The paper deals with the identification of variables and models that can explain why a certain Severity Level (SL) may be expected in the event of a certain type of crash at a specific point of an urban road network. Two official crash records, a weather database, a traffic data source, and information on the characteristics of the investigated urban road segments of Turin (Italy) for the seven years from 2006 to 2012 were used. Examination of the full database of 47,592 crash events, including property damage only crashes, reveals 9,785 injury crashes occurring along road segments only. Of these, 1,621 were found to be associated with a dataset of traffic flows aggregated in 5 minutes for the 35 minutes across each crash event, and to weather data recorded by the official weather station of Turin. Two different approaches, a back-propagation neural network model and a generalized linear mixed model were used. Results show the impact of flow and other variables on the SL that may characterize a crash; differences in the significant variables and performance of the two modelling approaches are also commented on in the manuscript.

Back-propagation neural networks and generalized linear mixed models to investigate vehicular flow and weather data relationships with crash severity in urban road segments

MUSSONE, LORENZO;
2017

Abstract

The paper deals with the identification of variables and models that can explain why a certain Severity Level (SL) may be expected in the event of a certain type of crash at a specific point of an urban road network. Two official crash records, a weather database, a traffic data source, and information on the characteristics of the investigated urban road segments of Turin (Italy) for the seven years from 2006 to 2012 were used. Examination of the full database of 47,592 crash events, including property damage only crashes, reveals 9,785 injury crashes occurring along road segments only. Of these, 1,621 were found to be associated with a dataset of traffic flows aggregated in 5 minutes for the 35 minutes across each crash event, and to weather data recorded by the official weather station of Turin. Two different approaches, a back-propagation neural network model and a generalized linear mixed model were used. Results show the impact of flow and other variables on the SL that may characterize a crash; differences in the significant variables and performance of the two modelling approaches are also commented on in the manuscript.
Transport Infrastructure and Systems
978-1-138-03009-1
File in questo prodotto:
File Dimensione Formato  
TIS2017_094.pdf

accesso aperto

: Publisher’s version
Dimensione 157.74 kB
Formato Adobe PDF
157.74 kB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11311/1006683
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? ND
social impact