Importance analysis of building components for earthquake losses based on the random forest algorithm
Earthquake damage to buildings always starts from a damaged component,so the importance analysis of building components helps identify the weakest component in the structure.Traditional researches often focused on structural components,and the importance of non-structural components was not well studied,resulting in the loss of non-structural components being much larger than the loss by structural components.In this paper,based on a large amount of detailed component earthquake loss data from the New Zealand earthquake,three importance analysis methods from random forest,namely the feature importance(FI),the permutation importance(PI)and the SHAP value(shap_values),were used to rank the importance of each component for earthquake losses.Assignment sort summation and normalized summation were used to assemble the results from the three methods representing the combined importance of each component.The normalized summation method can not only identify the relative importance of the components,but also quantify the significance of the differences among different components.The results demonstrate that of the 12 types of components analyzed,non-structural components are more important than structural components in earthquake loss contribution,and wall ornaments are the most important component,followed by slab foundations,but roof framing is the least influential.The results indicate that future earthquake resistance efforts should put more focus on wall ornaments and slab foundations.
万方数据
维普
