Distilling actionable insights through road network features to alleviate traffic congestion

Policy transfer is an efficient strategy for urban policymaking. However, the unique city-specific characteristics complicate the transfer of congestion-mitigating policies. We study seven highly congested cities worldwide and argue that, since regions with less congestion exist even within these cities, understanding how road network features explain the intra-city variations in con- gestion can significantly enhance the efficacy of mitigation measures. While our findings reveal extreme city-specific variations in feature importance, suggesting that supply-side interventions must be tailored to individual cities, the most effective features for demand-side interventions share notable similarities across different cities. In the short term, our results directly impact the ongoing mitigation measures in these cities, such as congestion pricing and personalised route choice applications, specifically in making them more robust and increasing public acceptance. In the long term, our insights encourage a re-evaluation of scepticism towards infrastructure invest- ments in mitigating congestion and highlight the potential for designing congestion-resistant future cities. Moreover, given the diverse selection of cities, the few common observations suggest that extrapolating those common observations to new cities will likely be effective, especially when a detailed analysis like ours is impossible due to data constraints. Finally, we provide key recommendations for researchers from the data-driven traffic prediction community regarding the need to focus on microscopic scales and for researchers in simulation-based counterfactual approaches to simulate the impact of road network features with values beyond the known ranges for feature values in the studied cities.