東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Linked to FindBook

Google Book

Amazon

博客來

Improving Solar Flare Forecasting Based on Time Series Magnetic Features Using Advanced Machine Learning.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Improving Solar Flare Forecasting Based on Time Series Magnetic Features Using Advanced Machine Learning./
Author:	Alvarez, Alejandro J.
Description:	1 online resource (276 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 84-09, Section: A.
Contained By:	Dissertations Abstracts International84-09A.
Subject:	Astrophysics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29999754click for full text (PQDT)
ISBN:	9798374417128

Improving Solar Flare Forecasting Based on Time Series Magnetic Features Using Advanced Machine Learning.
Alvarez, Alejandro J.

Improving Solar Flare Forecasting Based on Time Series Magnetic Features Using Advanced Machine Learning. - 1 online resource (276 pages)

Source: Dissertations Abstracts International, Volume: 84-09, Section: A.

Thesis (Ph.D.)--George Mason University, 2022.

Includes bibliographical references

The overarching goal of this dissertation study is to improve the prediction of solar flares, an energetic eruptive phenomenon occurring in the active regions of the Sun and a known driver of space weather. As humans have become more reliant on space-based services, like satellite communication, navigation, and weather forecasting, while continuing to strive towards the exploration of other worlds, knowledge and prediction of space weather becomes imperative. To achieve the stated goal, this research makes a comprehensive study ranging from building advanced data models to developing novel time-series classifiers. The work is grouped into six different but interconnected tasks. These include (1) the identification, analysis and use of a modern database of magnetic feature parameters extracted from high-resolution, high cadence magnetogram data from the SDO spacecraft; (2) characterizing, pre-processing and visualization of time-series magnetic feature data; (3) development of new sample definitions, new combinations and prediction models and implementation taking into consideration the flaring state of solar active regions; (4) prediction improvement using optimal sample time windows; (5) prediction improvement using time-series classification vs point-in-time classification; and (6) the development of a new multivariate time series classification (MTSC) algorithm capable of working directly with our data without the need for any data transformation while allowing for model visualization and gaining an understanding of the prediction models generated.The research work completed has made significant improvements in multiple aspects of the study of solar flare prediction. New definitions for solar samples (FAF, FAN, NAF, and NAN Sample Types) were developed and used to help transition the traditional flare versus non-flare analysis and prediction modeling allowing for more detailed and improved labeling of data based on active region state. Sample Types were further exploited by combining them to build various prediction models for either practical application (nominal flare prediction and all-clear flare prediction) and scientific research (prediction of disparate data types versus prediction of similar data types). The research-oriented prediction clearly shows that the similarity of data samples FAF and FAN from flare-producing active regions is the key obstacle to making further improvements in flare prediction. The study also consistently shows that the usage of shorter data sampling windows yields a better prediction as measured by a variety of scoring metrics. The optimal sampling window is probably 12 hours as it produces high skill-score predictions as well as a sufficiently long prediction window. Surprisingly, the study revealed a limited improvement that time-series flare classification has over point-in-time predictions based on the current database. Thus, there is a need to expand data sources and features used in flare prediction models to make significant improvement. The newly developed multivariate time series classification algorithm is novel and promising, as it generates satisfactory prediction results.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2023

Mode of access: World Wide Web

ISBN: 9798374417128Subjects--Topical Terms:

535904
Astrophysics.
Subjects--Index Terms:

Data samplingIndex Terms--Genre/Form:

542853
Electronic books.

Improving Solar Flare Forecasting Based on Time Series Magnetic Features Using Advanced Machine Learning.
LDR:04745nmm a2200445K 4500 001 2354815
005 20230501063927.5
006 m o d
007 cr mn ---uuuuu
008 241011s2022 xx obm 000 0 eng d
020 $a 9798374417128
035 $a (MiAaPQ)AAI29999754
035 $a AAI29999754
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Alvarez, Alejandro J. $3 1669560
245 1 0 $a Improving Solar Flare Forecasting Based on Time Series Magnetic Features Using Advanced Machine Learning.
264 0 $c 2022
300 $a 1 online resource (276 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 84-09, Section: A.
500 $a Advisor: Zhang, Jie; Kinser, Jason.
502 $a Thesis (Ph.D.)--George Mason University, 2022.
504 $a Includes bibliographical references
520 $a The overarching goal of this dissertation study is to improve the prediction of solar flares, an energetic eruptive phenomenon occurring in the active regions of the Sun and a known driver of space weather. As humans have become more reliant on space-based services, like satellite communication, navigation, and weather forecasting, while continuing to strive towards the exploration of other worlds, knowledge and prediction of space weather becomes imperative. To achieve the stated goal, this research makes a comprehensive study ranging from building advanced data models to developing novel time-series classifiers. The work is grouped into six different but interconnected tasks. These include (1) the identification, analysis and use of a modern database of magnetic feature parameters extracted from high-resolution, high cadence magnetogram data from the SDO spacecraft; (2) characterizing, pre-processing and visualization of time-series magnetic feature data; (3) development of new sample definitions, new combinations and prediction models and implementation taking into consideration the flaring state of solar active regions; (4) prediction improvement using optimal sample time windows; (5) prediction improvement using time-series classification vs point-in-time classification; and (6) the development of a new multivariate time series classification (MTSC) algorithm capable of working directly with our data without the need for any data transformation while allowing for model visualization and gaining an understanding of the prediction models generated.The research work completed has made significant improvements in multiple aspects of the study of solar flare prediction. New definitions for solar samples (FAF, FAN, NAF, and NAN Sample Types) were developed and used to help transition the traditional flare versus non-flare analysis and prediction modeling allowing for more detailed and improved labeling of data based on active region state. Sample Types were further exploited by combining them to build various prediction models for either practical application (nominal flare prediction and all-clear flare prediction) and scientific research (prediction of disparate data types versus prediction of similar data types). The research-oriented prediction clearly shows that the similarity of data samples FAF and FAN from flare-producing active regions is the key obstacle to making further improvements in flare prediction. The study also consistently shows that the usage of shorter data sampling windows yields a better prediction as measured by a variety of scoring metrics. The optimal sampling window is probably 12 hours as it produces high skill-score predictions as well as a sufficiently long prediction window. Surprisingly, the study revealed a limited improvement that time-series flare classification has over point-in-time predictions based on the current database. Thus, there is a need to expand data sources and features used in flare prediction models to make significant improvement. The newly developed multivariate time series classification algorithm is novel and promising, as it generates satisfactory prediction results.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2023
538 $a Mode of access: World Wide Web
650 4 $a Astrophysics. $3 535904
650 4 $a Information science. $3 554358
653 $a Data sampling
653 $a Multivariate time series classifier
653 $a Sample types
653 $a Selective sampling
653 $a Solar flare prediction
653 $a Solar weather
653 $a Magnetic features
653 $a Machine learning
653 $a Solar forecasting
655 7 $a Electronic books. $2 lcsh $3 542853
690 $a 0596
690 $a 0800
690 $a 0723
710 2 $a ProQuest Information and Learning Co. $3 783688
710 2 $a George Mason University. $b Computational Sciences and Informatics. $3 3169362
773 0 $t Dissertations Abstracts International $g 84-09A.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29999754 $z click for full text (PQDT)