頁籤選單縮合
題 名 | SOM類神經網路應用於降雨特性均一區劃分之研究=Using SOM Neural Network to Delimitate the Homogeneous Regions of Rainfall Properties |
---|---|
作 者 | 楊昌儒; 蔡長泰; 游保杉; | 書刊名 | 中國土木水利工程學刊 |
卷 期 | 11:2 1999.06[民88.06] |
頁 次 | 頁337-347 |
分類號 | 443.1 |
關鍵詞 | SOM類神經網路; 降雨特性均一區; SOM neural network; Homogeneous regions of rainfall properties; |
語 文 | 中文(Chinese) |
中文摘要 | 本文以臺灣水資源分區之北部地區38個雨量站之降雨特性資料為例,選取代表降 雨時空分佈特性之29個特徵變量,組成訓練範例,以SOM網路進行降雨特性均一區劃分,讓 網路依據訓練範例之內在特性,以非監督方式學習,自我組織、聚類,將相似降雨特性之測 站聚集在一起。經網路學習後,可將此38個測站劃分為3個降雨特性均一區,而此分區結果 可通過魏氏檢定,顯示聚類間具有明顯差異性;而且由降雨特性均一區與氣侯區域套疊結果 顯示,降雨特性均一區與氣侯區域相近,此顯示以SOM類神經網路,進行降雨特性均一區劃 分之合理性與可行性。 |
英文摘要 | This study is to conduct a regional analysis to delimitate the homogeneous regions of rainfall properties based on 38 recording rain gages located in water resources region of the northern Taiwan. A total of 29 analytic variables were selected, which represent the space positions of rain gages, annual rainfall variation and rainfall pattern, rainfall characteristics of the 38 recording rain gages. The SOM neural network was used to capture the native attributes between analytic variables. Through the unsupervised training procedure, three homogeneous regions were determined. All these regions passed the Wilk's test, indicating that the classification was accepatable and the identified regions corresponded well to the climate regions. It is concluded that the SOM neural network can be used as an effective means for identifying the homogeneous regions of rainfall properties. |
本系統中英文摘要資訊取自各篇刊載內容。