基于两阶段支持向量机的群体建筑物震害预测方法

Seismic Damage Prediction Method of Building Groups Based on a Two-stage Support Vector Machine

传统群体建筑物震害预测,多是采用与现有的建筑物震害数据类比的方法。由于特殊地质环境和特定地震情景的影响,外加人工统计的误差干扰,现有数据中存在相当数量的异常数据指标。这些数据噪声将严重影响群体建筑物震害预测的准确度。引入一种新型两阶段支持向量机方法,首先为正常数据和异常数据赋予不同的权重,接着用加权支持向量机方法建立群体建筑物震害预测模型。通过对汶川地震中640栋建筑进行交叉验证发现,提出的两阶段支持向量机方法不仅能有效识别出异常数据点,而且能快速准确地预测建筑物震害结果,可以用于实际的城市建筑物震害预测工作。

In conventional seismic damage prediction of building groups, the researchers always make a simple analogy with existing data of damaged buildings. Due to the influence of special geological environment,particular earthquake scenarios, together with the errors in handiwork statistics, there are a certain amount of outliers in dataset. The random noise in the dataset will have a serious impact on prediction accuracy. Thus, this paper introduces a two-stage support vector machine method. In the first step, the authors add different weightvalues to normal data and outliers respectively. Then a weighted support vector machine is proposed to build the prediction model of building groups. By using a cross-validation approach, the paper empirically tests the proposed model on 640 buildings in Wenchuan earthquake. The results show that the proposed method can not only effectively detect the outliers, but also make a fast accurate prediction. It is capable to be applied to the actual seismic damage prediction of urban buildings.