基于RF-LR的高速公路逃费车辆状态预测模型

State Prediction Model of Expressway Escaping Vehicles Based on RF-LR

【目的】为严格维护高速公路运营秩序和提高高速公路运行效率,对逃费车辆进行精准稽查以追缴过往车辆偷逃通行费。【方法】通过分析高速公路车辆通行卡大数据,采用随机森林(Randomforest,RF)筛选车型、车轴组数、总质量、超载率和通行费共5个逃费特征,经过虚拟化类型变量,利用合成少数类过采样技术(Syntheticminorityoversamplingtech-nique,SMOTE)算法平衡数据集,建立了逻辑回归(Logisticregression,LR)逃费车辆状态预测模型。【结果】基于RF-LR的高速公路逃费车辆状态预测模型有利于挖掘提取大数据中的有用信息,对逃费车辆状态预测具有较高的精度,预测正确率达到91.74%,预测精准率达到91.95%,召回率达到99.25%,预测性能较好。【结论】基于RF-LR的高速公路逃费车辆状态预测模型较以往简单预测模型有较大改进,消除了预测结果过拟合现象,提高了预测精度,预测结果可为高速公路运营管理提供参考,提高工作效率,实现对偷逃通行费行为的快速稽查。

[Purposes]In order to strictly control the operation order of expressway,improve the operation efficiency of expressway,check the vehicles of toll evasion accurately,and recover the tolls for passing vehicles.[Methods]By analyzing the big data of expressway traffic pass cards,using the random forest(RF)algorithm to screen the five escaping features of vehicle type,number of axle groups,total weight,overload rate,and toll.Through the virtualization of type variables and the balance of the data set with SMOTE algorithm,the logit regression(LR)state prediction model of toll evasion vehicles was established.[Findings]The experimental results show that the RF-LR based on prediction model is conducive to extract useful information from big data.The forecast of the state of toll escaping vehicles is of high accuracy,the predictive accuracy rate reaches 91.74%,the predictive precision rate reaches 91.95%,the recall rate reaches 99.25%.[Conclusions]The RF-LR based on predictive model is an improved version of the previous simple prediction model.It can also eliminate the over fitting phenomenon of prediction results and improve the prediction accuracy.The prediction of the model can provide reference for expressway operation and management,improve work efficiency,and realize rapid inspection of toll evasion.