基于机器学习的不可移动文物暴雨灾害风险评估——以山西省为例

Machine learning-based storm disaster risk assessment of immovable cultural relics:A case study of Shanxi Province

以山西省为例,将不可移动文物作为灾害的影响对象,通过分析暴雨灾害对其的影响因素构建指标体系,通过文物损毁情况划分风险并作为输出数据,将山西省6325处受灾不可移动文物按照6∶2∶2划分训练集、验证集和测试集,分别使用随机森林模型、支持向量机模型(support vector machine,SVM)、逻辑回归模型进行训练和验证,选取出最优模型,并对山西不可移动文物进行不同重现期暴雨下的风险预测。研究结果表明:基于机器学习的不可移动文物暴雨风险评估方法是可行且结果较为优异的;在3种模型中,随机森林的验证集准确率最高,为95.75%,测试集精度为94.70%;在山西省5、20、50 a重现期暴雨下的不可移动文物风险均呈现出北方低、南方高的态势,且在5 a重现期下高风险文物占比最多,高风险文物以古建筑和古遗址为主。

The immovable cultural relics in Shanxi Province were taken as the object affected by the rainstorm disasters,an index system was constructed by analyzing the factors affected by the rainstorm disasters,the risk was classified by the damage of cultural relics and it is used as the output data,6325 affected immovable cultural relics in Shanxi Province were divided into training set,validation set and test set according to 6∶2∶2,and the random forest model,support vector machine model and logistic regression model were used respectively.The optimal model was selected and the risk prediction of immovable cultural relics in Shanxi Province under different recurrence periods of heavy rainfall was carried out.The results show that the machine learning-based storm risk assessment method for immovable cultural relics is feasible and the results are significant.Among the three models,the random forest has the highest accuracy of 95.75%in the validation set and 94.70%in the test set.The risk of immovable cultural relics in north is low while high in south under the 5,20,and 50 a return periods in Shanxi Province,and the proportion of high-risk relics under the 5 a return period is the highest.The high-risk cultural relics under the 5 a recurrence period,and the high-risk cultural relics are mainly ancient buildings and ancient sites.