Identification of vehicular traffic patterns using density-based trajectory clustering algorithms

Abstract

Context: This paper focuses on identifying congestion patterns. Three data sets from the cities of Beijing, Guayaquil and Rome are used for this purpose. Method: The implementation of the Dyclee algorithm, modified to group cells of trajectories based on speeds, is carried out, with which experiments are performed in which the patterns of service volume and operability index are adequately calculated based on their results. The predominant research modality is bibliographic. However, the study also includes characteristics of a field research, because the algorithms were executed with GPS trajectory data from three different data sets. Results: To validate this research, two experiments were run. In the first experiment it was determined that the Dyclee algorithm correctly calculated the congestion patterns. In the second experiment it was found that TRADBSCAN obtained the best results with respect to validation metrics taking into account the established parameters. Conclusions: It was concluded that the Dyclee density-based clustering algorithm is able to identify vehicular congestion patterns.