网络行程时间可靠性评价方法与影响因素
Evaluation method and influence factors of network travel time reliability

采用区域划分方法研究了网络行程时间率的概率分布,提出了基于OD对的网络行程时间可靠性指标以评价城市交通可靠性; 选取影响行程时间可靠性指标的相关因素,建立了多元线性回归模型,用逐步回归法求解模型,并进行了模型参数显著性检验; 根据杭州市和北京市的网约车数据计算了网络行程时间可靠性指标,并与高峰拥堵延迟指数进行对比,分析了网络行程时间可靠性指标的时间和空间分布规律。研究结果表明:在多元线性回归模型中,规划行程时间率与等待时间、费用、距离、行程时间和OD对间行程次数这5个自变量拟合得到的决定系数为0.772,平均行程时间率与5个自变量拟合得到的决定系数为0.857,2个模型拟合程度均较好,回归模型显著; 规划行程时间率回归模型中等待时间、行程时间和实际行程距离的回归系数分别为0.386、0.399与-1.286,平均行程时间率回归模型中等待时间、行程时间和实际行程距离的回归系数分别为0.162、0.177与-0.676,2个交通可靠性指标都与等待时间和行程时间呈正相关,和实际行程距离呈负相关; 提出的网络行程时间可靠性指标与高峰拥堵延迟指数变化趋势一致,较好地符合现实交通状况,从多角度反映了交通可靠性特征,可以为路网规划提供决策支持,帮助居民更好地进行出行路径选择。
The probability distribution of network travel time rate was studied based on the regional division approach, and the travel time reliability indexes based on OD pairs were proposed to appraise the reliability of urban traffic. The multivariate linear regression model was established by choosing the relevant factors that influenced the travel time reliability indexes. The model was solved by the stepwise regression method, and the significance test was conducted to verify the estimated model parameters. The network travel time reliability indexes were calculated by Hangzhou and Beijing ride-hailing data and compared with the peak congestion delay indexes, then the temporal and spatial distributions of network travel time reliability indexes were analyzed. Research result shows that in multivariate linear regression models, the fitting determination coefficient between the planning travel time rate and five independent variables, including the waiting time, cost, distance, travel time, and number of trips for OD pairs is 0.772, and the fitting determination coefficient between the average travel time rate and five independent variables is 0.857, so both models have better fitting degrees and statistical significance. In the regression model of planning travel time rate, the regression coefficients of waiting time, travel time, and travel distance are 0.386, 0.399, and -1.286, respectively. In the regression model of average travel time rate, the regression coefficients of waiting time, travel time, and travel distance are 0.162, 0.177, and -0.676, respectively. The two traffic reliability indexes are positively correlated for the waiting time and travel time, and negatively correlated for the actual travel distance. The proposed network travel time reliability indexes are consistent with the peak congestion delay index and reflect the traffic reliability characteristics from various perspectives. They provide decision support for transportation planning and help residents choose propitious routes. 5 tabs, 7 figs, 26