基于CART回归树模型的深基坑施工安全事故分析与预测

Analysis and Prediction of Deep Excavation ConstructionAccidents Based on CART Model

深基坑施工一直被称为“刀尖上的舞蹈”,并日益成为施工安全事故最为频发的领域。本研究根据事故表征分析方法构建了深基坑施工事故表征信息的指标体系,采用CART回归树模型分析预测深基坑施工事故的分类情况。研究表明深基坑施工事故可以划分为决策树深度为9的交叉树形结构,并且是否存在违法招投标、基坑等级、项目所处区域、施工单位规模等事故表征信息对其影响较大,相较于一般的分类决策树算法,改进后的极端梯度提升算法的分类精度更高。研究结果丰富了深基坑施工事故研究,并为施工安全管理提供了理论指导和分析框架。

Concerning the high frequency of construction accidents,deep excavation engineering is one of the most dangerous works in terms of safety.This study aims to diagnose a characterization of the accident patterns in deep excavation construction projects.The classification and regression tree(CART)is used to diagnose accidental characteristics associated with deep excavation,and nine tree levels are identified.The findings indicated common characteristics of accident occurrences,including whether illegal bidding or not,deep excavation grade,location of construction project,and scale of construction enterprise.The use of extreme gradient boosting improves classification accuracy compared with the conventional decision tree algorithm.This study not only deepens the understanding of deep excavation construction accidents but also establishes a theoretical and practical framework for construction safety management.