论文部分内容阅读
为了验证生成式拓扑映射(GTM)算法分析区域交通状态的有效性,对北京市某包含29个检测器的实际路网连续28 d的数据集进行研究。试验中,通过GTM算法将原始数据从58维空间平滑地映射到2维的特征平面,有效实现了高维数据可视化;同时对状态进行无监督聚类和区域交通状态分析,并找出路网异常状态。试验结果表明:GTM算法可以应用在交通状态可视化、线圈故障诊断、交通拥堵的发现、交通状态演化研究等方面。
In order to verify the effectiveness of Generative Topology Mapping (GTM) algorithm in analyzing regional traffic conditions, a data set of 28 consecutive days in Beijing with 29 detectors was studied. In the experiment, the original data is mapped smoothly from the 58-dimensional space to the 2-dimensional feature plane by GTM algorithm, which effectively realizes the high-dimensional data visualization. At the same time, the state is supervised by unsupervised clustering and regional traffic status analysis, Abnormal state. The experimental results show that the GTM algorithm can be applied in the visualization of traffic conditions, the diagnosis of coil faults, the discovery of traffic congestion and the research of traffic status evolution.