Identifying strongly correlated groups of sections in a large motorway network

In a motorway network, correlations between the different links, i.e. between the parts of (different) motorways, are of considerable interest. Knowledge of fluxes and velocities on individual motorways is not sufficient, rather, their correlations determine or reflect, respectively, the functionality of and the dynamics on the network as a whole. These correlations are time dependent as the dynamics on the network is highly non-stationary, as it strongly varies during the day and over the week. Correlations are indispensable to detect risks of failure in a traffic network. Discovery of alternative routes less correlated with the vulnerable ones helps to make the traffic network robust and to avoid a collapse. Hence, the identification of, especially, groups of strongly correlated road sections is needed. To this end, we employ an optimized $k$-means clustering method. A major ingredient is the spectral information of certain correlation matrices in which the leading collective motion of the network has been removed. We identify strongly correlated groups of sections in the large motorway network of North Rhine-Westphalia (NRW), Germany. The groups classify the motorway sections in terms of spectral and geographic features as well as of traffic phases during different time periods. The representation and visualization of the groups on the real topology, i.e. on the road map, provides new results on the dynamics on the motorway network. Our approach is very general and can also be applied to other correlated complex systems.