http://scholars.ntou.edu.tw/handle/123456789/19373
DC Field | Value | Language |
---|---|---|
dc.contributor.author | 許振嘉 | en_US |
dc.contributor.author | 魏志強 | en_US |
dc.date.accessioned | 2021-12-17T06:03:47Z | - |
dc.date.available | 2021-12-17T06:03:47Z | - |
dc.date.issued | 2017-04 | - |
dc.identifier.uri | http://scholars.ntou.edu.tw/handle/123456789/19373 | - |
dc.description.abstract | 臺灣為海島型國家,北迴歸線經過且是熱帶與副熱帶氣候的交界處,在此緯度 亦為颱風之路徑,在臺灣較常受到颱風侵襲之季節為夏、秋兩季。颱風帶來豐沛的 雨量,為臺灣帶來不可或缺的水資源,但暴風雨所帶來的災害,則嚴重影響人們生 命財產的安全。又因每次颱風路徑不逕相同,且從不同方向登陸,受地形阻擋的影 響,使得風向、風速或海面風浪也有不同的變化,在颱風侵襲期間,若能準確地預 測浪高,則可以降低沿岸災害及保障人身安全,達到未雨綢繆的效果。本研究利用 邏輯斯回歸分類模型及機器學習之隨機森林法建立颱風警報時期的波浪預測,本 研究以臺灣東北部宜蘭、彭佳嶼測站和龜山島浮標為測試例。本研究收集中央氣象 局所屬的氣象站及浮標歷史觀測資料與颱風警報資料,資料年限為西元 2005 至 2012 年,選定海象、氣象與颱風相關的參數結合三種演算法建立整體學習。 | en_US |
dc.language.iso | zh | en_US |
dc.subject | 颱風 | en_US |
dc.subject | 波浪 | en_US |
dc.subject | 隨機森林 | en_US |
dc.subject | 邏輯斯回歸 | en_US |
dc.subject | 預測 | en_US |
dc.title | 應用機器學習法於颱風風浪分類之研究 | en_US |
dc.type | conference paper | en_US |
item.cerifentitytype | Publications | - |
item.openairetype | conference paper | - |
item.openairecristype | http://purl.org/coar/resource_type/c_5794 | - |
item.fulltext | no fulltext | - |
item.grantfulltext | none | - |
item.languageiso639-1 | zh | - |
crisitem.author.dept | College of Ocean Science and Resource | - |
crisitem.author.dept | Department of Marine Environmental Informatics | - |
crisitem.author.dept | National Taiwan Ocean University,NTOU | - |
crisitem.author.dept | Center of Excellence for Ocean Engineering | - |
crisitem.author.dept | Data Analysis and Administrative Support | - |
crisitem.author.orcid | 0000-0002-2965-7538 | - |
crisitem.author.parentorg | National Taiwan Ocean University,NTOU | - |
crisitem.author.parentorg | College of Ocean Science and Resource | - |
crisitem.author.parentorg | National Taiwan Ocean University,NTOU | - |
crisitem.author.parentorg | Center of Excellence for Ocean Engineering | - |
Appears in Collections: | 海洋環境資訊系 |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.