Association Rules to Identify Factors Affecting Risk and Severity of

Road Accidents

Irina Makarova

, Gulnara Yakupova

, Polina Buyvol

, Eduard Mukhametdinov

1 d

and Anton Pashkevich

Kazan Federal University, Syuyumbike prosp., 10a, 423822 Naberezhnye Chelny, Russian Federation

Politechnika Krakowska, Warszawska st., 24, Krakow, Poland

Keywords: Accidents, Association Rules, Factors, Safety, Severity of Consequences, Vehicles.

Abstract: The current increase in automobilization leads to a decrease in road safety. Therefore, the purpose of this

research is to analyze and identify the causes that significantly affect the risk and severity of accidents. For

this purpose, both histograms plotting and association rules were used. The statistics of traffic accidents in

Elabuga town for 2017-2018 were taken as initial information. To identify the most traumatic types of traffic

accidents, graphs of the number of accidents by types of accidents and the number of victims for 2017-2018

were constructed. It was found that the greatest number of injured (wounded or dead) is observed in collisions

and hitting a pedestrian. Then, road sections of the concentration of accidents were analyzed, the most

common types of accidents and the main violations of traffic rules that contribute to the occurrence of

accidents were determined. To identify hidden relationships between factors and accidents with severe

consequences, association rules were applied. As a result, the influence of weather conditions, quality of road

infrastructure and marking was established.

1 INTRODUCTION

The annual growth of the global fleet of vehicles with

new units using the existing limited road

infrastructure leads to an increase in traffic intensity

and the number of congestion on the roads, which, in

turn, causes many problems, including those related

to a decrease in traffic safety and an increase in the

negative impact on environment.

In this regard, in 2010, the European Commission

launched the European Smart Cities Initiative, which

includes four urban areas: buildings, heating and

cooling systems, electricity and transportation

(European Initiative on Smart Cities, 2010-2020).

The goal associated with transport is to create

intelligent public transport systems based on real-

time information, traffic management systems (TMS)

to prevent traffic collapses, as well as reduce the

environmental negative effects of the transport

https://orcid.org/0000-0002-6184-9900

https://orcid.org/0000-0001-6822-3700

https://orcid.org/0000-0002-5241-215X

https://orcid.org/0000-0003-0824-0001

https://orcid.org/0000-0002-4066-5440

system on the environment. However, as noted in

(Djahel, S., Doolan, R., Muntean, G.-M., Murph, J.,

2015), existing TMS do not provide sufficient and

accurate traffic information to provide detailed and

timely traffic monitoring and management.

On the other hand, over the past decade,

commercial companies and government

organizations have been actively developing research

areas related to autonomous vehicles, which, it seems,

will soon become widespread on the roads after legal

issues regarding the functioning of these new road

users are resolved. It seems that one of the main

problems that inevitably arise when changing the

composition of road users is their safe and efficient

operation in difficult road conditions, such as road

junctions. At road network locations where traffic

conflicts may occur, such as at intersections, it is

necessary to ensure that autonomous vehicles operate

safely and efficiently, and, more importantly, that

614

Makarova, I., Yakupova, G., Buyvol, P., Mukhametdinov, E. and Pashkevich, A.

Association Rules to Identify Factors Affecting Risk and Severity of Road Accidents.

DOI: 10.5220/0009836506140621

In Proceedings of the 6th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2020), pages 614-621

ISBN: 978-989-758-419-0

conventional people-driven vehicles at least maintain

their current level of safety.

In our opinion, a base of the most representative

"reference" scenarios of the interaction of

autonomous vehicles with traditional ones, especially

with the most vulnerable, should be created for this.

It is necessary to determine a list of driving situations

that should be evaluated in terms of possible conflict

avoidance, and then traffic control options for these

situations will be modeled and defined. Since there is

no information on the interaction of autonomous

vehicles with other participants in the movement, but

taken into account that their movement will be carried

out according to the given algorithms, at the first

stage it is necessary to analyze the existing statistics

of road traffic accidents, select their concentration,

and then use the simulation models to determine the

most dangerous scenarios.

2 OVERVIEW OF EXISTING

METHODS

Recently, the number of researches devoted to the

study of the significance of factors affecting the

severity of accidents has increased significantly (Zou,

X., Yue, W. L., Vu, H.L., 2018). An active field of

research by scientists from different countries is the

study of the complex relationship between

influencing factors and the severity of accidents using

statistical methods and machine learning algorithms:

classification and regression trees (Moral-García, S,

Castellano, J. G., Mantas, C. J., Montella, A.,

Abellán, J., 2019), neural networks (Theofilatos, A.,

Chen, C., Constantinos, A., 2019; Zheng, M., Li, T.,

Zhu, R., Chen, J., Ma, Z., Tang M., et al., 2019),

support vector methods (Chen, C., Zhang, G., Qian,

Z., Tarefder, R.A., Tian, Z., 2016), naive Bayes

classifier (Chen, C., Zhang, G., Yang, J., Milton,

J.C., Alcántara, A.D., 2016; Li, Z., Wu, Q., Ci, Y.,

Chen, C., Chen, X., Zhang, G., 2019), binary (Zhai,

X., Huang, H., Sze, N.N., Song, Z., Hon, K.K., 2019;

Salon, D., McIntyre, A., 2018; Jalayer, M.,

Shabanpour, R., Pour-Rouholamin, M., Golshani, N.,

Zhou, H., 2018; Rezapour, M., Moomen, M. and

Ksaibati, K., 2019; Sam, E.F., Daniels, S., Brijs, K.,

Brijs, T., G. Wets, 2018; Sam, E.F., Daniels, S., Brijs,

K., Brijs, T., G. Wets, 2018; Ahmed, M.M., Franke,

R., Ksaibati, K., Shinstine, D.S., 2018) and

polynomial (Penmetsa, P., Pulugurtha, S.S., 2018)

logistic regression, association rules (AR) (Montella,

A., 2011; Wu, P., Meng, X., Song, L., Zuo, W., 2019;

Weng, J., Zhu, J.-Z., Yan, X., Liu, Z., 2016; Nitsche,

P., Thomas, P., Stuetz, R., Welsh, R., 2017; Xu, C.,

Bao, J., Wang, C., Li, P., 2018). The accelerating

growth in computing power of computers and the

emergence of more sophisticated methods have

contributed to the rapid development of road safety

prediction models. Multivariate modeling and mining

methods are gradually replacing traditional one-

dimensional modeling methods based on the linear

model and the Poisson model.

When many researchers identify the relationship

of a large number of factors influencing to the

severity of the accident consequences, the method of

AR is widely used. So, as a result of research, the

authors of (Montella, A., 2011) identified the factors

leading to accidents at intersections and established

the interdependencies between these factors. In

general, they identified numerous factors related to

road and environmental problems, but not related to

pedestrian or vehicle. The most important factors

characterizing the geometry of the road were the

radius and angle of deviation. The significant role of

road markings and signs was also identified.

The authors of (Wu, P., Meng, X., Song, L., Zuo,

W., 2019) selected the city crossroads for analysis as

places that pose a serious security risk, since most

accidents within the city territory occur in places or

near junctions. They analyzed safety indicators for six

types of intersections and factors affecting the

severity of accidents. Fault tree analysis was used to

assess the risk of intersections, and AR were used to

analyze the nature of the severity of accidents. As a

result, four types of urban junctions with a high level

of accident risk and more than 4,000 rules describing

accidents with severe consequences were identified.

In (Weng, J., Zhu, J.-Z., Yan, X., Liu, Z., 2016), a

method based on AR is designed to analyze the

characteristics and factors contributing to emergency

situations during road repair work. Most AR include

conditions such as a speed of more than 40 km / h and

the use of traffic control devices.

The authors in the article (Nitsche, P., Thomas, P.,

Stuetz, R., Welsh, R., 2017) presents a data analysis

technique, including the preparation, analysis and

visualization of accident data, which allows

identifying critical pre-emergency scenarios at T -

and X- junctions as a basis for testing the safety of

autonomous vehicles. In this methodology, the k-

medoid method is used to form homogeneous groups

(clusters) among the array of accident records.

Subsequently to this clusters AR are applied to

generate typical motion scenarios and accident

patterns.

In (Xu, C., Bao, J., Wang, C., Li, P., 2018), the

method of AR was also used to study the factors

Association Rules to Identify Factors Affecting Risk and Severity of Road Accidents

615

contributing to the occurrence of serious traffic

accidents in China. The analysis showed that serious

accidents with victims are the result of complex

interactions between road users, the technical

characteristics of vehicles, the geometric parameters

of roads and environmental factors. As a result, the

authors received a number of reasons for the

occurrence of serious accidents with victims.

In Russian practice, as a rule, to analyze the

severity of injuries in road traffic accidents, models

based on the calculation of the total number of injured

(dead and wounded) are applied (Petrov, A., Petrova

D., 2016; Evtukov, S., Golov, E., Sazonova, T.,

2018), as well as visualization tools for these

indicators are used (Open data and traffic feedback,

2020, Accident map, 2020, Dornadzor, 2020). It

should also be noted that in the Russian Federation

there is no official universal procedure for the

comprehensive analysis of road accidents, approved

and adopted by government bodies. Despite this, the

duty of each subject of the road safety system is a

detailed analysis of the factors on the manifestation

of which this subject or the agencies involved with

him are able to influence.

In connection with the foregoing, today the search

for complex relationships between factors that affect

the severity of the consequences is an urgent task.

Therefore, the aim of this study is to analyze and

identify the causes that significantly affect the risk

and severity of accidents.

3 THE SOURCE DATA

Real data on accidents plays an important role in

improving the safety of road transport, since

information on the factors that led to malfunctions

and accidents is necessary to understand the causes of

malfunctions and how such events can be prevented

in the future.

In order to start any assessments of accidents, first

of all, it is necessary to have an appropriate

information array, the completeness and confidence

of which are of paramount importance in ensuring the

effectiveness of the analysis. Moreover, as noted in

(Imprialou, M., Quddus, M., 2019), insufficient

reporting of accidents is a recognized and studied

problem of road safety researches worldwide. This

fact is connected, firstly, with the failure to notify the

relevant state authorities about the accident, since its

participants agree to sign private payments for

insurance purposes, either there was no other-party

involvement (for example, an accident with

individual vehicles) or there were no obvious injuries

immediately after accident. And secondly, there is an

underestimation of the results of the road safety

analysis due to the non-presentation of certain

categories of injuries in the collected reports.

Consider the procedure for forming an data array

about accidents in the Russian Federation.

According to the Rules for the registration of road

traffic accidents (Decree of the Government of the

Russian Federation of June 29, 1995 N 647), all

accidents are divided from the point of view of

accounting into three groups.

The first group includes accidents in which people

died or were injured. Information about these

incidents is recorded in a special accident accounting

card and is included in the state statistical reporting.

According to the Decree of the Government of the

Russian Federation of November 19, 2008 N 859 “On

amendments to the rules for the registration of

accidents”, killed in the accident is person who

injured in an accident and who died within 30 days

from its consequences.

The second group includes accidents with

material damage without injuries, as well as an

accident in which people received bodily injuries that

did not cause any tangible harm to their health (while

the participants do not fall into the category of

“wounded”). Information about such accidents is not

included in the state statistical reporting, but is taken

into account and analyzed at the level of individual

cities and regions.

The third group includes individual incidents,

which, according to formal signs, can be qualified as

road transport, but information about them is not

included in the state statistical reporting, they are not

subject to accounting as road accidents.

However, to improve the degree of accounting, it

seems necessary to keep a complete record of

accidents with their classification, depending on the

severity of the consequences, as involving:

 material damage,

 light bodily harm,

 moderate to severe bodily injuries,

 the death of the victim,

 especially grave consequences (4 or more were

killed or 15 or more people were injured).

As the initial information, data collected by the

State Inspectorate for Road Safety in the Elabuga

town for 2017-2018 were used. According to the

classification of cities by population, Elabuga refers

to medium-sized towns (population 73,913 people).

The initial selection consisted of the following

factors, which were divided into four categories:

iMLTrans 2020 - Special Session on Intelligent Mobility, Logistics and Transport

616

 Characteristic of an accident - Type of

accident, traffic violations, injured and died,

children injured and died;

 Driver - Social characteristics of the driver,

Experience (years), Sex, Type and Degree of

intoxication (mcg / l), Direct traffic violations,

Related traffic violations;

 Vehicle - Number of vehicles involved in the

accident, Vehicle Type, Vehicle Brand and

Model, Faults, Damage;

 Road - Street, Deficiencies in the road

conditions, Type of road infrastructure in place,

Factors affecting traffic conditions (presence of

artificial bumps), Number of lanes, State of the

carriageway, Road profile, The lane in which

the accident occurred, Carriageway width,

Curb Width, Sidewalk Width, Dividing Strip

Width, View of the dividing strip;

 Environment - Lighting, Weather, Year,

Month, Day of the week, Hour.

4 METHODOLOGY

A graph

of the distribution of the accidents for 2017-

2018 was built. (Fig. 1). The most common types

were a collision, hitting a standing vehicle, hitting an

obstacle, hitting a pedestrian and exit from the road.

It is worth noting that the number of some accidents

in 2018 increased significantly (collisions), and the

number of less common ones was small.

Figure 1: A graph of the accidents grouped by the type of

accident for 2017-2018.

As a result of analyzing the histogram of the

distribution of the accidents by the number of victims

and by the type of accident, it was found that the

largest number of injured (wounded or dead) was

observed during collisions and hitting a pedestrian

(Fig. 2). Consequently, the question of finding the

main causes leading to an increase in the number of

collisions remains relevant. This is necessary for

adjustment the policy of regulating road safety.

Figure 2: Histogram of the distribution of the accidents by

the number of victims and by the type of accident for 2017-

2018.

Next, specific road sections of accident

concentration in Elabuga town were analyzed.

According to the results of the analysis of

accidents for a 2-year period, two of the most

dangerous sections of the road network were

identified, on which the largest number of accidents

occurred during the considered period:

- Mira street, b. 30 - b. 33;

- Neftyanikov street, b. 55 - b. 57.

It is worth noting that, in general, on these streets

the number of accidents has increased rapidly from

2017 to 2018 (Fig. 3).

Figure 3: A graph of the accidents in Mira street and

Neftyanikov street for 2017-2018.

When analyzing the distribution of accidents

occurred in these sections of accident concentration

by type, it is clear that on Mira street the most

common ones are collisions and hitting a standing

vehicle (Fig. 4). The main causes of road accidents

are violation of the rules for the vehicle location on

the roadway, wrong choice of distance, non-

compliance of lateral interval, non-compliance with

the order of travel (Fig. 5).

Association Rules to Identify Factors Affecting Risk and Severity of Road Accidents

617

Figure 4: A graph of the number of accidents on Mira street

grouped by type of accident for 2017-2018.

Figure 5: A graph of the number of accidents on Mira street

grouped by type of driving offence for 2017-2018.

It is worth noting that hitting a pedestrians with

serious consequences for human health were occurred

on this section of road. One of the contributing factors

is the multi-apartment residential buildings located

here, unregulated pedestrian crossing, as well as

places of attraction in the form of city malls (Fig. 6).

Figure 6: Road section of accident concentration on Mira

street, b. 30 - b. 34 for 2017-2018.

On Neftyanikov street for 2017-2018 the largest

number of accidents was recorded in Elabuga town

for the following types of accidents: collision and

hitting a standing vehicle (Fig. 7). The main driving

offences, as well as on Mira street, are violation of the

rules for the vehicle location on the roadway, wrong

choice of distance, non-compliance of lateral interval,

non-compliance with the order of travel (Fig. 8).

Figure 7: A graph of the number of accidents on

Neftyanikov street grouped by type of accident for 2017-

2018.

Figure 8: A graph of the number of accidents on

Neftyanikov street grouped by type of driving offence for

2017-2018.

On this section there are an unregulated pedestrian

crossing, a place of public transport stop, a place of

attraction in the form of an emergency department

(Fig. 9). The largest number of pedestrian accidents

with moderate to severe bodily injuries was recorded

here.

Such a type of accident as hitting a standing

vehicle at these dangerous sections can be explained

by a shortage of parking spaces for permanent storage

of the vehicle.

iMLTrans 2020 - Special Session on Intelligent Mobility, Logistics and Transport

618

Figure 9: Road section of accident concentration on

Neftyanikov street, b. 55 - b. 57 for 2017-2018.

In these sections of Mira street and Neftyanikov

street in the morning rush hour, the intensity of

vehicle traffic was measured, presented in Table 1.

Analysis of the capacity of these shows that

despite their low utilization coefficient (which is

equal to the ratio of traffic intensity to crossing

capacity), not exceeding 35% of the design capacity,

these sections occupy the first positions in terms of

accident rate. Therefore, we can assume the presence

of other factors that are not related to traffic intensity,

leading to a large number of accidents.

Table 1: Traffic intensity in the morning rush hour on Mira

street and Neftyanikov street.

Street

Traffic

intensity,

units / h

Crossing

capacity,

units / h

Utilization

coefficient

Prospect

Neftyanikov street

(from the side of

Proletarskaya

street

)

1265 3600 0,35

Mira street 705 2400 0,29

Neftyanikov street

(from the side of

Stroiteley street)

1126 3600 0,31

In this regard, a deeper analysis of the

interdependencies between the factors affecting the

occurrence of various types of accidents in these

sections was carried out. Most of the available

accident data used in this study is categorical, that is,

it is described by qualitative factors (also called

nominal factors). Although categories can be encoded

as numbers, for example, 1-woman, 2- man, these

numbers will not have mathematical meaning.

Therefore, special methods are needed to analyze

categorical data.

Recently, interest in the methods of “discovering

knowledge in databases” has been growing. One of

the common analytical methods of data processing is

affinity analysis. The method determines the mutual

relations between events occurring jointly.

One of the applications of affinitive analysis is

market basket analysis in order to detect associations

between different data, for example finding rules to

quantify the relationship between two or more data.

Such rules are called AR.

The basic concept in the theory of AR is a

transaction - a set of events that occur together.

An association rule consists of two sets of objects

called a antecedent and a consequent, written in the

form X → Y, which reads “from X follows Y”. Thus,

an association rule is formulated in the form "If an

antecedent, then a consequent." The antecedent is

often limited to the content of only one subject.

AR describe the relationship between sets of items

matching the antecedent and the consequent. This

relationship is characterized by two indicators -

support and confidence. Let D be a transaction

database, and N - the number of transactions in this

database. Each transaction Di represents a certain set

of items. S is a support, C is a confidence. Association

rule support is the number of transactions containing

both an antecedent and a consequent.

For example, it can written for the association

A → B:

S(A→B) = P(AՈB) (1)

A confidence of an association rule is a measure

of the rule accuracy, which is determined as the ratio

of the number of transactions containing both the

antecedent and the consequent to the number of

transactions containing only the antecedent. For

example, it can written for the association A → B:

С(A→B) = P (A|B) = P(AՈB)/P(A) (2)

If the support and confidence are high enough,

then this makes it possible to assert that any future

transaction that includes an antecedent will also

contain a consequent. However, it is also necessary to

estimate the degree of independence of the antecedent

and the consequent in order to avoid the situation of

obtaining “fictitious” rules, when the support and

confidence are high.

A lift is the ratio of the frequency of occurrence of

a antecedent in transactions, which also contain a

consequent, to the frequency of occurrence of the

consequent as a whole. A lift is defined as follows:

L(A→B) = C (A →B)/S(B) (3)

Association Rules to Identify Factors Affecting Risk and Severity of Road Accidents

619

In our case, association detection is the

identification of a combination of factors leading to

an accident. To identify a combination of the most

common factors that led to severe injuries or death of

road accident participants, 75 records of the accidents

were selected. The AR method was applied to these

data. By experience, the minimum values of support

were established - 4%, confidence - 60%, lift - 1, the

power of the rule set - 4, at which 152 rules are

formed - the amount acceptable for consideration and

interpretation. From a practical point of view the most

interesting and useful among the rules obtained are

given in table 2.

Table 2: The most informative AR.

Antecedent Consequent Support

Confi-

dence

Lift

"Zhiguli" VAZ-

2108, 09 and

modifications

Friday 4,11 60 3,982

November

Okruzhnoe

highway

4,11 60 4,867

Front Left Side

And Female

Wounded and

Female

4,11 60 4,867

Septemb

Rain 4,11 60 5,475

Gas station

Naberezhnye-

Chelninskoe

highway

4,11 60 6,257

Curb Width –

25 m

Friday 4,11 75 4,977

Ground stripe Mira street 5,48 66,7 6,083

February

Lacks of

winter street

cleaning and

Snowfall and

Curb Width –

30 m

5,48 66,7 9,733

Public transport

stop And Lack,

poor

distinguishability

of horizontal

marking of the

carriageway

Improper

usage, poor

visibility of

road signs

5,48 80 8,343

Lack of pedestrian

fences in the

required places

Adjustable

intersection

8,22 60 3,65

Adjustable

pedestrian

crossing

Lack of

pedestrian

fences in the

required places

11 61,5 4,492

The rules were filtered out by support, confidence and

lift.

An analysis of some of the obtained rules shows

the direct relationship of accidents with severe

consequences and the lacks of winter street cleaning

in February, leading to the appearance of snowfall on

the road. In three accidents in September it was

raining, also worsening road conditions. In eight

cases, in the immediate vicinity of the regulated

pedestrian crossing, there were no fences in the

required places, which could lead to violation of

traffic rules by pedestrians. The combination of “an

adjustable intersection” and “lack of pedestrian

fences in the necessary places” is also quite common

- in 6 cases. In 4 cases, grave consequences were

caused by poor visibility of the horizontal marking of

the roadway and poor visibility of road signs near a

public transport stop. On Friday, in three cases, an

accident occurred with the participation of VAZ-

2108, 09 “Zhiguli” and modifications, as well as with

a curb width of 25 meters. In three cases, injuries

occurred as a result of a hitting to the left side, that is

from the driver. Also, according to the received rules,

accidents with severe consequences on Mira street

more often occur in places where this street has

ground stripe (4 cases), on Naberezhno-Chelninsky

highway at a gas station (3 cases) and on Okrugny

highway in November (3 cases).

5 CONCLUSION

When developing a road safety policy, the

government of any developed or developing country

needs to analyze factors that affect not only the

likelihood of an accident, but also determine the

severity of the consequences.

To identify the most traumatic types of traffic

accidents, graphs of the number of accidents by types

of accidents and the number of victims for 2017-2018

were constructed. It was found that the greatest

number of injured (wounded or dead) is observed in

collisions and hitting a pedestrian. Then, road

sections of the concentration of accidents in Elabuga

town were analyzed, the most common types of

accidents and the main violations of traffic rules that

contribute to the occurrence of accidents were

determined. At the same time, the analysis of the

traffic intensity of these streets showed their low load.

Therefore, to identify hidden relationships between

factors and accidents with severe consequences, AR

were applied. As a result, the influence of weather

conditions, quality of road infrastructure and marking

was established.

The resulting rules will allow to build simulation

models, with the help of which, and taking into

account the different situations that occur in a real

transport system, it can be possible to determine the

most effective measures to prevent accidents and

mitigate the severity of the consequences.

iMLTrans 2020 - Special Session on Intelligent Mobility, Logistics and Transport

620

ACKNOWLEDGMENTS

This work was supported by the Russian Foundation

for Basic Research: grant No. 19-29-06008\19.

REFERENCES

Accident map. https://dtp-stat.ru/(accessed 05.01.2020).

Ahmed, M.M., Franke, R., Ksaibati, K. and Shinstine, D.S.,

2018. Effects of truck traffic on crash injury severity on

rural highways in Wyoming using Bayesian binary logit

models. In Accident Analysis & Prevention, vol.117,

pp. 106-113.

Chen, C., Zhang, G., Qian, Z., Tarefder, R.A., Tian, Z.,

2016. Investigating driver injury severity patterns in

rollover crashes using support vector machine models.

In Accident Analysis & Prevention, vol.90, pp. 128–

139.

Chen, C., Zhang, G., Yang, J., Milton, J.C., Alcántara,

A.D., 2016. An explanatory analysis of driver injury

severity in rear-end crashes using a decision

table/Naïve Bayes (DTNB) hybrid classifier. In

Accident Analysis & Prevention, vol. 90, pp. 95-107.

Decree of the Government of the Russian Federation of

June 29, 1995 N 647. Meeting of the legislation of the

Russian Federation, 1995, N 28, art. 2681.

Djahel, S., Doolan, R., Muntean, G.-M., Murph, J., 2015. A

Communications-oriented Perspective on Traffic

Management Systems for Smart Cities: Challenges and

Innovative Approaches. In IEEE Communications

Surveys & Tutorials, vol. 17, iss. 1, pp.125-151.

Dornadzor. https://dornadzor-sz.ru/(accessed 05.01.2020).

European Initiative on Smart Cities, 2010-2020,

http://setis.ec.europa.eu/set-plan-

implementation/technologyroadmaps/european-

initiative-smart-cities.

Evtukov, S., Golov, E., Sazonova, T., 2018. Prospects of

scientific research in the field of active and passive

safety of vehicles. In MATEC Web of Conferences, vol.

239, paper № 040182018.

Imprialou, M., Quddus, M., 2019. Crash data quality for

road safety research: Current state and future directions.

In Accident Analysis & Prevention, vol.130, pp. 84-90.

Jalayer, M., Shabanpour, R., Pour-Rouholamin, M.,

Golshani, N., Zhou, H., 2018. Wrong-way driving

crashes: A random-parameters ordered probit analysis

of injury severity. In Accident Analysis & Prevention,

vol.117, pp. 128-135.

Li, Z., Wu, Q., Ci, Y., Chen, C., Chen, X., Zhang, G., 2019.

Using latent class analysis and mixed logit model to

explore risk factors on driver injury severity in single-

vehicle crashes. In Accident Analysis & Prevention,

vol.129, pp. 230–240.

Montella, A., 2011. Identifying crash contributory factors

at urban roundabouts and using association rules to

explore their relationships to different crash types. In

Accident Analysis & Prevention, vol. 43, iss. 4, pp.

1451-1463.

Moral-García, S, Castellano, J. G., Mantas, C. J., Montella,

A., Abellán, J. 2019. Decision Tree Ensemble Method

for Analyzing Traffic Accidents of Novice Drivers in

Urban Areas. In Entropy, vol. 21(4), pp.360.

Nitsche, P., Thomas, P., Stuetz, R., Welsh, R., 2017. Pre-

crash scenarios at road junctions: A clustering method

for car crash data. In Accident Analysis & Prevention,

vol.107, pp. 137-151.

Open data and traffic feedback. Безопасные дороги.рф.

(accessed 05.01.2020).

Penmetsa, P., Pulugurtha, S.S., 2018. Modeling crash injury

severity by road feature to improve safety. In Traf Inj

Prev, vol. 19, iss. 1, pp. 102-109.

Petrov, A., Petrova D., 2016. Assessment of Spatial

Unevenness of Road Accidents Severity as Instrument

of Preventive Protection from Emergency Situations in

Road Complex. In IOP Conf Ser: Materials Sc Eng, vol.

142, iss. 1, paper № 012116.

Rezapour, M., Moomen, M. and Ksaibati, K., 2019. Ordered

logistic models of influencing factors on crash injury

severity of single and multiple-vehicle downgrade

crashes: A case study in Wyoming. In Jour Saf Res, vol.

68, pp. 107-118.

Salon, D., McIntyre, A., 2018. Determinants of pedestrian

and bicyclist crash severity by party at fault in San

Francisco, CA. In Accident Analysis & Prevention, vol.

110, pp. 149-160.

Sam, E.F., Daniels, S., Brijs, K., Brijs, T. and G. Wets, 2018.

Modelling public bus/minibus transport accident

severity in Ghana. In Accident Analysis & Prevention,

vol. 119, pp. 114-121.

Theofilatos, A., Chen, C., Constantinos, A., 2019.

Comparing machine learning and deep learning methods

for real-time crash prediction. In Journal of the

Transportation Research Board, Vol.2673 issue: 8, pp.

169-178.

Weng, J., Zhu, J.-Z., Yan, X., Liu, Z., 2016. Investigation of

work zone crash casualty patterns using association rules.

In Accident Analysis & Prevention, vol.92, pp. 43-52.

Wu, P., Meng, X., Song, L., Zuo, W., 2019. Crash Risk

Evaluation and Crash Severity Pattern Analysis for

Different Types of Urban Junctions: Fault Tree Analysis

and Association Rules Approaches. In Trans Res Rec,

vol. 2673, iss. 1, pp. 403-416.

Xu, C., Bao, J., Wang, C., Li, P., 2018. Association rule

analysis of factors contributing to extraordinarily severe

traffic crashes in China. In Journal of Safety Research,

vol. 67, pp. 65-75.

Zhai, X., Huang, H., Sze, N.N., Song, Z. and Hon, K.K.,

2019. Diagnostic analysis of the effects of weather

condition on pedestrian crash severity. In Accident

Analysis & Prevention, vol. 122, pp. 318-324.

Zheng, M., Li, T., Zhu, R., Chen, J., Ma, Z., Tang M., et al.

2019. Traffic accident’s severity prediction: a deep

learning approach based CNN network. In IEEE Access

PP (99), vol.7, pp. 39897-39910.

Zou, X., Yue, W. L., Vu, H.L., 2018. Visualization and

analysis of mapping knowledge domain of road safety

studies. In Accident Analysis & Prevention, vol.118, pp.

131-145.

Association Rules to Identify Factors Affecting Risk and Severity of Road Accidents

621