Predicting Urban Mobility

A very curious study has been published by some Chinese researchers few days ago. According to the researchers, urban mobility and transportation system can be viewed as complex network with time-varying traffic flows as links to connect adjacent regions as networked nodes. By computing urban traffic evolution on such temporal complex network with PageRank, it is found that for most regions, there exists a linear relation between the traffic congestion measure at present time and the PageRank value of the last time. Since the PageRank measure of a region does result from the mutual interactions of the whole network, it implies that the traffic state of a local region does not evolve independently but is affected by the evolution of the whole network. As a result, the PageRank values can act as signatures in predicting upcoming traffic congestions. The researchers observe the aforementioned laws experimentally based on the trajectory data of 12000 taxies in Beijing city for one month.

It is the first time that PageRank is applied to model and predict traffic state evolution from a global point of view. Although there exist other methodologies such as deep learning theory to predict traffic congestions on large-scale transportation networks, the mechanism regarding how urban traffic evolves is not explicitly visible, nor explainable, due to the complex coupling within the deep neural networks. In contrast, we obtain an explicit linear relationship between the PageRank value and the upcoming congestion degree.

Previous works on traffic modeling and urban mobility are mostly based on such an assumption that the traffic state at a given site is only dependent on the traffic flow propagations from its nearby regions. However, the big data available such as the GPS traces of a large number of taxis in a city allow a new point of view to revisit the urban traffic modeling problem. This study aims to reveal whether and how the dynamical behavior of the entire urban transportation system does affect the evolution of the congestion degree of a local region. The distinctions and contributions of this study can be summarized as follows: (1) Urban transportation system is modeling as temporal complex network with time-varying directional links weighted by the traffic flows entering and leaving a region, where the whole city is partitioned into a couple of blocks acting as the nodes. (2) researchers employ the PageRank algorithm to simulate the evolution of the established network model, where the traffic flows into and out of a region are viewed as influence delivery and the global impact arising from such influence delivery between adjacent regions is computed in the framework of PageRank. (3) It is found that the congestion degree of a local region is not only affected by the traffic states of its neighboring regions but also those of the whole network. (4) It is verified experimentally that there exists a positively linear relation between last-time PageRank value and current congestion index for most regions such that PageRank indices can act as signals to warn upcoming traffic congestions.

Read the full paper at: