以亂數基礎分類法和Fuzzy-C-means分群法分析土石流判釋問題

Resolving the Discrimination on the Occurrence of Debris-flow Factor Analysis through Entropy-Based Classification and Fuzzy C-mean Approaches

萬絢(Shiuan Wan) ； 閻嘉義(Jia-Yi Yen)

《水保技術》 4卷1期 (2009/01) Pp. 37-46

https://doi.org/10.30040/JSWCT.200901.0004

摘要

過去通常以統計的方式將颱風或豪大雨中，發生土石流與未發生的集水區，以歸納出土石流發生的地文條件與水文條件，找出發生與未發生的潛勢溪流中，各因子的範圍值；然而隨著描述土石流發生的因子逐漸增多，過去學者往往以主觀判斷何項為主要因子，但隨科技的進步，集水區量測技術大幅度改善，不過資訊量的增加並不代表描述能力增加，亦有可能加入不必要的因子，甚至代入混淆的因子致使判釋土石流上發生錯誤。本文有鑑於此，以亂數基礎分類法Entropy Based Classification (EBC)和模糊分群法Fuzzy-C mean (FCM)兩套方法處理土石災害主要問題。EBC主要功能是找出影響土石流的主要因子和因子的臨界值，而FCM 是找出影響土石流知識中心和土石流知識界線（垂直平分線），這兩套方法充分解決了主觀判斷之誤，可建立客觀的判識標準。

關鍵字

土石流災害 ； Entropy Based Classification ； Fuzzy-c mean

並列摘要

In the past, significant efforts were dedicated to explore and collect data related to debris flow occurrence such as those pertaining to rainfall, geology, and landforms. These employed (1) the planning of potential area of debris-flows (2) the threshold values of influence factors (3) the investigation of in-situ debris-flows zone. Most of the past study/literature works focused on statistical analysis of searching threshold values and influence factors. The flaws appeared while each of the debris flows occurs at different amount of rainfall of typhoon. The influence factors are determined subjectively. The present study analyzed the debris flow through two stages: (1) an advanced Data-Mining approaches (Entropy-based Classification, EBC) to extract threshold values and influence factors, and (2) using Clustering analysis (Fuzzy C-mean) to generate the knowledge center and separate line of occurrence/non-occurrence debris flow on-site data. Based on this objective process, some of the subjective results rendered statistically by past study are clarified.

並列關鍵字

Debris-flow ； Entropy Based Classification ； Fuzzy-c mean