Performance Analysis of Artificial Neural Network Approach on Solar Radio Burst Detection

Abstract

Solar radio burst is defined as a massive solar radio emission related to the solar flare event occurrences. It is related to space weather events and will triggered an interference in our radio waves signal and affected the electromagnetic spectrum on earth. The solar flare could strike and condemn entire communications line including satellite operation, navigation system, Global Positioning System (GPS), international electrical grid and many more. Solar radio burst is the early warning sign that can helps reducing the effect by taking a precaution action by shutting down system. Because the solar radio is in the low frequency range, the detector system consist of low-frequency receiver is used to detect the burst event. As for Malaysia, solar radio observations are currently carried out using Compact Astronomical Low-cost, Low Frequency Instrument for Spectroscopy and Transportable Observatory (CALLISTO) which been placed at the Malaysia Space Agency (MYSA) Banting, Selangor. The application of Artificial Neural Network (ANN) helps in preparing the proper prediction on solar radio burst using solar radiation readings from the spectrometer. ANN is divided into two main group which are static and dynamic neural network. In static neural network, the data propagates in a single direction from input to the output whereas, in dynamic neural network, the data propagates regardless of its direction. In this paper, both static and dynamic neural network had been applied to the data obtained from CALLISTO to develop a solar radiation prediction model to detect the solar radio burst. Based from the results, it is shown that dynamic neural network given the best results compared to the static neural network.