Affiliations: Department of Computer Science, University of Bath
Bath BA2 7AY, United Kingdom | Polish-Japanese Institute of Information Technology
Koszykowa 86, 02-008, Warsaw, Poland | Institute of Mathematics, Warsaw University, Banacha
2, Warsaw 02-095, Poland
Abstract: Sunspots are the subject of interest to many astronomers and solar
physicists. Sunspot observation, analysis and classification form an important
part of furthering the knowledge about the Sun. Sunspot classification is a
manual and very labor intensive process that could be automated if successfully
learned by a machine. This paper presents machine learning approaches to the
problem of sunspot classification. The classification scheme attempted was the
seven-class Modified Zurich scheme [18]. The data was obtained by processing
NASA SOHO/MDI satellite images to extract individual sunspots and their
attributes. A series of experiments were performed on the training dataset with
an aim of learning sunspot classification and improving prediction accuracy.
The experiments involved using decision trees, rough sets, hierarchical
clustering and layered learning methods. Sunspots were characterized by their
visual properties like size, shape, positions, and were manually classified by
comparing extracted sunspots with corresponding active region maps (ARMaps)
from the Mees Observatory at the Institute for Astronomy, University of
Hawaii.
