Exploring spatio-temporal traffic performance variation through clustering of descriptive travel time statistics

Characterizing link-types and day-types in road networks is vital for understanding recurring traffic state patterns. Link-types and day-types in road networks describe road segments and days based on their specific characteristics. For long-term planning, clustering can be used to categorize links and days with similar characteristics and patterns that may indicate degraded performance in the road network in the future. In this paper, we apply cluster analysis to automate this process and identify similarities among links and days to find potential infrastructure deficiencies and recurring traffic states. Our study uses k-means on descriptive statistics to reveal link-types and day-types. Applying our method to high-resolution travel speed data from a road in Sweden reveals distinct characteristics based on the link and day. The results indicate that the relative difference between the measured travel speed and the free-flow travel speed is negative on links with higher free-flow travel speeds. Additionally, the variability in travel speeds is greater on links with lower free-flow travel speeds.