基于LightGBM和TOPSIS法的城市轨道交通关键节点识别

Identification of key nodes in urban rail transit based on LightGBM and TOPSIS method

为提高城市轨道交通网络关键节点的识别精度,基于复杂网络理论、交通网络性能特征,考虑轨道交通站点的城市活力信息,建立关键节点评价指标,采用轻量级梯度提升决策树(light gradient boosting machine,LightGBM)机器学习算法计算逼近理想解排序(technique for order preference by similarity to ideal solution,TOPSIS)法中各项评价指标的权重,提出融合LightGBM算法和TOPSIS法的城市轨道交通关键节点识别模型。以杭州城市轨道交通网络为例,针对识别出的前15个关键节点进行动态攻击,通过计算删除关键节点后的网络效率与最大连通子图比例验证模型的识别精度。结果表明:删除前5个关键节点后,网络效率与最大连通子图比例分别为46.89%、56.47%,在一定程度上破坏了轨道交通的网络结构;删除前15个关键节点时,网络效率与最大连通子图比例分别下降至25.61%与16.6%,轨道交通网络基本完全破坏。基于LightGBM和TOPSIS法的城市轨道交通关键节点模型可有效识别交通网络中的关键节点,识别精度较高。

In order to improve the accuracy of identifying key nodes in urban rail transit networks,a city rail transit key node identification model is proposed which combines the light gradient boosting machine(LightGBM) machine learning algorithm and the technique for order preference by similarity to ideal solution(TOPSIS) method based on complex network theory and traffic network performance characteristics,considering the urban vitality information of rail transit stations.Taking Hangzhou urban rail transit network as an example,the identified top 15 key nodes are subjected to dynamic attacks,and the accuracy of the model is verified by comparing the network efficiency and the proportion of the largest connected subgraph before and after the removal of key nodes.The results show that after removing the top 5 key nodes,the network efficiency and the proportion of the largest connected subgraph are 46.89% and 56.47%,respectively,which to some extent disrupt the network structure of rail transit.When the top 15 key nodes are removed,the network efficiency and the proportion of the largest connected subgraph decrease to 25.61% and 16.6%,respectively,which indicates that the rail transit network is almost completely disrupted.The city rail transit key node model based on LightGBM and TOPSIS can effectively identify key nodes in traffic networks with high accuracy.