Ανάλυση hot-spot για οδικά δεδομένα μέσω παράλληλης επεξεργασίας
Hot-spot analysis on road traffic data using parallel system approach
View/
Keywords
Hot-spot ; Συμφόρηση ; Κατανεμημένο σύστημα ; Επεξεργασία γράφουςAbstract
Traffic analysis systems are an important area of study in the modern era. Due to the
continuous increase in number of vehicles and the increasing complexity of the structure
of road networks in densely populated areas, it is common for areas with high levels
of congestion to appear at times on given parts of the road network. The analysis and
identification of these points provides important information about the structure and weaknesses
of the road network. It is therefore necessary to find methods that can perform statistical
analysis spatially and temporally for the appearance of hot-spot areas of interest. At the
same time, these methods must be able to manage the large volume of data resulting
from vehicle trajectories and extract useful information even from sparse samples from
the trajectories.
The objective of this work is to create a model that highlights hot-spot road sections with
statistically significant values of traffic congestion in different analysis time windows. The
proposed model is applied to sparsely sampled vehicle trajectory data and implemented in
a distributed systems environment.
The Getis-Ord index G∗i is used to analyze hotspots, which performs calculations on
spatiotemporal data. Additionally, a Gaussian kernel is used as a distance weight function
for the relationship between data proximity.
The proposed methodology is an extension of previous research. The proposed algorithm
aims to address problems that arise in previous models. It pursues this goal by improving
the parallelization of calculations through the grouping of spatiotemporally close data,
thereby reducing the need to search for neighboring edges in different execution systems.
The analyses and the proposed algorithm were implemented in the Apache Spark environment.
The results obtained are executed on a distributed Okeanos system.