In this paper, we investigate the day-to-day regularity of urban congestion patterns. We first partition link speed data every 10 min into 3D clusters that propose a parsimonious sketch of the congestion pulse. We then gather days with similar patterns and use consensus clustering methods to produce a unique global pattern that fits multiple days, uncovering the day-to-day regularity. We show that the network of Amsterdam over 35 days can be synthesized into only 4 consensual 3D speed maps with 9 clusters. This paves the way for a cutting-edge systematic method for travel time predictions in cities. By matching the current observation to historical consensual 3D speed maps, we design an efficient real-time method that successfully predicts 84% trips travel times with an error margin below 25%. The new concept of consensual 3D speed maps allows us to extract the essence out of large amounts of link speed observations and as a result reveals a global and previously mostly hidden picture of traffic dynamics at the whole city scale, which may be more regular and predictable than expected.@en

Today, the deployment of sensing technology permits the collection of massive amounts of spatiotemporal data in urban areas. These data can provide comprehensive traffic state conditions for an urban network and for a particular day. However, data are often too numerous and too detailed to be of direct use, particularly for applications such as delivery tour planning, trip advisors, and dynamic route guidance. A rough estimate of travel times and their variability may be sufficient if the information is available at the full city scale. The concept of the spatiotemporal speed cluster map is a promising avenue for these applications. However, the data preparation for creating these maps is challenging and rarely discussed. In this study, that challenge is addressed by introducing generic methodologies for mapping the data to a geographic information system network, coarsening the network to reduce the network complexity at the city scale, and estimating the speed from the travel time data, including missing data. This methodology is demonstrated on the large-scale urban network of Amsterdam, Netherlands, with real travel time data. The preprocessed data are used to build the spatiotemporal speed cluster by using three partitioning techniques: normalized cut, density-based spatial clustering of applications with noise, and growing neural gas (GNG). A new posttreatment methodology is introduced for density-based spatial clustering and GNG, which are based on data point clustering, to generate connected zones. A preliminary cross comparison of the clustering techniques shows that GNG performs best in generating zones with minimum internal variance, the normalized cut computes three-dimensional zones with the best intercluster dissimilarity, and GNG has the fastest computation time.@en