基于高斯过程回归模型的洪涝灾害损失预测研究——以重庆市为例

快速准确的预测洪涝灾害各项损失是开展洪涝灾害应急管理工作的基础，而预测技术和方法则是洪涝灾害损失预测的核心与关键。从灾害风险构成因素和数据易获取性2方面构建了洪涝灾害损失评估指标体系，分别从致灾因子、孕灾环境、承灾体和应急能力4个方面选取了13项损失评估输入指标，提出基于高斯过程回归模型的洪涝灾害损失预测方法，并应用于重庆市洪灾受灾人数、农作物受灾面积和直接经济损失的预测。实例表明，高斯过程回归方法对上文提到的3种损失情况预测结果的残差平方和分别为0.99、0.1、12.67，拟合精度分别达到99.85%、99.97%、96.1%，相较于多层感知器神经网络和支持向量机等方法更具优越性。

Study on Flood Disaster Loss Prediction Based on Gaussian Process Regression Model：A Case Study of Chongqing City

Fast and accurate prediction of flood damage is the basis for emergency management of flood disasters, and forecasting techniques and methods are the core and key to flood disaster loss prediction. This paper constructs a flood disaster assessment index system from the aspects of disaster risk component and data accessibility, and selects 13 loss assessment input indicators from disaster-causing factors, affected environment, disaster-bearing object and emergency response capability. A prediction method of flood damage loss based on Gaussian process regression model is proposed, which is applied to the prediction of the flood damage loss in Chongqing of affected number of people, crop affected area and direct economic losses. The example shows that the residual square sum of the three loss cases mentioned above is 0.99, 0.1 and 12.67 respectively, and the fitting precision is 99.85%, 99.97% and 96.1% respectively. Compared with the multi-layer perceptron neural network and support vector machine, the Gauss process regression method is more superior.