Congestion Estimation Through Multiple Congestion Metrics: A Case Study of Chicago

This paper estimates the congestion status of community areas in the city of Chicago through multiple congestion metrics with different data-driven methodologies. Using publicly accessible taxi trip data, we compute two congestion metrics: mean velocity and the congested vehicle miles traveled (VMT) ratio. A single-layered perceptron can effectively estimate the congestion index based on the mean velocity for each community area at each 15-minute interval, using the historical congestion estimates data as the ground truth. We use K-Means clustering to assign the congestion index to the entire city of Chicago in different time periods based on the congested VMT ratio. The integrated congestion index combines the two congestion indexes through agglomerative hierarchical clustering. Congestion maps are plotted to assess the congestion status of the city and to evaluate the validity of the integrated congestion index.