基于XGBoost模型的不同地区危化品道路运输事故分析

XGBoost Method to Analyze Hazardous Materials Road Transport Accidents from Different Administrative Districts

为降低危化品道路运输事故对人员、财产和环境造成的破坏,保障危化品道路运输的安全性,提出具体有针对性的事故预防措施,探索多种数据挖掘方法去识别不同地区危化品道路运输事故发生的原因。首先,考虑到危化品道路运输事故的稀缺性,将逻辑回归(LR)、支持向量机(SVM)、多层感知机(MLP)和XGBoost模型基于相同的训练和测试数据进行性能评估,确定最佳的危化品道路运输事故数据分析模型;其次,针对我国不同地区之间地理环境、自然资源、人口、经济、危化品需求等方面存在的差异性,利用最佳的分析模型分别探索各事故特征在不同地区对事故严重程度的影响。研究结果表明,XGBoost模型在危化品道路运输事故数据分析中表现最佳,在对事故严重程度有显著影响的特征方面,各地区之间存在一定的差异。同一特征对事故严重程度的影响在不同地区之间也有差异。

In order to reduce the damage to people,property and environment caused by hazardous materials road transportation accidents and to ensure the safety of hazardous materials road transportation,various data mining methods are explored to identify the causes of hazardous materials road transportation accidents in different regions.Firstly,given the scarcity of hazardous materials road transportation accidents,the performance of logistic regression (LR),support vector machine (SVM),multilayer perceptron (MLP) and XGBoost models was evaluated,based on the same training and testing data to find the best analytical models.Secondly,to address the variability in geographic location,natural resources,population,economy,and demand for hazardous materials among different regions in China,the best analysis model was used to analyze the impact of each accident feature on the severity of accidents in different regions.The results indicate that the proposed XGBoost method has the best modeling performance.There is some variation between regions in the features that have a significant impact on accident severity.The influence of the same feature on the severity of an accident even varies from region to region.