2014 IEEE 17th International Conference on

Intelligent Transportation Systems (ITSC)

October 8-11, 2014. Qingdao, China

Driving Risk Assessment using Cluster Analysis based on

Naturalistic Driving Data*
Yang Zheng, Jianqiang Wang, Xiaofei Li, Chenfei Yu, Kenji Kodaka and Keqiang Li

example, Al-Ghamid et al. showed that the location and cause



Abstract—In addition to the real traffic accident data,

naturalistic driving data can allow researchers gain of accidents were most significantly associated with accident
insights into the factors that cause risk/hazard situations. severity using logistic regression based on accident-related
This paper considers a comprehensive naturalistic data [7]. Chang et al. proposed a classification and regression
driving experiment to collect detailed driving data on tree model to establish the relationship among injury severity,
actual Chinese roads. Using acquired real-world driving driver/vehicle characteristics, environment factors and
accident severity using office recorded vehicle accident data
data, a near-crash database is built, which contains
[8]. These studies have typically been based on official traffic
vehicle status, potential crash object, driving
accident statistics which have two major limitations: 1) lack
environment and road type, and weather condition. of detailed driving data; 2) difficult to collect and acquire
K-means cluster analysis is applied to classify the (usually collected by traffic police agency). Hence, the
near-crash cases into different driving risk levels using studies stated above do not consider the relationship between
braking process features, namely maximum deceleration, the detailed driving data (e.g. vehicle speed, acceleration,
average deceleration and percentage reduction in the braking and steering information) and accident severity.
vehicle kinetic energy. The results indicate that the Recent developments in vehicle instrumentation techniques
velocity when braking and triggering factors have strong have made monitoring the naturalistic driving behavior and
relationship with the driving risk level involved in obtaining detailed driving data both technologically possible
near-crash cases. and economically feasible. For example, NHTSA sponsored
the project ‘100-Car Naturalistic Driving Study’, which is the
I. INTRODUCTION first large-scale instrumented vehicle study undertaken to
collect naturalistic driving data in the United States [9]. With
Over the last two decades, significant progress has been
access to naturalistic driving data, many researchers have
made in all aspects of vehicle safety system [1]. Efforts that
proposed new methods and gained new insights in traffic
aim to advance a safer vehicle traffic system can mainly be
safety involving drivers, vehicles and roadways [10]-[13].
divided into two areas: 1) active safety [2][3], and 2) passive
Malta et al. focused the pedal signals and driver speech to
safety [4]. Although many encouraging achievements have
better understand the driver behavior under potential threats
been made, the number of road fatalities remains
using a large real-world driving database [10]. Aoude et al.
unacceptably high, and traffic accidents are considered as a
developed SVMs and a hidden Markov model for driver
major public health problem [5]. Because responsibility for
behavior classification at intersections and validated the
traffic accidents involves the vehicle, driver and road, we
proposed algorithms using naturalistic intersection data [11].
must not only improve the safety performance of vehicles but
also better understand the factors that affect driving risk and This paper focuses on the analysis of the factors that affect
identify the factors that result in accidents to make road the driving risk using naturalistic driving data. In this study,
transportation much safer. we first conducted a comprehensive naturalistic driving
experiment to collect detailed driving data on actual Chinese
Many research activities have been conducted to seek
roads and then built a near-crash database through designing a
better understanding of the factors that affect the probability
novel data transcription protocol. The driving risk level under
and injury severity of crashes in the hope of providing police
near-crash cases is represented by the braking process
countermeasures to reduce the number of crashes [6]. For
characteristics. The K-means cluster method is adopted to
classify the near-crash cases into different risk level groups
*Research supported by National Natural Science Foundation of China based on these three braking process features. The results
(Grant No. 51175290) and the joint project of Tsinghua and Honda. indicate that the velocity when braking and triggering factors
Yang Zheng, Xiaofei Li and Keqiang Li are with the State Key Laboratory have the largest influence on the driving risk level.
of Automotive Safety and Energy, Tsinghua University, Beijing
100084,China.(e-mail:zhengy13@mails.tsinghua.edu.cn,lixf11@mails.tsing
hua.edu.cn, likq@tsinghua.edu.cn.) II. NATURALISTIC DRIVING DATA AND DATABASE DESIGN
Jianqiang Wang is with the State Key Laboratory of Automotive Safety To build a firm research foundation for driving risk
and Energy, Tsinghua University, Beijing 100084, China. (Corresponding assessment and enhanced driving safety, two components are
author; phone: ( +86-10-62795774; e-mail: wjqlws@tsinghua.edu.cn)
Chenfei Yu was with the State Key Laboratory of Automotive Safety and essential: 1) actual driving data and 2) careful experimental
Energy, Tsinghua University, Beijing 100084, China. She is currently with design. In contrast to field operational tests, data collection is
Pan Asia Technical Automotive Center Co., Ltd., Shanghai 201201, China. performed through naturalistic and low intervention method
(e-mail: yuchenfei@tsinghua.org.cn) in actual traffic condition. This section introduces the
Kenji Kodaka is with Honda R&D Co., Ltd. Automobile R&D Center, experimental equipment and experiment design, describes
Tochigi 321-3393, Japan. (e-mail: Kenji_Kodaka@n.t.rd.honda.co.jp.)

978-1-4799-6078-1/14/$31.00 ©2014 IEEE 2584

Authorized licensed use limited to: Indian Institute of Technology Hyderabad. Downloaded on June 10,2024 at 07:01:08 UTC from IEEE Xplore. Restrictions apply.
data transcription protocol and then builds the near-crash Among the 31 drivers, 9 were female and 22 were male
database. with regular driving license. They were 43 years old, on
average (age range from 25 to 67 years) and have possessed a
A. Data-collection Equipment driving license for a mean period of 16 years (ranging from 3
The naturalistic driving experiments were conducted on a to 48 years).
Honda Crosstour. The vehicle was provided with instruments
9
to collect driver, vehicular and road data under real-world braking signal
conditions. The data-collection system installed in the Longitudinal Acceleration
6
experimental vehicle included GPS, vehicle sensors, two Lateral Acceleration

Acceleration(m/s 2)
driving recorders (DR) and four CCD cameras (Fig. 1). The
3
four cameras recorded detailed video scenes including 1)
Forward view, 2) Right-side forward view, 3) Left-side
0
forward view, 4) Driver’s facial expression. One DR recorded
the vehicle speed obtained by GPS, brake signal, steering
signal, three-axis acceleration information and detailed video -3

collected by the facial-expression and forward view cameras.

The other DR recorded the video collected by both left-side -6
-10 -5 0 5 10
and right-side forward view cameras for the convenience to Time(s)
code the incidents. Figure 2. Example of recorded driving signals
In our study, we focused on the risk factors and driver
behaviour under near-crash scenario in the naturalistic driving TABLE I. SCHEDULE OF THE ENTIRE EXPERIMENT
experiment. Near-crash case means the driver must perform Time period Morning Afternoon Night
emergency braking operation; otherwise a real crash will
occur. For the experimental data collection, a near-crash case Hours 140 220 50
means that the deceleration of the experimental vehicle
reaches a threshold value instead of happening actual TABLE II. ROAD TYPES IN THE EXPERIMENT
accidents. Hence, the data-collection system recorded the
vehicle state (speed, brake signal, steering signal and Road type 1 2 3 4
three-axis acceleration) and four video scenes when a large Kilometres 1800 1210 4100 1650
deceleration was detected. The recording time started 1: Highway, 2: City ring road, 3: Inner-city road, 4: Rural road
approximately from 10 s before the triggering point to 5 s after
the triggering point, which means that a typical near-crash C. Hand Labelling of Near-crash Database
case has approximately a 15-s signal and video sequence. Fig. Altogether, we obtained 912 near-crash cases throughout
2 shows examples of the recorded driving signals. the 60-day naturalistic driving experiment with the 31 drivers.
Deciding the protocol of labelling the multi-modal
information is critical in properly associating the near-crash
driving situation with the recorded driving state signal and
video. A novel data transcription protocol that considers a
comprehensive cross section of the factors that could affect the
drivers and their responses is proposed in this study. The
proposed protocol comprises the following five major
categories:
1) Vehicle status
2) Potential crash object
3) Driving environment and road type
Figure 1. Experimental vehicle and equipment 4) Weather condition
B. Experiment Design 5) Driver information and driver actions
The naturalistic driving route contained all road types: The designed transcription protocol is comprehensive and
inner-city highway, city ring road, inter-city road (mixed contains important attributes that describe the potential
traffic conditions) and rural road (poor road structure and factors contributing to the driving risk, providing potential for
crowded living quarters). A total of 31 drivers, who have analysing the relationship among the driving risk,
signed the informed consent form, participated in these driver/vehicle characteristics and road environment. Graduate
naturalistic driving experiments at their normal driving state. students with driving license served as volunteer taggers, who
The experiment lasted for 60 days at 6–7 h/day, which resulted manually labelled the recorded 912 near-crash cases
in an approximately 400-h naturalistic driving time and over according to the designed transcription protocol. Finally, we
8500-km naturalistic driving range. The schedule of the entire developed a near-crash database. TABLE III lists the
experiment plan is listed in TABLE I, and TABLE II lists the definition of the transcription protocol.
naturalistic driving distance on the different road types.

2585
Authorized licensed use limited to: Indian Institute of Technology Hyderabad. Downloaded on June 10,2024 at 07:01:08 UTC from IEEE Xplore. Restrictions apply.
 TABLE III. DEFINITION OF TRANSCRIPTION PROTOCOL

Variable Code Type Description

Vehicle Status
Velocity when braking V_BRA Continuous The vehicle speed when the driver triggers the braking signal or the turn
point of acceleration signal (m/s)
Maximum deceleration D_MAX Continuous The maximum deceleration during the emergency braking process (m/s 2)
Time interval of braking T_IN Continuous The time interval between the braking signal trigging and the time point of
maximum deceleration
Velocity Reduction V_RED Continuous The vehicle speed reduction form the braking signal trigging to the time
point of maximum deceleration
Vehicle status before braking process V_STA Qualitative 1, Deceleration process; 2, Acceleration process; 3, Constant speed;
Vehicle maneuver V_MAN Qualitative 1, Straight going; 2, Right turn; 3, Left turn; 4, Lane change; 5, Others
Potential crash object
Crash Object Type O_TYP Qualitative 1, Vehicle; 2, Single-track vehicle (motorcycle and bicycle); 3, Pedestrian;
4, Others
Potential crash type P_CRA Qualitative 1, Rear end; 2, Conflict during intersection; 3, Jump out; 4, Opposite
driving conflict; 5, Cut-in conflict; 6, Others
Triggering factors T_FAC Qualitative 0, Non-host vehicle factors; 1, Traffic light; 2, Lane reduction; 3, Lane
change; 4, Collision avoidance; 5, Others
Driving environment and road type
Near crash location N_LOC Qualitative 1, Intersection; 2, Non-intersection
Road type R_TYP Qualitative 1, Structure road; 2, Normal road; 3, Hybrid road; 4: Rural road
Parking vehicle along the road side P_PLA Qualitative 0, No; 1,Yes
Barriers for the opposing traffic flow B_TRA Qualitative 0, No; 1,Yes
Barriers for vehicles and pedestrian B_VEH Qualitative 0, No; 1, Yes
Weather Condition
Weather WEA Qualitative 1, Sunny;2:Cloudy; 3: Others
Light condition L_CON Qualitative 1, Lightness; 2, Little dim;
Driver information and actions
Gender GEN Qualitative 1, Male; 2, Female
Age AGE Continuous The driver’s age (years)
Time span with driving license T_DIR Continuous The time period that owning the valid driving license
Steering light S_LIG Qualitative 0, No; 1, Yes
Vehicle horns V_HON Qualitative 0, No; 1, Yes
Second Task S_TASK Qualitative 0, No; 1, Talking; 3, Others

(1) Maximum deceleration during the braking process .

III. DEFINITION AND CLUSTER OF DRIVING RISK (2) Average deceleration from the braking triggering
A. Definition of Driving Risk point to the point of maximum deceleration
(3) Percentage reduction in the vehicle kinetic energy
In this paper, driving risk is defined as a potential threat from the braking triggering point to the point of
that may cause vehicle crashes or other accidents. Usually, the
maximum deceleration
consequence of driving risk that involves the driver is mainly
reflected by the emergency braking operation. Hence, the The average deceleration can be calculated by
driving risk level can be represented by the braking process the following formula:
characteristics. Fig. 3 shows the key point used to define a
typical deceleration curve of the braking process. The ∫ ( ) [ ( ) ( )]. (1)
following three features are adopted to represent the driving
risk level involved in a typical near-crash case during where ( ) ( ) denotes the vehicle velocity and
naturalistic driving: acceleration respectively. The percentage reduction in the
vehicle kinetic energy can be calculated as following:

2586
Authorized licensed use limited to: Indian Institute of Technology Hyderabad. Downloaded on June 10,2024 at 07:01:08 UTC from IEEE Xplore. Restrictions apply.
 deceleration of the moderate-risk group is also much higher
( ) ( ) ( )
[ ] (2) than that of the low-risk group.
( ) ( )

where denote the vehicle mass. 1

6
braking signal 0.8
Longitudinal Acceleration
0.6
3 High risk

E
Acceleration(m/s 2)

η
0.4 Moderate risk
Low risk
0 0.2

0
8
-3
6 8
t0 amin
4 6
t1 4
2 2
-6
-10 -5 0 5 10 2
- aaverage (m/s ) 0 0
- amin (m/s 2)
Time(s)

Figure 3. Key features of driving-risk level Figure 4. K-means cluster results

B. Cluster for Driving Risk

IV. CLUSTER RESULT ANALYSIS
The main criterion in evaluating the driving risk level is
the braking process feature, defined as A. Data Distribution on Driving risk Level
According to the proposed data transcription protocol
[ ] . (3) (shown in TABLE III. ), the distribution of near-crash cases
Cluster analysis is a valid and objective approach to on the driving risk levels in terms of 19 potential risk factors
classify driving risks in different near crashes into different is shown in TABLE IV.
risk levels and has been used in individual driver risk The frequency information listed in TABLE IV indicates
assessment research [12]. The K-means cluster method, which that traffic light in the fifth potential risk variable T_FAC is
is popular for cluster analysis in data mining, is employed to an important factor on the driving risk level because a
classify the driving risks involved in different near-crash cases relatively high proportion of near-crash cases caused by
into different risk groups based on the feature . Using a sudden changes of the traffic light occurs in the moderate-
pre-determined number of clusters, the K-means cluster and high-risk groups (55.3% and 35.0%, respectively).
method partitions the observations into clusters, where each Meanwhile, the proportion of near-crash cases caused by
observation belongs to a cluster whose mean is closest to its other triggering factors conforms to the overall distribution of
value [14]. The K-means method minimises the within-cluster the driving-risk levels. From the sixth potential risk variable
sum of squares: N_LOC, we can find similar statistical results where
𝑘 near-crash cases that occur at the intersection are relatively
arg ∑ ∑‖ 𝑗 𝜇‖ (4) higher in moderate- and high-risk groups (44.5% and 12.6%,
𝑆 respectively) than those outside the intersection area (38.0%
= 𝑋𝑗 ∈𝑆𝑖
and 5.2%, respectively). The other meaningful findings listed
where [ ] is the set of observed data, which in TABLE IV show that the higher the braking speed is, the
represents the braking process feature higher is the proportion of near-cash cases in the moderate-
[ ] in the context of this paper; and high- risk groups. The proportions in the moderate- and
high-risk groups are, respectively, 46.4% and 13.9% when
[ ] represents the set of clusters and 𝜇
the speed at the braking point lies in the range from 10 to 20
denotes the mean point of cluster set . m/s, whereas those when the speed at the braking point lies
The driving risk level under each near-crash case is from 0 to 10 m/s are, respectively, 34.7% and 2.8%, as shown
classified into one of the three clusters: 1) low-risk group, 2) in the 19th potential risk variable V_BRA.
moderate-risk group, 3) high-risk group. Near crashes in the B. Risk factors affecting Driving Risk
clusters with the highest maximum deceleration are
considered to be high driving-risk group. The output of the In this section, the variable importance obtained form
cluster analysis is shown in Fig. 4. TABLE V summarises the decision tree is used to quantify the influence of potential risk
statistical characteristics of these three driving-risk groups. factors on driving risk level. This analysis is performed using
The number distribution of the different risk groups follows a SPSS software. For detail description of decision tree and
pyramid structure, which means that the high-risk group has variable importance ranking, please refer to [15].
minimum near-crash cases, whereas the low-risk group has the TABLE VI lists the normalized importance of the
largest number of near-crash cases. We can see that the potential risk factors. We can easily see that the two variables,
maximum deceleration of the high-risk group is more than two namely, velocity when braking (V_BRA), triggering factor
times that of the low-risk group, and the maximum (T_FAC), have the largest influence on the driving-risk level,
which are conformed to aforementioned analysis.

2587
Authorized licensed use limited to: Indian Institute of Technology Hyderabad. Downloaded on June 10,2024 at 07:01:08 UTC from IEEE Xplore. Restrictions apply.
 TABLE IV. DISTRIBUTION OF DRIVING-RISK LEVELS BY POTENTIAL RISK VARIABLES

Driving risk level Driving risk level

Variable Variable Descript
Num Description Count LR MR HR Num Count LR MR HR
Code Code ion
52.0% 40.2% 7.8% 52.0% 40.2% 7.8%
1 V_STA Deceleration 9 B_TRA
265 48.7% 43.8% 7.5% No 372 58.9% 37.4% 3.8%
process
Acceleration
531 55.4% 37.5% 7.2% Yes 540 47.2% 42.2% 10.6%
process
Constant speed 116 44.0% 44.8% 11.2% 10 B_VEH No 537 55.9% 37.8% 6.3%
2 V_MAN Straight going 778 51.0% 40.4% 8.6% Yes 375 46.4% 43.7% 9.9%

Right turn 38 65.8% 31.6% 2.6% 11 WEA Sunny 727 51.2% 41.3% 7.6%
Left turn 41 65.9% 34.1% 0.0% Cloudy 147 55.8% 34.7% 9.5%
lane change 46 45.7% 50.0% 4.3% Others 38 52.6% 42.1% 5.3%
Other 9 44.4% 44.4% 11.1% 12 L_CON Lighted 796 52.3% 40.2% 7.5%
3 O_TYP Slightl
Vehicle 596 55.0% 40.4% 4.5% 116 50.0% 40.5% 9.5%
y dim
Single-track 13 GEN
98 72.4% 21.4% 6.1% Male 661 51.0% 40.5% 8.5%
vehicle
Pedestrian 69 60.9% 37.7% 1.4% Female 251 54.6% 39.4% 6.0%
Others 149 22.1% 53.0% 24.8% 14 AGE  30 145 50.3% 41.4% 8.3%

4 P_CRA Rear end 349 51.3% 45.0% 3.7% 31-40 291 54.0% 39.9% 6.2%
Conflict during
70 61.4% 32.9% 5.7% 41-50 232 48.7% 40.9% 10.3%
intersection
Jump out 65 60.0% 36.9% 3.1% 51-60 202 56.9% 35.6% 7.4%
Opposite
46 67.4% 28.3% 4.3% 60  42 38.1% 57.1% 4.8%
driving conflict
Cut-in conflict 191 63.4% 30.4% 6.3% 15 T_DIR  10 305 50.8% 40.0% 9.2%
Others 191 63.4% 30.4% 6.3% 11-20 380 56.1% 37.1% 6.8%
5 T_FAC Non-host
723 57.7% 37.6% 4.7% 21-30 157 46.5% 44.6% 8.9%
vehicle factors
Traffic light 103 9.7% 55.3% 35.0% 30  70 47.1% 48.6% 4.3%

Lane reduction 9 77.8% 22.2% 0.0% 16 S_LIG No 784 51.3% 40.3% 8.4%
Lane change 33 48.5% 48.5% 3.0% Yes 128 56.3% 39.8% 3.9%
Collision 17 V_HON
26 57.7% 42.3% 0.0% No 859 51.1% 41.0% 7.9%
avoidance
Others 18 57.7% 42.3% 0.0% Yes 53 66.0% 28.3% 5.7%
6 N_LOC Intersection 317 42.9% 44.5% 12.6% 18 S_TASK No 784 52.0% 40.4% 7.5%

Non-intersectio
595 56.8% 38.0% 5.2% Talking 125 51.2% 39.2% 9.6%
n
7 R_TYP Structured road 285 46.0% 43.5% 10.5% others 3 66.7% 33.3% 0.0%
Normal road 238 46.2% 43.3% 10.5% 19 V_BRA  0,10 501 62.5% 34.7% 2.8%

Hybrid road 251 62.5% 31.9% 5.6% 10,20 388 39.7% 46.4% 13.9%

Rural road 138 55.1% 43.5% 1.4%  20,+  23 30.4% 56.5% 13.0%

8 P_VEH No 586 48.1% 42.8% 9.0%

Yes 326 58.9% 35.6% 5.5%

Note: Num denotes the index of potential risk variables, and LR: low-risk group; MR: moderate-risk group; HR: high-risk group

2588
Authorized licensed use limited to: Indian Institute of Technology Hyderabad. Downloaded on June 10,2024 at 07:01:08 UTC from IEEE Xplore. Restrictions apply.
 TABLE V. CHARACTERISTIC OF DRIVING RISK GROUPS

Mean of features of the braking process

Risk groups Number of near crash cases Percentage
(m/s2) (m/s2)
Low-risk group 474 52.0 % -1.931 -1.027 30.9 %
Moderate-risk group 367 40.2 % -3.278 -1.717 56.6 %
High-risk group 71 7.8 % -5.385 -3.125 66.1 %

results indicate that the velocity when braking and triggering

TABLE VI. IMPORTANCE OF THE POTENTIAL FACTORS factors have the largest influence on the driving risk level,
which, to some extent, are in accordance with some previous
Variables Normalised importance studies.
V_BRA 100.0%
REFERENCES
T_FAC 96.7% [1] D. Caveney, “Cooperative vehicular safety applications,” Control
O_TYP 82.9% Systems, IEEE, 2010, pp. 38-53.
[2] M. Nagai, “The perspectives of research for enhancing active safety
P_CRA 75.9% based on advanced control technology,” Vehicle System Dynamics,
2007, pp. 413-431.
AGE 11.7% [3] D. Zhang, K. Li, J. Wang, “A curving ACC system with coordination
control of longitudinal car-following and lateral stability,” Vehicle
1) Velocity when Braking System Dynamics, 2012, pp. 1085-1102.
As shown in TABLE VI, velocity when braking (V_BRA) [4] A. Jarašūniene, G. Jakubauskas, “Improvement of road safety using
is the most important potential risk variable. Intuitively, the passive and active intelligent vehicle safety systems,” Transport, 2007,
higher the vehicle speed is, the greater is the kinetic energy of pp. 284-289.
[5] DTM-China (Ministry of Public Security, Department of Traffic
the lone-driver-vehicle system. If a potential threat or sudden Management). Annual report of road traffic accidents statistics in P.R.
change of the object status occurs in the driving environment, China. Scientific Research Institute of Traffic Management, Ministry
the lone-driver-vehicle system will become more unstable of Public Security, Beijing (in Chinese). 2010.
and risky, meaning that the driving risk level involved in a [6] D. Lord, F. Mannering, “The statistical analysis of crash-frequency
near-crash case will be high as the vehicle velocity increases. data: a review and assessment of methodological alternatives,”
Transportation Research Part A: Policy and Practice, 2010, pp.
Some other research studies have indicated that driving speed 291-305.
is an important factor in road safety [16]. Elvik et.al pointed [7] A. S. Al-Ghamdi, “Using logistic regression to estimate the influence
out that speed not only affects the severity of a crash but also of accident factors on accident severity,” Accident Analysis &
is related to the risk of being involved in a crash [17]. Prevention, 2002, pp. 729-741.
[8] L.Y. Chang, H.W. Wang, “Analysis of traffic injury severity: An
2) Triggering Factors application of non-parametric classification tree techniques,” Accident
Triggering factor (T_FAC) is the second important Analysis & Prevention, 2006, pp. 1019-1027.
potential variable with a 96.7% normalized importance. [9] S.G. Klauer, T.A. Dingus, V.L. Neale, et al. “The impact of driver
inattention on near-crash/crash risk: An analysis using the 100-car
TABLE VI shows that traffic light in the fifth potential risk naturalistic driving study data,” 2006.
factor variable T_FAC has significant effect on the driving [10] L. Malta, C. Miyajima, K. Takeda, “A study of driver behavior under
risk level involved in near-crash cases because a relatively potential threats in vehicle traffic,” Intelligent Transportation Systems,
high proportion of near-crash cases caused by sudden IEEE Transactions on, 2009, pp. 201-210.
changes in the traffic light occurs in moderate- or high-risk [11] G.S. Aoude, V.R. Desaraju, L.H. Stephens, “Driver behavior
classification at intersections and validation on large naturalistic data
groups (55.3% and 35.0%, respectively). This result agrees set,” Intelligent Transportation Systems, IEEE Transactions on, 2012,
with other previous studies in vehicle accidents that result pp. 724-736.
from dilemma zone at signalized intersection [11][18]. [12] F. Guo, Y. Fang, “Individual driver risk assessment using naturalistic
driving data,” Accident Analysis & Prevention, 2013, pp. 3-9.
V. CONCLUSIONS [13] F. Guo, S.G. Klauer, J. M. Hankey, “Near crashes as crash surrogate for
naturalistic driving studies,” Transportation Research Record: Journal
In this study, we obtained 912 near-crash cases through a of the Transportation Research Board, 2010, pp. 66-74.
60-day naturalistic driving experiment employing 31 drivers. [14] L. Kaufman, P. J. Rousseeuw, “Finding groups in data: an introduction
A comprehensive transcription protocol, which contains to cluster analysis,” John Wiley & Sons, 2009.
[15] L. Breiman, J. Friedman, C.J. Stone, et al. “Classification and
important attributes describing the conditions contributing to regression trees,” CRC press, 1984.
driving risk, was designed to provide the possibility of [16] L. Aarts, I. Van Schagen, “Driving speed and the risk of road crashes: A
analyzing the relationship between driving risk, review,” Accident Analysis & Prevention, 2006, pp. 215-224.
driver/vehicle characteristics and road environment. In this [17] R. Elvik, P. Christensen, A. Amundsen, “Speed and road accidents. An
paper, the driving risk level under near-crash cases has been evaluation of the Power Model,” TØI report, 2004.
[18] H.Rakha, I. El-Shawarby, J.R. Setti, “Characterizing driver behavior on
represented by the braking process characteristics, namely, 1) signalized intersection approaches at the onset of a yellow-phase
maximum deceleration, 2) average deceleration and 3) trigger,” Intelligent Transportation Systems, IEEE Transactions on, pp.
percentage reduction in the vehicle kinetic energy. K-means 630-640.
cluster is used to cluster different driving risk levels involved
in a near-crash case using the braking process features. The

2589
Authorized licensed use limited to: Indian Institute of Technology Hyderabad. Downloaded on June 10,2024 at 07:01:08 UTC from IEEE Xplore. Restrictions apply.