Risk Assessment on Railway Line in Mountainous Area under Rainstorm and Flood Situation Based on Random Forest Algorithm:A Case Study of the Yuanping Section of Shuohuang Railway
The split process of random forest decision tree is affected by the unbalance and variability of data set,which will lead to the bias of weight allocation and prediction results.For railway in mountainous area,the influencing factors of flood risk are numerous,and the differences between data are large.How to improve the influence of data imbalance and variability is directly related to the ranking of the importance of each factor and the accuracy of evaluation results.In this regard,nine risk indicators from disaster causing factors and disaster pregnant environment were selected based on the theory of watershed disaster system to construct the flood risk assessment data set for railway in mountainous area.On this basis,the improved random forest algo-rithm based on Gini split function and Sigmoid split function were respectively used for model training and comparative analy-sis.The optimal model was used to forecast the flood risk along the Yuanping section of Shuohuang Railway,and the results of field investigation were used to verify the forecast results.The results show topographic factors(elevation,slope and aspect)are closely related to flood risk.As the internal cause of flood risk,disaster environment plays a decisive role in the flood risk as-sessment of railway in mountainous area.The improved random forest algorithm based on Sigmoid split function can improve the information purity,reduce the data variation and improve the accuracy during the split process,and the prediction results are in good agreement with the field situation.The research results can provide a reference for the risk assessment of railway in mountainous area under rainstorm and flood situation.
mountain railwaysrandom forestrisk assessmentrainstorm and flood
万方数据
