基于行程时间影响的关键路段识别与查找

从路段实际功能出发，提出基于路段与路径行程时间序列的相关性识别关键路段的方法.借鉴蒙特卡洛思想，以真实数据构造10万条随机路径验证该方法的可行性，并识别出对上海市路网行程时间有关键影响的路段集合.以上述集合为参照，利用模糊聚类及迭代累计平方和算法提取路段行程时间序列特征并构造两个新变量，结合基础属性建立二项Logit模型，从而主动查找关键路段.比较该模型与基础模型、随机分类器查找效果表明：基于最大归一化行程时间曲线聚类，其结果对关键路段识别模型的性能有提升效用；行程时间对数差分序列的结构性变点在路网和路段级别均有明显时间聚集特性，虽然其个数与路段关键性无明显关系，但其与常见波动程度指标相关性小，可保留用于描述行程时间波动常发性和聚集性.

Identification and Retrieval of Critical Segments Based on Travel Time Effect

Considering the actual functions of road segments in the network, a method for identifying critical segments is proposed based on the correlation between the segment and the route travel- time series. Drawing lessons from the Monte Carlo method, 100 thousand random routes created with ground truth data verify this method's feasibility. Several segments that have a crucial impact on the travel time of Shanghai's road network are identified to form a set. Taking the set as a reference, a Binomial Logit Model is built to retrieve critical segments actively, with some new attributes of the segment's travel time series extracted by the Fuzzy Clustering Method and Iterative Cumulative Sums of Squares Algorithm, and some primary attributes as well. The comparison of the model with the basic model and the random classifier shows that the maximum normalized travel time curve's clustering results can improve the critical link identification model's effectiveness. As to structural change points of the travel time logarithmic difference sequence, they show apparent time aggregation characteristics at the road network and road section levels. Although the number of structural change points has no significant correlation with the link's criticality, it has little correlation with the typical volatility index. It can be retained as an index to describe the frequent and aggregation of travel time fluctuations.