基于随机森林算法的洪水灾害风险评估研究

根据流域灾害系统理论,在考虑致灾因子、孕灾环境和承灾体的基础上,选取9个风险评价指标,运用样本数据进行人工识别风险并得到训练样本,采用随机森林算法构建基于随机森林的洪水灾害风险评估模型。然后采用随机森林自评估工具,分析建立的洪水灾害风险评估模型的误差和指标,同时构建支持向量机模型作为对比方案,并采用五折交叉验证方法对基于随机森林算法的洪水灾害风险评估模型和支持向量机模型进行验证。最后以海河流域邱庄段为研究对象,分别运用基于随机森林算法的洪水灾害风险评估模型和基于支持向量机模型对相同的数据集进行评估和对比,结果显示,12 h内降雨总量、洪水持续时间和土壤含水量是引发洪水的主要因素,而基于随机森林算法的洪水灾害风险评估的训练精度及测试精度均高于支持向量机模型。

Risk assessment of flood disasters based on random forests

According to the theory of catchment disaster system, firstly 9 risk evaluation indices are selected considering disaster-causing factors, disaster-pregnant environment and disaster-bearing bodies, the sample data are used to manually identify risks and to obtain training samples, and a risk assessment model for flood disasters is established by using the random forest algorithm based on the random forests. Secondly, the self-assessment tools of the random forests are employed to analyze the model errors and indexes, and the support vector machine model is established as a comparison scheme. The 5-fold cross-validation method is used to validate the above two models. Finally, the Qiuzhuang catchment of Haihe River Basin is taken as the research object to evaluate and compare the evaluation models based on the random forests and the support vector machine with the same data. The results show that the total rainfall in 12 hours, duration of floods and water content of soils are the main factors to cause floods. The training and testing accuracies of the risk assessment of flood disasters based on the random forests are both higher than those of the support vector machine.