Ανάδειξη Hot Spot κυκλοφοριακής συμφόρησης σε δεδομένα τροχιάς οχημάτων
Traffic congestion Hot Spot analysis over vehicle trajectory data
View/
Keywords
Getis-Ord statistic ; Gaussian kernel ; Hot Spot analysis ; Traffic congestion ; Road networkAbstract
Hot spot analysis is the problem of identifying statistically significant spatial clusters. Τhe aim of this thesis is to identify spatio-temporal road segments with statistically significant amount of traffic congestion for massive trajectory data of moving objects.
Hotspots are identified using the modified statistical index Getis Ord Gi*, which is appropriately tailored in order to be implemented on a spatio-temporal graph. The spatial-temporal weight function applied to index Getis Ord Gi* is the Gaussian Kernel.
Two parallel and scalable algorithms are proposed for the identification of traffic congestion hotspots. In the first algorithm, the spatial autocorrelation index is calculated per graph edge, taking into consideration all the successor edges of the spatial unit under examination. In the second algorithm, the index is calculated per graph edge taking into consideration the successor edges within a user-defined distance. The goal is to determine the point where the algorithm’s optimal execution time intersects with the quality of the outcome. The algorithms are developed in Apache Spark and deployed using Google Cloud Platform. The performance of the algorithms is experimentally evaluated for different analysis parameters and the quality of the outcome both analytically and via visualization.