支援向量機於水域生態環境評估

Application of Support Vector Machine in Assessment of Aquatic Ecological Environment

陳盈守(Ying-Shou Chen)

指導教授 ： 張斐章

國立臺灣大學/生物資源暨農學院/生物環境系統工程學研究所/碩士(2005年)

https://doi.org/10.6342/NTU.2005.01123

摘要

本研究結合水文與水質等多項因子與魚類調查資料作為河川水域生態之評估項目，首先將樣區魚類出現的種類轉化為魚類評估性指標，再藉由支援向量機(SVM)來模擬魚類評估指標和水域環境監測因子之間的相互關係，以建立水域生態評估模式。支援向量機(SVM)為一新興之機械學習方法，具有完備的統計理論與最佳化過程，近年來被廣泛應用於各項科學之資料分類或推估問題；本研究以支援向量機理論為依據建構水域評估模式，並與K-means、Fuzzy C-means、Discrimination等傳統分類或辨識模式作比較。
本研究選用嘉義縣的赤蘭溪和台中縣的筏子溪作為研究區域，經赤蘭溪資料驗證結果顯示，支援向量機(SVM)於水域生態之評估，效果明顯優於其他傳統分類模式。使用赤蘭溪和筏子溪的資料進行測試，其正確率亦高達85%，顯示支援向量機(SVM)的確能有效地利用過去環境的水文與水質監測資料配合魚類調查資料來建立良好的水域生態評估模式。

關鍵字

支援向量機 ； 魚類評估指標 ； 水域生態

並列摘要

The main purpose of this study is to propose an aquatic ecological assessment model by the Support Vector Machine (SVM). A number of hydrological factors and water quality with investigated information of fishes are first collected and used as the base of assessment of river ecological condition. The type of collected fishes in a specific river section is then transferred into fish indexes to indicate the suitability of the investigating aquatic ecology. The SVM is then used to construct the aquatic assessment model based on the hydrological factors and water quality information.
The SVM is a linear machine rooted in the statistical learning theory to construct a hyperplane as the decision surface in such a way that the margin of separation between positive and negative examples is maximized. The SVM has been broadly used in solving many scientific problems, especially in pattern classification and nonlinear regression. For the purpose of comparison, three traditional classifying methods, i.e. K-means, fuzzy C-mean, and discrimination, were also used to construct the assessment model.
The measured hydrological data and investigated fishes data obtained in Chihlan creek and Fazih creek River were used to train and test the models. The results demonstrate that the SVM get much better performance than the three traditional classify methods. The cross validation results of accuracy in both creeks obtained by SVM are over 91%. The results suggest that the SVM is a powerful and suitable tool for building an aquatic ecological assessment model.

並列關鍵字

Support vector machine ； fish indexes ； aquatic ecological

參考文獻

20. 肖篤寧，1992，「景觀生態學理論、方法及應用」，地景企業股份有限公司。

35. 曾國源，2003，「以類神經網路架構土石流預警系統之研究」，國立台灣大學生物環境系統工程學研究所論文。

10. 李明儒、林信輝等，2003，「生物整合指標(IBI)之應用探討」，水土保持學報35(1):81-96。

41. Bernhard Scholkopf and Alexander J. Smola, 2002, “Learning with kernels: support vector machines, regularization, optimization, and beyond”, Cambridge, Mass.: MIT Press.

42. Bezdek JC., 1973, “Fuzzy Mathematics in Pattern Classification. PhD. Thesis. Applied Mathematics Center”, Cornell University.