應用案例式推理法推估地區降雨量研究

本研究將2004至2007年間之975日降雨資料依大氣壓力、風速、溼度、溫度、輻射、風向等屬性，建立起資料庫，依序以案例式推理法推估每日降雨量，將每日之推估降雨量與實際降雨量作比較，求其相對誤差，進而來探討苗栗地區降雨量的準確性。研究過程係以歐幾里得距離法及權重法計算相似度，藉比較相似度進行雨量推估及分析。研究結果中，以歐幾里得距離法求相似度時，所得推估降雨量與實際降雨量之相對誤差總平均為29.46%；以權重法求相似度時，所得推估降雨量與實際雨量之相對誤差總平均為24.59%。此方法可提供未來降雨量預測上參考。

關鍵字

地區降雨；權重相似度法；歐幾里得距離法

並列摘要

In this study the 975 daily rainfall data of Miaoli area from 2004 to 2007 were set up in an information file with the atmospheric pressure, the wind velocity, the humidity and temperature, the radiation, the wind direction and rainfall quantity etc. The Case-Based Reasoning method was used to estimate the rainfalls day by day and then compared the estimated rainfalls with the observed rainfalls to find the relative errors. Then the accuracy of the estimated rainfalls at Miaoli area were discussed.The similarities were calculated from the Euclidean distance method and the weighted method in this study. By comparing the similarities of the data, the rainfalls were estimated and analyzed. For the results, when the Euclidean distance method was used to find the similarity, the total average error between the estimated rainfalls and the observed rainfalls is 29.46%; when the weighted method was used to find the similarity, the total average error between the estimated rainfalls and the observed rainfalls is 34.37%.The method of this study can be used as a reference for the rainfall prediction in the future.