1. Introduction
Unsafe and insufficient road infrastructure is a fundamental issue for the occurrence of road traffic crashes (RTC) and their outcome. RTC occurrences are significantly influenced by road geometric formation [
1]. It can be fragmented into alignment, profile, and cross-section. Mostly, road width, cross slope, road margins, traffic separators, and curbs can be considered basic physical elements [
2,
3]. The intention of geometric design is to optimize efficiency and safety so that it minimizes cost and environmental damage.
To analyze the impact of road geometric formation on traffic crashes and their severity levels, the study used recent 5-year Budapest city road traffic crash data that was collected from Hungary’s transport authority from 2017 to 2021. The study area was selected due to different factors such as the nature of the data, availability, and convenience of data. In Budapest city, there was a high concentration of traffic crashes that were recorded yearly. In the past 5 years, around 17,006 road traffic crashes have been registered. The crash’s outcome has been classified as fatality (216), serious injury (3999), and minor injury (12,791).
Different studies showed that road infrastructure, mostly geometry formation, had its own impacts on road traffic crashes. Pembuain et al. (2018) indicated that the elements in road infrastructure formation had a significant effect on the risk of road traffic accidents [
4]. The report in Australia showed that the road is a causation factor in about 30% of all crashes [
5]. A road defect directly triggers a crash, where some element of the road environment misleads a road user and thereby creates human errors [
6]. The study showed that two-lane rural highways reduce crash rates by 44% versus high crash-rate infrastructure, at the 99% confidence level [
7]. One of the important and cogent measures for reducing road crash fatalities is continuously improving and maintaining the good shape and condition of our roads [
8]. Road safety can be improved by implementing principles of road safety infrastructure management (RIS) [
9]. It showed that road geometry elements can mislead road users. Two-lane roads highly contribute to crashes; in this case, it is important to examine the impact of other lane formations on RTC and its severity level.
A study in Texas indicated that severe crashes are likely to occur on horizontal curves with higher degrees of curvature compared to curves with smaller degrees of curvature [
10,
11,
12]. Sarbaz Othman et al. (2009) stated that large-radius right-turn curves were more dangerous than left curves during lane-changing maneuvers. However, sharper curves are more dangerous in both left and right curves [
13]. Overtaking on right curves was sensitive to the radius and the interaction of the radius with road conditions, while the left curves were more sensitive to super-elevation [
14]. Even though the study shows that horizontal curves are a cause of severe road traffic crashes, it is better to analyze the overall impact of curved roads on RTC compared to other road geometric formations.
On straight roads, speed and distance were influenced by road traffic accidents. The longest distance offered the highest risk of fatal injuries [
15]. Willy et al. (2020) discovered a strong relationship between side freedom and accident number [
16]. MBESSA Michel et al. (2020) stated that the reduction in road geometric formation has a significant impact on crash occurrence [
17]. In Singapore, road crashes at intersections contribute around 35% of the reported accidents and show that vehicle type, road type, collision type, driver’s characteristics, and time of day are important determinants of the severity of crashes at intersections [
18]. The study in the U.S. showed that the relative ratio analysis showed that intersection-related crashes were almost 335 times as likely to have “turned with an obstructed view” as the critical reason for non-intersection-related crashes [
19]. So, road geometric formations, such as horizontal curves, side freedom, intersections, and straight roads have their own impact on the occurrences of road traffic crashes. It was better to visualize the interaction between stated variables and the number of lanes with intersection type.
Abbasi et al. (2022) stated that developing artificial lighting at intersections and LED-raised pavement markers on two-lane rural roads could lead to enhanced road safety under dark LCs [
20]. Mehdi Hosseinpour et al. (2013) study results of REGOPM on crash severity showed that horizontal curvature, paved shoulder width, terrain type, and side friction were associated with more severe crashes, whereas land use, access points, and the presence of a median reduced the probability of severe crashes [
21,
22]. In this case, further analysis was required to define the effect of natural light conditions on RTC.
Different research on road geometric formation and its impact on road traffic crashes showed different attributes and their level of contribution. Most of the studies focused on a microscopic level, such as selected road segments, rural road networks, intersections, specific numbers of lanes, small road networks, etc. According to my knowledge, there was a research gap on the impact of road geometric formation on road traffic crashes and their severity levels at a macroscopic level, such as at the country or city level. Budapest City was selected due to the fact that the area was urban, with highly networked roads, large territory, and has a high concentration of vehicles and a dense population. In fact, the number of road traffic accidents recorded was also high, and the nature of the data and its accessibility encourages the choice of the city. Moreover, as per study review in Budapest city, the impact of road geometric formation on road traffic crashes at the macroscopic level and their severity rate was not taken into consideration. In spite of that, further investigation was needed to direct the impacts of road geometric formation on road traffic crashes and their severity levels.
The main objective of this study was to analyze the impact of road geometric formation on road traffic crashes and their severity levels. For further analysis, the study emphasizes the road’s geometric formation (straight, intersection, horizontal curvature, number of lanes, etc.) and number of lanes. In addition to that, this study tried to highlight the probable causes of road traffic crashes and the relationship between variables. The study used relative frequency distribution for statistical analysis, multinomial logistic regression to analyze the main determinant factor of a road traffic crash, and multilayer perceptron artificial neural network (MLP-ANN) to demonstrate the influences and interaction of road geometric formation and the number of lanes on a road traffic crash and its outcome. In addition to that, it used the severity index (SI) to show the severity level of road geometric formation and the number of lanes that lead to a road traffic crash and its outcome. For explanation, description, and summarizing the output, the study used both inferential and descriptive statistics.
3. Result and Discussion
These sections of the study attempt to identify the output of the analysis and discuss the findings. It contains the road traffic crash distribution and its frequency in terms of independent variables, and cross-tabulation between dependent and independent variables to signify the impact of other variables on road traffic crashes and their outcome. Moreover, it cross-tabulates the independent variables against each other. In general, this study tried to bring reliable findings and their implications that define the impact of road geometric formation on traffic crashes and their severity level. For more information, please see the study findings discussed below:
3.1. Road Network and Traffic Crash Distribution
Figure 1 shown below indicates that since 2017, road traffic crashes in the past 5 years have been evenly distributed on the road network. It indicated that a high concentration of road traffic crashes was recorded in the downtown (center) of the city. To locate the concentration of road traffic crashes, the study used Q-GIS.
3.2. Road Traffic Crash District Distribution and District Level of Concentration
The study site has 23 districts. The study tried to rate the level of a road traffic crash in percentage based on its concentration. As a result, the study grouped the area into three; 0–3%, 3–6%, and 6–9% as high, intermediate, and low, respectively. The maximum and minimum road traffic crash concentrations were observed in District XIV with 8.5% and XXIII with 1.9%, respectively. For more information, see
Figure 2 below and
Table A1.
Based on the above
Figure 2b, the red color indicates a highly concentrated road traffic crash that was observed in districts X, XI, XIII, and XIV. Relatively, the green- and orange-colored districts had a low and intermediate concentration of road traffic crashes and their outcomes.
As a result, further investigation and remedial action were needed in those stated districts by the stakeholders to minimize road traffic crash severity in the study area. 3.3. Road Geometry Formation and Frequency of Road Traffic Crash
This part of the study deals with the frequency of road traffic crashes in line with the geometric form of the road.
Table 2 shown below indicates how road traffic crashes and their outcomes vary with road geometric formation.
According to
Table 2, approximately 63.7% and 31.3% of road traffic accidents occurred on straight and intersection sections of the road network, respectively. Even if a high number of road traffic crashes are observed on the straight road segments, the intersection part of the road also contributes to significant road traffic crashes. As a result, the stakeholders must identify causes and problems on the straight part of the road to minimize road traffic crashes in the study area. Crashes are not only high at the straight part of the road, in comparison with areal coverage, the rate of crashes at intersections is relatively high. Further investigation was also needed at the intersection part of the road.
3.4. Road Geometry Formation and Road Traffic Crash Outcome
As shown in
Table 3 below, a high number of fatalities and injuries were registered on the straight part of the road segment. A relatively significant number of fatalities and injuries are also registered at intersections. Even though the straight section of the road had a high number of accidents, in terms of areal coverage, the intersection part of the road also played a significant role in the occurrences of road traffic accidents. As a result, to minimize the number of deaths, further investigation was needed on both straight and intersection parts of the road.
3.5. Road Geometric Formation and Severity Level of Traffic Crash
Distinct from accident frequency, the crash severity index provides the severity of each crash outcome registered during a specific time. Based on Equation (1) above, this study tried to analyze the severity level of the road geometric formation based on the number of deaths and injuries.
Table 4 shown above indicates that the number of road traffic crash outcomes varies in different parts of the road section. A high level of severity in terms of death, serious and slight injuries is observed at the horizontal curve, straight and intersection parts of the road, respectively. Even if the number of deaths is high on the straight part of the road, the level of severity in terms of deaths cannot indicate this part of the road. As a result, in the specified study area, horizontally curved road geometry was the most severe. In order to reduce the severity level of the road in the study area, additional research into the horizontal curve of the road network was required.
3.6. Multinomial Logistic (MNL) Regression
On the basis of the nature of the data, this study categorized road traffic crashes and their outcomes as dependent and independent variables. The outcomes of the crash were fatalities, serious injuries, and slight injuries.
The result of the analysis indicated that there was a relationship between road traffic accidents and their potential determinants. The results of the cause-effect analysis of the variables listed in
Table 5 revealed that fatality is highly related to light conditions, collision type, alcohol consumption, and speed limit. Meanwhile, serious injuries are highly related to collision type, road geometric formation, and the reason for the occurrence of road traffic crashes. It considers slight injuries as the reference category. So, light condition, collision type, alcohol consumption, road geometry formation, speed limit, and reason for road traffic crashes were determinant factors and had a significant effect on the occurrences of road traffic accidents at a
p-value of 0.05.
As a result, stakeholders must concentrate on those factors in order to reduce road traffic accidents. On the basis of the above findings, the model that helps to determine the significant level of road traffic accidents can be expressed using the following traffic accident (
A) predicting equation.
3.7. Collision Type and Road Traffic Crash Outcome
A collision in transportation happens between a vehicle and an object, a vehicle and a pedestrian, etc., and that plays a significant role in the occurrences of a road traffic crash. As shown in
Table 6 below, collisions of vehicles with pedestrians result in a high number of road traffic fatalities. Collisions with vehicles, particularly rear-end collisions, play a significant role. Not only were there fatalities, but there were also numerous injuries as a result of rear-end collisions, and vehicle collisions with pedestrians.
As a result, a high number of road traffic deaths and injuries were registered due to collisions happening between vehicles and pedestrians, vehicles and vehicles (rear-end collision). Further investigation was needed to minimize pedestrian deaths and injuries.
3.8. Artificial Neural Network (ANN): Multilayer Perceptron (MLP)
In this study, MLP-ANN was used to analyze and show the relationship between dependent and independent variables. To extract the output, the study used Statistical Package for Social Science (SPSS-20). It provides specific information on which variables have a significant impact on the occurrence and outcome of road traffic accidents.
Figure 3 shown below indicates the impact of road geometric formation and the number of lanes on road traffic crashes and their outcome.
The model attempted to understand the relationship between the training data and be evaluated on the test data. In this case, 70% of the data is used for training and 30% for testing. The output of the model summary between the input of road geometry and number of lanes and the output of road traffic accidents indicated that the percent of incorrect predictions for training was 25%.
Table 7 below indicates that in both cases, the incorrect prediction for testing was less than 25%. So, the model was a good fit with a training and testing error of 25%. This shows that the model prediction level was correct and accurate above 75% [
40].
The blue and gray lines in
Figure 3 show the positive and negative bond between the dependent and independent variables, whose synaptic weight is >0 and <0, respectively. In this diagram, road geometry, and the number of lanes were input (independent variables), and crash outcome (road traffic accident) was output (dependent variable).
The intersection part of the road has a strong and positive contribution to the occurrences of road traffic crashes that cause fatalities, whereas the horizontal curve of the road network has a strong and positive impact on the occurrence of slight injuries. The straight section of the road has a positive but relatively minor impact on the occurrence of traffic accidents.
In addition to that, one-lane and two-lane roads had a strong and positive contribution to the occurrence of road traffic injuries. Simultaneously, three-lane and above-average roads have a significant positive and strong impact on the occurrences of minor injuries. It also has a relatively positive and slightly stronger impact on the occurrences of death. So, this indicates that road geometric formation has an impact on the occurrence of road traffic injuries.
3.9. Road Traffic Crash Level of Severity and Road Geometric Formation
The outcome of the crash analysis indicated that a highly severe accident was registered on a straight road. Based on
Table 8 shown below, from total road traffic crash outcomes, a high number of deaths and injuries were registered on the straight road sections. Relatively, in line with coverage, the level of accidents and injuries was also high at the intersection part of the road.
As a result, further investigation was needed to minimize the number of deaths and injuries that happen at straight and intersection parts of the road network in the study area.
3.10. Road Geometric Formation and Related Causes of Road Traffic Crash
As shown in
Table A3, a high number of road traffic crashes was registered due to the improper use of traffic control devices such as traffic signs, signals, and marks. The causes and primary reason for the occurrences of a road traffic crash at an intersection, horizontal curve, and straight part of the road segment was improper use of road sign, road pavement condition, and stopping sight distance problem, respectively. So, improper use of road traffic control devices has a significant impact on the occurrences of a road traffic crash that causes enormous loss of life and physical damage.
3.11. Road Geometric Formation and Traffic Collision Type
Road traffic collisions are a probable cause of road traffic crashes and accidents in the transportation system.
Table 9 clearly defines the relationship between road geometric formation and road traffic collisions. Based on statistical data collected from secondary sources, collisions between vehicles(rear-end collisions) highly happen in all road geometric formations of the road network. In addition to that, the collision of vehicles with pedestrians plays a great role in the occurrence of road traffic crashes and accidents. Mostly, road traffic crashes happen on straight road segments that are the result of rear-end collisions.
As a result, a high number of road traffic crashes were registered due to collisions happening between vehicles, and vehicles with pedestrians on a stated road geometric formation. Mostly, vehicle collisions, such as rear-end collisions, play a significant role in the occurrences of road traffic crashes on straight and intersection segments of the road.
3.12. Road Geometric Formation and Traffic Crash Hourly Distribution
Table A2 indicated that in 1-hour distribution, a high number of road traffic crashes were registered from 16:01–17:00. Concomitantly, in the 3-hour distribution in
Table 10 below, the maximum crash was registered at 15:01–18:00. As a result, this study considers 16:01–17:00 as a peak hour for road traffic crashes at all road geometric formations.
3.13. Road Traffic Crash Frequency and Number of Lane
According to
Table 11, a high number of road traffic crashes were observed at one-lane road geometric formations that accounted for more than 55.85%.
3.14. Road Traffic Crash Outcome and Number of Lane
Table 12 shows that one-lane road geometry formations had the highest number of road traffic deaths and injuries. It shows that as the number of lanes increased, there was a reduction in the number of crashes and their outcome. So, road with three lanes and above has minor role in the occurrence and outcome of the crash.
3.15. Interaction of Road Geometric Formation and Number of Lane on Traffic Crash
Table 13 below shows that although the number of road traffic crashes on a one-lane road is high, in all lane formations, a high number of road crashes and their outcomes were registered at the straight and intersection parts of the road geometry formation.
3.16. Number of Lane and Severity Level of Road Traffic Crash
Based on equation 1 stated above, this study tried to analyze the severity level of the road lane number based on the number of road traffic fatalities and injuries registered.
Table 14 shown above indicates that even if the number of road traffic crash outcomes varies according to the number of lanes, a high level of severity in terms of fatalities, serious injuries, and slight injuries was observed at three and above, one-, and two-lane roads, respectively. Even if the number of deaths and slight injuries is high on one-lane roads, the level of severity is high at three-lane and two-lane road geometry configurations.
3.17. Causes and Hourly Distribution of Road Traffic Crash
As shown in
Table A4 below, a high number of road traffic crashes were registered due to the improper use of traffic control devices such as traffic signs, signals, and marks. In the 3-hour distribution, a high road traffic crash happened at 15:01–18:00 due to the improper use of a road traffic control device. In addition to that, a high concentration of road traffic crashes was registered at 16:01–17:00 based on a 1-hour distribution, and the study considers this time as the peak hour for the occurrences of road traffic crashes.
3.18. Causes and Road Traffic Crash Outcome
Road traffic accidents such as fatalities, serious injuries, and slight injuries are the outcomes of road traffic crashes that happen for different reasons. Based on
Table A5 below, the study indicated that the maximum number of deaths and injuries registered were caused by the improper use of road traffic signs.
3.19. Frequency of Road Traffic Crash and Responsible Body
As shown in
Table 15, different participants contributed to the occurrences of the road traffic crash to varying degrees. This study indicated that drivers contributed more than 82.6% of the road traffic crash frequency in the study area. This shows the driver was responsible for the death and injuries of road users. Furthermore, pedestrians also play a significant role in the occurrence of road traffic accidents.
As indicated above, most road traffic crashes happen due to the improper use of road traffic signs by drivers and pedestrians. As a result, it was advisable for the stakeholders to train road users, mostly drivers, how to properly use road traffic control devices and apply law enforcement that is used to minimize the occurrences of road traffic crashes and their outcomes in the study area.
3.20. Road Traffic Crash Outcome and Responsible Body
Table 16, shown below, indicates that drivers can contribute to a huge loss of life and injuries. As a result, the driver bears a large portion of the blame for road traffic accidents that result in significant loss of life and physical harm to other road users.
3.21. Road Geometric Formation and Responsible Body in Traffic Crash
Table 17, shown below, indicates that the driver was highly responsible for the occurrence of road traffic crashes on all parts of the road network. In addition to driver problems and errors, vehicular failure plays a vital role in the occurrence of road traffic crashes at the intersection parts of the road network. Pedestrians also play a significant role in the occurrence of road traffic crashes. So, drivers and pedestrians have a significant role in the occurrences of road traffic crashes at the stated road geometric formation.
4. Conclusions and Recommendation
Road geometric formation has an impact on the occurrences of road traffic crashes and their outcomes. The aim of this study was to analyze the impact of road geometric formation on road traffic crashes and their outcome. For further analyze the impact of road geometric formation on road traffic crash and its outcome, the study used Budapest city road traffic crash data based on the nature, convenience, and availability. The study used tools such as Microsoft Excel, the Statistical Package for Social Science (SPSS-20), and Quantum Geographic Information System (Q-GIS) for data organization and analysis. Multinomial logistic (MNL) regression is used to identify determinants of road traffic crashes. The Severity Index (SI) is used to rate the level of severity of determinant factors. A Multilayer Perceptron Artificial Neural Network (MLP-ANN) is used to analyze the determinant factors and the impacts of road geometric formation. The study used both inferential and descriptive statistics.
Multicollinearity tests, p-value, overdispersion, and other statistical model testing were undertaken to analyze the significance of the data and variable for modeling and analysis. This study considers a variable which had a correlation coefficient less than 80%. The maximum variable similarity was observed between light conditions and hourly distribution with a value of 28.9%. Alcohol consumption and collision type were also responsible for 24.5% of the similarities. As a result, all variables were used for analysis and modeling.
Both dependent and independent variables for model selection were undertaken with an overdispersion test by comparing the mean and variance of the data. The analysis indicated that more than 80% of the data was not overdispersed. As a result, multinomial logistic regression was selected for analysis determinant factor. In addition, the p-value was used to determine the level of accuracy. The variable used for model development was considered with a p-value of 0.05. This means the accuracy level of the analysis was more than 95%.
In Multilayer Perceptron Artificial Neural Network analysis, the percentage of incorrect predictions both for testing and training was analyzed. The analysis indicated that the percentage of incorrect predictions was less than 25%, which shows the accuracy level of the model was 75% and above. So, the model was a good fit with some training and testing errors.
The result of the study indicated that a high frequency of road traffic crashes and their outcomes are observed in straight road geometric formations that account for around 63.7%. The horizontal curve of the road is a severe road geometric. In addition to that, a high number of road traffic crashes and their outcomes were registered on one-lane roads. However, the severity level was high at three-lanes and above.
The Multinomial Logistic (MNL) Regression Model output indicated that light conditions, collision type, alcohol consumption, speed limit, and road geometric formation have significant impacts on the occurrences of road traffic crashes and their outcomes. In terms of collisions, a high number of road traffic deaths and injuries were registered due to collisions happening between vehicles and vehicles with pedestrians. Vehicle collisions (especially rear-end collisions) play an important role in the occurrence of road traffic crashes at straight and intersection parts of the road geometry.
According to the Multilayer Perceptron Artificial Neural Network (MLP-ANN), road horizontal curved geometry has a positive and significant relationship with road traffic fatalities. The intersection and horizontal curve had positive interactions and a strong impact on the occurrences of slight injuries. Concomitantly, the one-lane road has positive and strong interaction with road traffic and causes serious and slight injuries. Meanwhile, two-lane roads had a positive and strong interaction with road traffic and slight injuries. In this case, one-lane and three-lane roads have a positive impact on the number of fatalities on the road.
While comparing the statistical analysis results of the models stated above, even if in MNL-regression model, road geometry was not a determinant factor for fatal road traffic accidents, the MLP-ANN model indicates that intersection road geometric formation has a strong and positive impact on the occurrences of fatal road traffic accidents. While the severity index analysis indicated that the highest severity value was registered on the horizontal curvature of road geometry, in contrast, the MLP-ANN showed that the impact of horizontal curve road geometry has a strong and positive impact on the incidence of slight injuries. On three-lane and above roads, which are highly severe road networks in this study, the finding of the MLP-ANN model supports that three-lane above-roads had a slightly strong and positive impact on the occurrences of fatal road traffic accidents.
A high number of road traffic crashes were registered due to the improper use of traffic control devices. The primary reasons for the occurrences of a road traffic crash at an intersection, horizontal curve, and straight road geometric formation were the improper use of road traffic signs; road pavement condition, and stopping sight distance problems, respectively. For the occurrences of road traffic crashes and their outcomes, 16:01–17:00 was a peak hour at all geometries of the road. In this study, the driver played a vital role in the occurrences of road traffic crashes and their outcomes accounted for more than 82.6%. This shows the driver was responsible for the occurrences of road traffic crashes and their outcomes in all of the road geometric formations. Since most road traffic crashes and their outcomes happen due to the improper use of road traffic signs, this study recommends and advises stakeholders to train road users, primarily drivers, on how to properly use road traffic control devices and respect rules and regulations to reduce the occurrences and outcomes of road traffic crashes.
In addition to that, to overcome this problem, further investigation is needed to understand the reason behind why road users were not properly using traffic control devices, and detailed remedial action must be undertaken on traffic control device usage. In general, the stakeholders must emphasize the above-indicated determinant factors to mitigate the problem.