View : 977 Download: 0

Classification of Typhoon Track Patterns Affecting the Korean Peninsula Using Data Mining

Title
Classification of Typhoon Track Patterns Affecting the Korean Peninsula Using Data Mining
Authors
김혜진
Issue Date
2017
Department/Major
대학원 통계학과
Publisher
이화여자대학교 대학원
Degree
Master
Advisors
오만숙
Abstract
During each typhoon season, typhoon landfalls cause a great amount of social and economic damage. Classifying typhoon track patterns and predicting the track pattern before entering Korean Peninsula are important to predict vulnerable area by the typhoon. Typhoons affecting the Korean Peninsula develops in the North Pacific Ocean and go northward and then enter the Korean Peninsula via the West Sea, the South Sea or the East Sea. Before predicting types of typhoon track patterns, this paper defines the range of the 'Korean Peninsula' and the range of 'influence to the Korean Peninsula'. With these definitions, this study classifies typhoons into three track patterns based on numerical and official standards. 'W (West)', 'S (South)' and 'E (East)'. The pattern of a typhoon is predicted as 1-day and 2-day forecast before it enters the Korean Peninsula. We use classification methods used in Data Mining such as Ensemble methods (Bagging, Random Forest, GBM) and SVM. The performance of prediction model is measured by mean of Cross-Validation error.;태풍의 경로를 유형화하여 한반도 진입 전에 예측하는 것은 해당 태풍으로 인한 향후 피해 지역을 예측할 수 있게 해준다. 열대 해상에서 발생하여 북상하는 태풍은 한반도에 서해, 남해 또는 동해를 통해 진입한다. 본 논문에서는 수치적인 기준으로 '한반도'의 범위와 3개 진입 방향을 규정한 후, 한반도에 영향을 미치는 태풍의 경로를 3개로 유형화하였다. 이에 대한 multi-class classification 방법으로 앙상블기법(Bagging, Random Forest, GBM)과 SVM을 사용하였다. 또, 이를 한반도 상륙 1일전인 위도 31°와 2일전인 위도 28°에서 예측하였다.
Fulltext
Show the fulltext
Appears in Collections:
일반대학원 > 통계학과 > Theses_Master
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

BROWSE