新增车辆通行拥堵预测模型

Prediction Model of New Increased Vehicle Traffic Congestion

为对新增车辆的通行拥堵进行预测,首先使用K-Medoids聚类算法将交通流运行状态划分为顺畅、阻滞、拥堵三类,然后引入交通流特征参数构建累积Logistic回归模型量化新增车辆对路段运行状态的影响,最后基于支持向量回归机预测新增车辆通行时间。研究结果表明:当只考虑车流量、限行时段和二者之间的交互作用时,模型预测道路状态的正确率达到82.36%,此时车流量在非限行时段每增加一辆车,发生比从顺畅状态转为非顺畅状态的概率是原来的1.087倍;当考虑车流量、黄牌车比例、限行时段、外地车比例及后两者的交互作用时,模型预测通行时间MSE最小,预测效果最优。

In order to predict the traffic congestion of new increased vehicles,the K-Medoids clustering algorithm was firstly used to classify the traffic flow state into smooth,block,and congested;then the traffic flow characteristic parameters were introduced to construct a cumulative Logistic regression model to quantify the impact of new vehicles on the running state of the road segment;finally,the new increased vehicles travel time was predicted based on the support vector regression machine. The research results show that when only the flow rate,the time limit period and the interaction between the two are considered,the correct rate of the model in predicting road state reaches 82.36%;at this time,the probability of traffic flow changing from smooth state to non-smooth state is 1.087 times of the original for each additional car in the non-restricted period;when considering the flow rate,the proportion of yellow card,the time limit period,the proportion of the foreign car and the interaction between the latter two,the model predicts the travel time MSE to be the smallest and the prediction effect is optimal.