A spatiotemporal analysis of traffic congestion patterns using clustering algorithms : a case study of Casablanca

Analyzing the flow of vehicles within a metropolitan area is essential to understand traffic congestion better and evaluate urban mobility for more environmentally and sustainable targets. This study uses real-time big data to conduct a spatiotemporal analysis of traffic congestion patterns in Casablanca, Morocco. This is the first attempt to explore the spatiotemporal pattern of traffic congestion performance in one of North Africa's largest cities. The data collected includes hourly real-time traffic performance index data for one week, obtained using the Waze Application Programming Interface (API) to collect real-time duration and distance during a trip of 440 trajectories. Using this data, we computed the Travel Time Index (TTI) to study traffic congestion on a macro level (zones) without limiting our research to clustering the zones by applying the K-means algorithm to classify each zone of Casablanca from low, medium, and congested, to high congested communes. We also investigated the impact of different land-use patterns, such as density, parks, residential, industrial, highway, parking, commercial, and tram. We trained a generative model (Naïve Bayes classifier) to study explanatory variables that contribute to increasing levels of traffic congestion. As a result, we found that these land use factors significantly influence the worsening of traffic congestion. Residential areas, industrial areas, park areas, parking, commercial buildings, and tramways were identified as the primary contributors to traffic congestion.