支援向量機於降雨引致崩塌潛勢分析之研究

RAINFALL INDUCED-LANDSLIDE SUSCEPTIBILITY ANALYSIS USING SUPPORT VECTOR MACHINE

黃雅喬(Ya-Chiao Huang) ； 林國峰(Gwo-Fong Lin) ； 張明瑞(Ming-Jui Chang) ； 何瑞益(Jui-Yi Ho)

《中國土木水利工程學刊》 28卷1期 (2016/03) Pp. 57-66

https://doi.org/10.6652/JoCICHE.201603_28(1).0006

摘要

台灣山坡地範圍占全面積73.6%，颱風侵襲帶來的大量強降雨，易於山區產生嚴重的崩塌災害，造成社會經濟損失與危害人民生命安全。崩塌發生的影響因子眾多，且因子間具高度非線性關係，本研究首創選用在處理非線性問題具有高精度與高效能之支援向量機，探討因子與崩塌之相互關係，建置高屏溪流域崩塌潛勢評估模式。結果顯示，崩塌模擬之準確率為92.9%，總準確率為71.3%，且AUC值高達0.792，顯示所建立模式可提供良好的崩塌預測。此外，可根據此崩塌潛勢評估模式所得之崩塌潛勢圖，提前掌握高屏溪流域中可能之崩塌區域，以減少財產損失與人員傷亡。

關鍵字

降雨 ； 崩塌潛勢 ； 支援向量機

並列摘要

Landslides usually occur in Taiwan and result in property losses and casualties due to the large hilly and mountainous area with high steepness and heavy rainfall caused by typhoons. The occurrence of landslide is a very complex and highly nonlinear phenomenon. The support vector machine, which performs precisely and efficiently in nonlinear problems without too many assumptions, is firstly employed herein to develop a landslide susceptibility model for the Kao-Ping River basin in southern Taiwan. The model performance is checked using the confusion matrix and the area under the receiver operating characteristic curve (AUC). The validation results show that the true positive rate, the accuracy, and the AUC are 92.9%, 71.3% and 0.792, respectively. This indicates that the proposed model can provide reasonable landslide prediction. The proposed model can also provide the landslide susceptibility map which will be a helpful tool for landslide warning.

並列關鍵字

rainfall ； landslide susceptibility ； support vector machine