@CRUiSE

Advanced Impact Integration Platform for Cooperative Road Use .

THE IMPORTANCE OF STATISTICS TO SUPPORT ROAD

SAFETY ANALYSIS
Margarida Coelho, Scientific Coordinator, DEM/TEMA
Mariana Vilaça, DEM/TEMA

Aveiro, May 8, 2019

 @CRUISE

MOTIVATION
Pedestrians

- Worldwide 1.2 million people die every

year in road crashes;
- Traffic-related injury is the leading cause
of death for 15-29 years old;
- VRUs are more likely to suffer the most Cyclists
severe consequences in collisions.
- Pedestrians and cyclists are the majority
of people killed and injured in urban
centers.

Source:
Adminaite, D., Allsop, R., & Jost, G. (2015). Making Walking and Cycling on Europe’s
Roads Safer. European Transport Safety Council. Brussels.
European Commission (2018). Pedestrians and cyclists. Directorate General for
Transport, February, 2018.
 @CRUISE

RESEARCH WORK #1

Statistical Analysis of the Occurrence and Severity

of Crashes Involving Vulnerable Road Users –
Aveiro case study

Mariana Vilaça1, Nélia Sousa2, Margarida C. Coelho1

1 DEM/TEMA, University of Aveiro
2 DMat/ CIDMA, University of Aveiro

EWGT2017 & Transportation Research Procedia

 @CRUISE

OBJECTIVES
- To develop a predictive model to estimate the likelihood of occurrence a crash
involving VRUs
- To identify risk factors that can affect pedestrians and cyclists injury severity
- To perform a spatial and temporal analysis of crashes involving VRUs (pedestrians
and cyclists), considering the level of injury severity

Source: Mariana Vilaça, Nélia Silva, Margarida C. Coelho (2017) Statistical Analysis of
the Occurrence and Severity of Crashes Involving Vulnerable Road Users: Portugal
Experience, Transportation Research Procedia 27 (2017), pp. 452–459.
 @CRUISE

METHODOLOGY
Crashes Database

246 Crashes involving 257 pedestrians and cyclists injured for

the period 2012- 2015

Database: Records from PSP and ANSR

Parameters: date, time, meteorological conditions, type of

intersection, gender and age of VRU, injury severity.

Source: Mariana Vilaça, Nélia Silva, Margarida C. Coelho (2017) Statistical

Analysis of the Occurrence and Severity of Crashes Involving Vulnerable Road Users:
Portugal Experience, Transportation Research Procedia 27 (2017), pp. 452–459.
 @CRUISE

METHODOLOGY
Pedestrians and cyclists counts, by gender
Weekdays : Wednesday and Thursday
Hours: 8-10a.m.; 12-2p.m. and 5-7p.m

Analyses of VRU exposure

Data analysis

231 pedestrians 28 cyclists per

per hour hour

Source: Google Maps

50-59% female 74 -77% male
Source: Mariana Vilaça, Nélia Silva, Margarida C. Coelho
(2017) Statistical Analysis of the Occurrence and Severity of
Crashes Involving Vulnerable Road Users: Portugal Experience,
Transportation Research Procedia 27 (2017), pp. 452–459.
 @CRUISE

Results Distribution by gender and age

71,9% female
28,1% male
ANNUAL EVOLUTION
28,8% female

71,2% male

Age Range % VRUs injured

5 to 15 years old 11%
16 to 26 years old 22%
27 to 37 years old 10%
38 to 48 years old 14%
49 to 59 years old 14%
60 to 70 years old 14%
+ 71 years old 15%

Source: Mariana Vilaça, Nélia Silva, Margarida C. Coelho (2017) Statistical

Analysis of the Occurrence and Severity of Crashes Involving Vulnerable Road Users:
Portugal Experience, Transportation Research Procedia 27 (2017), pp. 452–459.
 @CRUISE

Results
Pedestrians
8:30a.m. 4:00p.m. 7:30p.m.

Distribution of road
crashes per hour

Cyclists

6:30p.m.

5:00p.m.

8:00p.m.

Source: Mariana Vilaça, Nélia Silva, Margarida C. Coelho (2017) Statistical Analysis of the Occurrence and Severity of
Crashes Involving Vulnerable Road Users: Portugal Experience, Transportation Research Procedia 27 (2017), pp. 452–459.
 @CRUISE

Results
Pedestrians Distribution of road crashes per
type of traffic singularity

Cyclists

Source: Mariana Vilaça, Nélia Silva, Margarida C. Coelho (2017) Statistical Analysis of the Occurrence and Severity of
Crashes Involving Vulnerable Road Users: Portugal Experience, Transportation Research Procedia 27 (2017), pp. 452–459.
 @CRUISE

MAIN CONCLUSIONS

Predictive Model
- Compared with a cyclist, pedestrians are 2.7x more likely to
experience crashes in urban street segments, 10.6x more likely to
experience crashes at a crosswalk and 3.5x more likely to have a
female pedestrian involved
- The probability of pedestrian crashes that occur with good
meteorological conditions decrease 77% when compared with cyclists.

Source: Mariana Vilaça, Nélia Silva, Margarida C. Coelho (2017) Statistical Analysis of the Occurrence and Severity of
Crashes Involving Vulnerable Road Users: Portugal Experience, Transportation Research Procedia 27 (2017), pp. 452–459.
 @CRUISE

MAIN CONCLUSIONS

General
The number of crashes at intersections has increased for pedestrians and
cyclists during the study period
35% of crashes occur during the afternoon peak hour
56.2% of crashes occur at pedestrian crosswalks
73.2% of crashes occur under good meteorological conditions

Source: Mariana Vilaça, Nélia Silva, Margarida C. Coelho (2017) Statistical Analysis of the Occurrence and Severity of
Crashes Involving Vulnerable Road Users: Portugal Experience, Transportation Research Procedia 27 (2017), pp. 452–459.
 @CRUISE

RESEARCH WORK #2

Frequency and severity of crashes involving vulnerable

road users – An integrated spatial and temporal analysis
(Aveiro, Porto and Lisbon Case Studies)

Mariana Vilaça, Eloísa Macedo, Pavlos Tafidis, Margarida C. Coelho

DEM/TEMA, University of Aveiro
TRB2018
 @CRUISE

OBJECTIVES

- To perform an integrated spatial and temporal analysis of crashes involving VRUs,

considering severity of injuries;
- To develop a predictive model to estimate the likelihood of occurrence of a crash
involving VRUs.

Source: Mariana Vilaça, Eloisa Macedo, Pavlos Tafidis, Margarida C. Coelho, Frequency and severity of crashes involving
vulnerable road users – An integrated spatial and temporal analysis, 97th Transportation Research Board Annual Meeting,
Washington D.C., 2018.
 @CRUISE

METHODOLOGY Kernel Density Estimation (KDE)

Light Injuries 1

Serious Injuries 3

Fatalities 5

Temporal Analysis

Multinomial Logistic Regression

Source: Mariana Vilaça, Eloisa Macedo, Pavlos
Tafidis, Margarida C. Coelho, Frequency and exp 𝒙𝛵 𝜷
𝜋 𝒙 = ,
1 + 𝑒𝑥𝑝 𝒙𝛵 𝜷
severity of crashes involving vulnerable road users –
An integrated spatial and temporal analysis, 97th
Variables: gender; age group; injury severity;
Transportation Research Board Annual Meeting,
weekday; time period and weather conditions
Washington D.C., 2018.
 @CRUISE

RESULTS

Source: Mariana Vilaça, Eloisa Macedo, Pavlos Tafidis, Margarida C. Coelho,

Frequency and severity of crashes involving vulnerable road users – An integrated
spatial and temporal analysis, 97th Transportation Research Board Annual
Meeting, Washington D.C., 2018.
 @CRUISE

MAIN CONCLUSIONS
- Spatial analysis: most injuries occur near high attraction places, with
speed limits under 50km/h
- Most injuries occur during afternoon peak hours (4-7p.m)
- Cyclists between 25-49 yr. are the age group with more injuries
- Pedestrians between 18-49 yr. are the age group with more
occurrences in Aveiro, while the 65+ years group has more occurrences in
Porto and Lisbon
- MLR models revealed that the VRU gender and age group, and
weather conditions are statistically significant predictor variables
Source: Mariana Vilaça, Eloisa Macedo, Pavlos Tafidis, Margarida C. Coelho,
Frequency and severity of crashes involving vulnerable road users – An integrated spatial
and temporal analysis, 97th Transportation Research Board Annual Meeting,
Washington D.C., 2018.
 @CRUISE

RESEARCH WORK #3

A Rare Event Modelling Approach to Assess Injury

Severity Risk of Vulnerable Road Users

Mariana Vilaça, Eloísa Macedo, Margarida C. Coelho

DEM/TEMA, University of Aveiro

Published in Safety Journal

 @CRUISE

OBJECTIVES

- To develop an approach to reveal the most significant risk factors that can possibly
affect VRU injury severity when involved in a motor vehicle crash
- To examine 3 resampling techniques in an attempt to improve the model’s
classification performance, since prediction model performance can be biased when
imbalanced data is used
- To use 2 supervised methods: decision tree and logistic regression

Source: Mariana Vilaça, Eloísa Macedo and Margarida C. Coelho, A Rare Event
Modelling Approach to Assess Injury Severity Risk of Vulnerable Road Users, Safety,
in press.
 @CRUISE

METHODOLOGY Aveiro; Porto; Lisbon

2012-2017

Source: Mariana Vilaça, Eloísa Macedo and Margarida C. Coelho, A Rare Event
Modelling Approach to Assess Injury Severity Risk of Vulnerable Road Users, Safety,
in press.
 @CRUISE

MAIN RESULTS AND CONCLUSIONS

- Machine learning classifiers were able to correctly classify both the
majority and the minority classes with relatively high accuracy
- Oversampling technique (used to balance the dataset) always
improves the effectiveness of both classifiers (decision tree and logistic
regression) to identify risk factors
- Road markings, road conditions and luminosity significantly affect the
severity of a pedestrian’s injury
- Age group and temporal variables (month, weekday and time period)
are the risk factors that were revealed to be the most significant to
predict the severity of a cyclist’s injury
Source: Mariana Vilaça, Eloísa Macedo and Margarida C. Coelho, A Rare Event
Modelling Approach to Assess Injury Severity Risk of Vulnerable Road Users, Safety,
in press.
 @CRUISE

CONCLUDING REMARKS
• Access to real data is crucial.
• Statistical analysis lead to a deep
knowledge on transportation
related phenomena.
• Identification of risk factors is
relevant to help decision-makers in
the definition of road safety
policies and strategies that aim to
reduce the injury severity of crashes
between VRUs and motor vehicles
 @CRUISE

CONTACTS http://transportes-tema.web.ua.pt

[email protected]

Acknowledgements

HTTPS://PROJECT-CRUISE.WEEBLY.COM/

AVEIRO, MAY 8, 2019