Feature extraction of urban traffic network data based on Local Tangent Space Alignment

This paper introduces a nonlinear dimensionality reduction method based on manifold learning, namely the Local Tangent Space Alignment, into the field of feature extraction in urban traffic networks. By constructing an approximation for the tangent space at each data point in high dimensionality, this method is capable of learning the local geometry of the manifold. Those tangent spaces are then aligned to give the global coordinates of the data points in a lower-dimensional space with respect to the underlying manifold. Experiment shows that compared with other dimensionality reduction methods such as Principal Component Analysis and Locally Linear Embedding, LTSA has a more impressing performance in extracting as well as visualizing the spatiotemporal features of the traffic networks.