SUPERVISED LEARNING BASED IDENTIFICATION AND PREDICTION OF M AND X CLASS PRE-FLARE PROCESS USING SHARP DATA

Solar flare is one of the most energetic phenomena in the solar system. Solar flares can affect the dynamics of plasma and energy in the solar system including Earth. However, until now, the theory of solar flare formation is still not yet fully understood. Nevertheless, the role of magnetism in...

Full description

Saved in:
Bibliographic Details
Main Author: Alif Fernanda, Chandra
Format: Theses
Language:Indonesia
Online Access:https://digilib.itb.ac.id/gdl/view/75485
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Institut Teknologi Bandung
Language: Indonesia
id id-itb.:75485
spelling id-itb.:754852023-08-01T14:14:47ZSUPERVISED LEARNING BASED IDENTIFICATION AND PREDICTION OF M AND X CLASS PRE-FLARE PROCESS USING SHARP DATA Alif Fernanda, Chandra Indonesia Theses solar flare, solar flare prediction, supervised learning, SHARP, Python INSTITUT TEKNOLOGI BANDUNG https://digilib.itb.ac.id/gdl/view/75485 Solar flare is one of the most energetic phenomena in the solar system. Solar flares can affect the dynamics of plasma and energy in the solar system including Earth. However, until now, the theory of solar flare formation is still not yet fully understood. Nevertheless, the role of magnetism in generating solar flares has been confirmed. On the other hand, solar flares as an energy release process are certainly preceded by an energy accumulation process. The pre-flare energy accumulation process is believed to be reflected in its magnetic parameters. Therefore, the analysis of magnetic parameters is essential to understand the occurrence of solar flares. This thesis aims to analyze the role of magnetic parameters obtained from the Spaceweather HMI Active Region Patch (SHARP) database in predicting the occurrence of M and X-class solar flares. Additionally, this thesis also aims to predict the occurrence of M and X-class solar flares by examining the most influential magnetic parameters in their occurrence. Eighteen magnetic parameters from the SHARP database from May 10, 2010, to December 31, 2021, were reviewed as potential features for prediction using support vector machine. Then, the temporal positions of M and X-class solar flares relative to the peak of each magnetic parameter in their active region were examined to see the role of these parameters as precursors to solar flares. Moreover, significant changes or fluctuations that occur before solar flares are assumed to be a pre-flare process. The Short Time Fourier Transform (STFT) was used to observe this process more clearly. Based on the temporal distance distribution of M and X-class solar flares relative to the peak of each magnetic parameter, no consistent parameter was found to act as a precursor marker for solar flares. This was concluded from the high standard deviation values of the temporal distance distribution (¿60 hours) and an average value close to 0 hours. This implies a unique pre-flare process for each solar flare. Additionally, the significance ranking of magnetic parameters as features in identifying the pre-flare process was obtained. A ranking of 12 magnetic parameters showed correlation evolution with the pre-flare process, namely TOTPOT, ABSNJZH, SAVNCPP, TOTUSJH, TOTUSJZ, USFLUX, AREA ACR, MEANPOT, SHRGT45, MEANSHR, MEANGAM, and R VALUE. Considering the significance of these features, the optimal number of features to be used in the support vector machine was determined to be 7 features, which resulted in a relative precision increase of 40% compared to feature reduction. text
institution Institut Teknologi Bandung
building Institut Teknologi Bandung Library
continent Asia
country Indonesia
Indonesia
content_provider Institut Teknologi Bandung
collection Digital ITB
language Indonesia
description Solar flare is one of the most energetic phenomena in the solar system. Solar flares can affect the dynamics of plasma and energy in the solar system including Earth. However, until now, the theory of solar flare formation is still not yet fully understood. Nevertheless, the role of magnetism in generating solar flares has been confirmed. On the other hand, solar flares as an energy release process are certainly preceded by an energy accumulation process. The pre-flare energy accumulation process is believed to be reflected in its magnetic parameters. Therefore, the analysis of magnetic parameters is essential to understand the occurrence of solar flares. This thesis aims to analyze the role of magnetic parameters obtained from the Spaceweather HMI Active Region Patch (SHARP) database in predicting the occurrence of M and X-class solar flares. Additionally, this thesis also aims to predict the occurrence of M and X-class solar flares by examining the most influential magnetic parameters in their occurrence. Eighteen magnetic parameters from the SHARP database from May 10, 2010, to December 31, 2021, were reviewed as potential features for prediction using support vector machine. Then, the temporal positions of M and X-class solar flares relative to the peak of each magnetic parameter in their active region were examined to see the role of these parameters as precursors to solar flares. Moreover, significant changes or fluctuations that occur before solar flares are assumed to be a pre-flare process. The Short Time Fourier Transform (STFT) was used to observe this process more clearly. Based on the temporal distance distribution of M and X-class solar flares relative to the peak of each magnetic parameter, no consistent parameter was found to act as a precursor marker for solar flares. This was concluded from the high standard deviation values of the temporal distance distribution (¿60 hours) and an average value close to 0 hours. This implies a unique pre-flare process for each solar flare. Additionally, the significance ranking of magnetic parameters as features in identifying the pre-flare process was obtained. A ranking of 12 magnetic parameters showed correlation evolution with the pre-flare process, namely TOTPOT, ABSNJZH, SAVNCPP, TOTUSJH, TOTUSJZ, USFLUX, AREA ACR, MEANPOT, SHRGT45, MEANSHR, MEANGAM, and R VALUE. Considering the significance of these features, the optimal number of features to be used in the support vector machine was determined to be 7 features, which resulted in a relative precision increase of 40% compared to feature reduction.
format Theses
author Alif Fernanda, Chandra
spellingShingle Alif Fernanda, Chandra
SUPERVISED LEARNING BASED IDENTIFICATION AND PREDICTION OF M AND X CLASS PRE-FLARE PROCESS USING SHARP DATA
author_facet Alif Fernanda, Chandra
author_sort Alif Fernanda, Chandra
title SUPERVISED LEARNING BASED IDENTIFICATION AND PREDICTION OF M AND X CLASS PRE-FLARE PROCESS USING SHARP DATA
title_short SUPERVISED LEARNING BASED IDENTIFICATION AND PREDICTION OF M AND X CLASS PRE-FLARE PROCESS USING SHARP DATA
title_full SUPERVISED LEARNING BASED IDENTIFICATION AND PREDICTION OF M AND X CLASS PRE-FLARE PROCESS USING SHARP DATA
title_fullStr SUPERVISED LEARNING BASED IDENTIFICATION AND PREDICTION OF M AND X CLASS PRE-FLARE PROCESS USING SHARP DATA
title_full_unstemmed SUPERVISED LEARNING BASED IDENTIFICATION AND PREDICTION OF M AND X CLASS PRE-FLARE PROCESS USING SHARP DATA
title_sort supervised learning based identification and prediction of m and x class pre-flare process using sharp data
url https://digilib.itb.ac.id/gdl/view/75485
_version_ 1822007697395941376