Traffic state classification in grid area based on floating car data

Traffic congestion is a common phenomenon in many cities, leading to safety issues and exacerbating environmental pollution. The scientific classification of traffic conditions is fundamental to understanding the mechanisms of congestion evolution and is a prerequisite for solving congestion problems. This paper proposed a traffic condition classification method based on an urban road grid model. The grid model was constructed by matching floating vehicle trajectories with a grid map, from which seven feature parameters were extracted: average speed, average acceleration, speed standard deviation, number of stops, 95th percentile speed, speed ratio, and stop time. Principal component analysis (PCA) was then used to retain the three most valuable principal components. Bayesian optimization (BO) was employed to assist the K-Means algorithm in finding the optimal initial cluster centers. This results in an improved BO-KMeans clustering algorithm that classifies traffic conditions into four categories: unclogged, sluggish, moderate congestion, and heavy congestion. The scientific classification of traffic states is fundamental to research on the causes of urban road congestion and the mechanisms of its evolution. It plays a crucial role in addressing traffic congestion issues and advancing the development of smart cities.