以崩塌率依據建構邏輯式迴歸崩塌潛勢評估模式

The Landslide Susceptibility Model Based on Landslide Ratio by Using Logistic Regression

10.29417/JCSWC.201412_45(4).0005

吳俊鋐(Chun-Hung Wu)

邏輯式迴歸 ； 崩塌潛勢模式 ； 旗山溪集水區 ； 崩塌率 ； Logistic regression ； landslide susceptibility model ； Chishan watershed ； landslide ratio

中華水土保持學報

45卷4期（2014 / 12 / 01）

257 - 265

繁體中文

本研究以2009年莫拉克颱風後崩塌分佈，探討以一般方式及以崩塌率為依據方式建構邏輯式迴歸崩塌潛勢評估模式差異。研究結果顯示以崩塌率為依據方式可提高模式正確率5.0%，並降低原一般方式預測錯誤區域面積及修正所產生的兩個誤差：連續變數型資料無法完全模擬崩塌率分佈及類別變數係數與崩塌率相似性不高。本研究建議在以邏輯式迴歸模式建構崩塌潛勢評估模式過程，應先以崩塌率進行分類，方可得更好的預測結果。

The This study builds constructs the a landslide susceptibility assessment model based onthat uses the landslide inventory data induced byfrom 2009 Typhoon Morakot, in 2009, for in the Chishan watershed, and discusses determines the difference between the original logistic regression landslide susceptibility model and the logistic regression landslide susceptibility model based onthat uses landslide ratio. The result shows that the correct rateaccuracy of the logistic regression landslide susceptibility model based onthat uses the landslide ratio is 5% betterbetter than the original logistic regression landslide susceptibility model by 5.0%, and can revisecorrects two errors produced that occur in the original logistic regression landslide susceptibility model. One of these two errors is that the continuous variables can't do not simulate the distribution of the landslide ratio, and the other is that the similaritythere is some disagreement between the values of the coefficients values offor categorical variables and the landslide ratio is not high. The This study suggests shows that classification based onthat uses the landslide ratio in the processes of building theto construct a logistic regression landslide susceptibility model can results in a better prediction.