network. We compared 3 methods of
artificial neural networks to get the best result of classification namely backpropagation,
probabilistic and combination between self-organizing map and k-nearest neighbor. Among
three of them, probabilistic neural network gave the best classification result.