Data compression techniques for urban traffic data

Abstract

With the development of inexpensive sensors such as GPS probes, Data Driven Intelligent Transport Systems (D[superscript 2]ITS) can acquire traffic data with high spatial and temporal resolution. The large amount of collected information can help improve the performance of ITS applications like traffic management and prediction. The huge volume of data, however, puts serious strain on the resources of these systems. Traffic networks exhibit strong spatial and temporal relationships. We propose to exploit these relationships to find low-dimensional representations of large urban networks for data compression. In this paper, we study different techniques for compressing traffic data, obtained from large urban road networks. We use Discrete Cosine Transform (DCT) and Principal Component Analysis (PCA) for 2-way network representation and Tensor Decomposition for 3-way network representation. We apply these techniques to find low-dimensional structures of large networks, and use these low-dimensional structures for data compression.