论文部分内容阅读
针对城市道路交通流数据(流量、速度、密度或各种综合评价指标等)时间序列具有高维、高噪声的特性,在时间序列维约简的基础上,以5元组的形式对分割后的时间序列进行模式表示,并据此定义5种常见的时间序列形状相似性距离。为确定交通数据时间序列模式相似性度量方法,使用分层聚类算法分析不同相似性距离在交通流状态辨识中的效果。以上海南北高架东侧间部分路段定点线圈检测数据为例,通过对比5种模式相似性距离的聚类效果,发现模式距离效果最优,使用欧式距离和它组合后进行聚类能够实现交通拥挤态势模式相似性和差异性的辨识。
Aiming at the characteristics of high-dimensional and high-noise time series of urban road traffic flow data (traffic, speed, density or various comprehensive evaluation indexes), based on the time series dimension reduction, The time series of the model representation, and thus define five common time series shape similarity distance. In order to determine the similarity of time series of traffic data, hierarchical clustering algorithm was used to analyze the effect of different similarity distances in traffic flow state identification. Taking the fixed-point coil detection data of some sections of the east-north of Shanghai North-South Elevated Elevation as an example, we find that the model distance is the best by comparing the clustering results of the similarity distances of the five modes. Using the Euclidean distances and clustering after combining it can achieve traffic congestion Identification of Similarity and Difference of Situation Patterns.