基于XGBoost-SHAP模型的高速公路合流区通行能力影响因素研究

Study on the Influence Factors of Highway Merging Zone Capacity Based on XGBoos t- SHAP Modeling

设计了合流区通行能力仿真实验,收集不同影响因素下通行能力数据,利用XGBoost模型搭建通行能力计算模型并通过Optuna框架进行超参数优化,融合SHAP可解释模型对各因素的影响程度及贡献进行量化。结果表明,随着大型车比例和匝道汇入流量的增加,对合流区通行能力的负向影响逐渐增加,展现出很强的负向关联,而加速车道长度对合流区通行能力的影响在0~250 m内呈现出正向关联,超出250 m,两者不再呈现相关性。因此,通过管控汇入流量和提高加速车道长度可显著提升合流区通行能力。

This paper designs a simulation experiment on the capacity of merging area,collects the capacity data under different influencing factors,builds a capacity calculation model by using XGBoost model and optimizes the hyperparameters through Optuna framework,and integrates the SHAP interpretable model to quantify the degree of influence and the contribution of each factor.The results show that the negative impact on the merging zone capacity gradually increases with the increase of the proportion of large vehicles and ramp admittance flow,showing a strong negative correlation,while the impact of the acceleration lane length on the merging zone capacity shows a positive correlation in the range of 0~250 m,and the two no longer show correlation beyond 250 m.Therefore,the merging zone capacity can be significantly improved by controlling the converging flow and increasing the length of the acceleration lane.