泥石流危险性评价:模糊c均值聚类-支持向量机法

泥石流是一种能够造成灾难性后果的严重自然灾害，准确可靠的泥石流危险性评价对于其预警及防治工作来说至关重要。泥石流的危险性评价方法有很多，模糊c均值聚类（FCM）方法是其中一种应用广泛的分类方法；相比其他方法而言，其无需主观确定边界，并且能以各级隶属度矩阵为输出结果，方便应用。支持向量机（SVM）是基于结构风险最小化为目标的机器学习理论，以支持向量为算法支撑，具有一定的鲁棒性，并且适合在小样本条件下进行分类。本文选用FCM和SVM联合的方法，开展泥石流危险性的评价；对北京房山区南窖沟泥石流危险性进行分析，并对比其他评价方法所得结果，证明本文提出的评价方法具有较好的效果。

Risk Assessment of Debris Flow: A Method of SVM Based on FCM

Debris flow is a kind of nature hazard which can produce serious consequences, and the accuracy of debris flow risk classification is important for the early warning and disaster prevention.Now there are a lot of algorithms used to evaluate the risk of debris flow. Fuzzy c means (FCM) algorithm is one of the widely used algorithms. Compared with general classification method, FCM does not need to determine risk boundaries artificially, and it outputs membership degree matrix for each risk degree. Support vector machine (SVM) is a machine algorithm based on structure risk minimization. It establishes classification model by support vectors and has better Robustness. It is well applied to small samples. In this paper, the SVM method based on FCM is used for debris flow risk evaluation, which gets a good classification performance.