Association Rules Between Urban Road Traffic Accidents and Violations Considering Temporal and Spatial Constraints: A Case Study of Beijing

Abstract

Traffic violations are among the leading causes of accidents and significantly compromise urban road safety. This study analyzed traffic violation and incident data collected by automated enforcement systems in urban Beijing from 2019 to 2023, consisting of 3264 traffic accident records and 147,876 traffic violation records. Through a spatiotemporal data association method, 2126 violations directly associated with accidents were identified. The FP-growth algorithm was then applied to derive 18 robust association rules encompassing five categories of accidents and four categories of violations. The findings indicate that the correlation between traffic accidents and violations displays clear peak periods during the morning (8:00–9:00) and evening (17:00–18:00). Violations such as red light running, stopping beyond the stop line during a red light, and ignoring prohibitions strongly correlate with traffic accidents under specific spatiotemporal conditions. Illegally parked vehicles not only reduce road transport efficiency but also significantly elevate the risk of traffic accidents in the surrounding area. The association rules identified in this study can assist traffic managers in formulating more effective measures to mitigate traffic violations, tackle traffic accidents at their source, enhance urban traffic safety, and promote the long-term sustainability of urban transportation systems.

1. Introduction

In 2023, China recorded 10 million road traffic accidents [1], with urban roadways accounting for the highest proportion. Driver violations emerged as the leading cause of these incidents. Common violations include speeding, drunk driving, running red lights, failure to yield as required, and illegal parking [2,3,4,5,6]. Drivers are the main subjects of traffic violations, and implementing specific punitive measures can effectively deter such unlawful actions [7]. Thus, identifying the traffic environments where specific types of violations are prone to trigger accidents is essential for guiding traffic management in developing targeted control strategies. This approach enhances urban road safety and supports the sustainable development of urban transportation systems.

Association rule mining (ARM) is one of the key research areas in the field of data mining [8,9,10]. ARM allows the discovery of frequently occurring itemsets within large datasets, thus revealing the underlying relationships among data items. For large and complex road traffic accident datasets, the application of suitable algorithms, such as Apriori and FP-growth [11,12], enables the identification of frequently occurring combinations of traffic accidents and violations. Through this process, valuable association rules can be discovered, shedding light on the relationship between traffic violations and accidents. The application of association rule mining to traffic accident data, aimed at analyzing accident causes and contributing factors, has emerged as a key research focus in the field of traffic safety. For instance, Kong et al. [13] employed a classification-based association rule mining algorithm to analyze the relationship between speeding violations and traffic accidents on highways in Michigan, USA. The study revealed that when speeds exceed 140 km/h, the likelihood of accidents is highest, with the most severe consequences. Similarly, Mercedes et al. [14] applied an enhanced FP-growth algorithm to analyze the impact of various traffic violations on accident severity in Spain between 2003 and 2005, and further estimated the financial costs of these violations. Mohamed et al. [15] analyzed multiple datasets from the Abu Dhabi Traffic Police, applying an enhanced FP-growth algorithm to examine the relationship between drivers’ traffic violation records and accident risk. The findings revealed that running red lights and speeding are the most significant risk factors among traffic violations contributing to accidents. Roni et al. [16] conducted an association rule analysis to investigate the effect of traffic violations on the occurrence of subsequent accidents, revealing that red light violations significantly elevate the risk of accidents on adjacent roads. Jiang et al. [17] employed a multiple linear regression model to examine the relationship between traffic violations and accidents at signal-controlled intersections. The findings revealed that violations committed during the operation of commercial vehicles, wrong-way driving, speeding, and red light violations are the key factors contributing to accident occurrence. As discussed earlier, substantial progress has been achieved in exploring the relationship between traffic violations and accidents through association rule mining. However, existing data for analyzing the correlation between traffic violations and accidents are largely derived from traffic accident reports. For example, Zhang et al. [18] analyzed the traffic accident data report from Guangdong Province (2006–2010) to examine the influence of various factors (such as driver characteristics, vehicle conditions, and road environment) on the occurrence and severity of traffic accidents. The study found that traffic violations and driver characteristics significantly affect the severity of accidents. Since traffic accident reports are subjectively compiled by traffic officers despite being based on objective circumstances, differences in experience, expertise, and judgment standards among officers result in variations in assessing the accident scene and assigning responsibility. Consequently, this introduces a level of subjectivity into the data [19]. Moreover, research has demonstrated that traffic officers frequently overlook minor accidents, which leads to a substantial lack of such data in accident reports [20,21]. When performing correlation analyses between traffic accidents and violations, this missing data can result in biases in identifying potential association rules. Additionally, from a data structure perspective, traffic violations and accidents are correlated across both temporal and spatial dimensions. Within specific time and spatial constraints, traffic violations have the potential to escalate into traffic accidents [22,23]. For instance, after a driver runs a red light, the vehicle may be involved in a side-impact collision at the next intersection [24]. Similarly, when an accident causes traffic congestion, illegal maneuvers such as improper lane changes or wrong-way driving may occur, potentially leading to secondary accidents [25]. From a data association perspective, there is a clear correlation between traffic violations and traffic accidents. Violations such as speeding [26], driving under the influence (DUI) [27], running red lights [28], and illegal parking [29] increase both the likelihood of traffic accidents and the severity of their outcomes.

To address the aforementioned issues, this study proposed the integration of traffic violation data automatically captured by electronic enforcement systems with traffic incident data. The electronic enforcement system offers 24 h uninterrupted surveillance, enabling the comprehensive collection of all traffic violation data within its monitoring range. As electronic enforcement systems have been widely implemented in urban traffic management across China, leveraging this data to explore the intrinsic relationship between traffic violations and accidents has gradually emerged as a significant research direction in the field of traffic safety [30,31,32]. Traffic incident data are recorded by the traffic command center via traffic accident emergency calls, offering a more comprehensive dataset. Compared to traditional traffic accident reports, integrating the automatically captured traffic violation data from electronic enforcement systems with traffic incident data allows for more thorough coverage of different types of urban road traffic accidents. Therefore, this study examined both temporal and spatial dimensions to determine whether traffic violations (or accidents) occur before or after a traffic accident (or violation) either ahead or behind on the road. By integrating additional data such as weather conditions, traffic congestion, time of occurrence, and land use, an association rule model was developed to explore the intrinsic relationships and potential patterns between traffic accidents and violations. The findings offer traffic management authorities more objective and valuable data insights, aiding in the formulation of management strategies that are aligned with local traffic characteristics.

2. Analytical Methods

2.1. Data Preprocessing

The dataset used in this study was derived from traffic accident and violation records in Beijing’s central urban area from 2019 to 2023, comprising 3264 traffic accidents and 147,876 traffic violations. Traffic accident data were extracted from traffic incident reports and contained information such as the time of the accident report, the location of the accident, and the type of accident. Traffic violation data were obtained from urban electronic enforcement systems, capturing details such as the time of the violation, location (camera position), type of violation, and violation photographs. To ensure data accuracy and alignment with the actual traffic environment, we manually verified the geographic locations of traffic accidents and violations generated by the system. First, the address information in the data were converted into geographic coordinates, followed by manual verification to ensure that the locations of accidents and violations displayed on the map match the actual traffic environment. If inconsistencies in geographic coordinates were identified, we reprocessed these coordinates to ensure their spatial accuracy and consistency. The traffic accident and violation data were each standardized, containing the time of occurrence, location, and the type of accident or violation. Examples of standardized traffic accident and violation data are presented in Table 1.

2.2. The Method for Linking Traffic Accidents and Violations Based on Temporal and Spatial Dimensions

When the location and time of a traffic accident and violation nearly coincide, they can be classified as correlated data by comparing the types of the accident and violation. For instance, a traffic accident triggered by a vehicle running a red light at the same intersection. However, there are instances where traffic accidents and violations differ in time and space yet exhibit an intrinsic correlation. For instance, if a vehicle illegally parks in a bike lane, forcing bicycles to use the motor vehicle lane, and an accident between a motor vehicle and a bicycle occurs, the accident typically does not happen at the illegal parking location but rather some distance away. Therefore, this study aimed to link traffic accident and violation data within temporal and spatial constraints. Let the latitude and longitude of the traffic accident location be x_a (m_i,n_i) and the latitude and longitude of the traffic violation location be x_v (e_i,f_i), where i = 1, 2, 3,…, n. The spatial distance between the two points is calculated as [33,34]:

(1)$d=2r\arcsin \alpha$

where r represents the Earth’s radius, expressed in meters, and

(2)$\alpha =\sqrt{{\sin }^2{\displaystyle \frac{n_i-f_i}{2}}+\cos n_i\cos f_i{\sin }^2{\displaystyle \frac{m_i-e_i}{2}}}$

When the shortest distance between traffic accident data and traffic violation data is below a given threshold, the two data points are considered spatially related. If a traffic accident location is associated with multiple traffic violation locations, the violation data with the shortest distance is chosen.

Let the time of the traffic accident be denoted as t_ai and the time of the traffic violation as t_vi, i = 1, 2, 3,…, n. The time difference between the two points is then calculated as:

(3)$t=\left|t_{ai}-t_{vi}\right|$

When the relative time between traffic accident data and traffic violation data falls below a given threshold, the two data points are considered temporally related. The method for linking traffic accident and violation data under spatiotemporal constraints requires that both the shortest distance threshold and the minimum time threshold be met simultaneously.

2.3. Association Rule Mining Method Based on Frequent Pattern Growth Association Rules

The FP-growth algorithm employs a frequent pattern tree (FP-tree) to compress frequent itemsets, dividing the compressed database into a set of conditional datasets and then mining association rules from each of these datasets. The algorithm traverses all frequent itemsets directly on the FP-tree without the need to generate candidate itemsets, which can significantly reduce the database’s storage space [35,36,37]. Therefore, this study employed the FP-growth algorithm to analyze the association rules between traffic accident and violation data.

Association rules are a commonly used concept in data mining, employed to discover relationships and patterns among items in a dataset. In association rules, support, confidence, and lift are three important metrics used to evaluate the validity and strength of the associations.

Support refers to the probability of transactions in the database N containing the itemset {X,Y}, that is, the probability of both X and Y appearing together in all itemsets, which is calculated as:

(4)$Support=\left(X\to Y\right)=P\left(X,Y\right)={\displaystyle \frac{\left|X{\displaystyle \cup Y}\right|}{\left|N\right|}}$

where Support represents the support value, which is the number of times X and Y appear together, and N is the total number of events.

Confidence refers to the probability of Y occurring given that the antecedent X has occurred, which is calculated as:

(5)$Confidence=\left(X\to Y\right)=P\left(Y\left|X\right.\right)={\displaystyle \frac{Support\left(X\to Y\right)}{Support\left(X\right)}}$

Lift measures the degree of association between the antecedent and consequent in an association rule. A lift value > 1 indicates a positive correlation, with higher values signifying stronger positive correlation. A lift value = 1 indicates no correlation, meaning the items are independent of each other, while a lift value < 1 suggests a weak correlation between the two.

(6)$Lift\left(X\to Y\right)={\displaystyle \frac{Confidence\left(X\to Y\right)}{Support\left(Y\right)}}$

The steps of the FP-growth algorithm are as follows: Step 1: Calculate the frequency of each item, sort them in descending order, and construct the FP-tree for the transaction based on the frequent item sets. Step 2: Calculate the minimum support, eliminate items with support below the threshold, and construct the header table for the remaining items. Step 3: Traverse the FP-tree from bottom to top based on the header table to generate the conditional pattern base. Step 4: Based on the conditional pattern base, check whether the conditional FP-tree is empty or contains only a single path. If it does, proceed to Step 5; otherwise, return to Step 3. Step 5: Combine the nodes in the path and generate the final frequent patterns by combining them with the suffix of frequent items.

2.4. Data Filtering Method

The association method based on time and space was used to analyze the standardized traffic accident and violation data. This method required that the shortest distance threshold and the minimum time threshold be satisfied simultaneously to establish a valid association between traffic accidents and violations. For the minimum thresholds based on time and space constraints, the shortest spatial thresholds of 50 m, 100 m, 150 m, 300 m, and 450 m, along with the minimum time thresholds of 5 min, 10 min, 15 min, 30 min, and 45 min, were selected to link traffic accident and violation data, as shown in Table 2. As indicated in Table 2, larger time and space thresholds lead to a greater amount of associated data. However, once the thresholds exceed a certain limit, the frequency of duplicate associations between traffic accidents and violations also increases. Based on previous research conclusions, the optimal time window for traffic violations is around 30 min [22]. When the time threshold is set to 30 min, the overlap in time and space results in multiple violations being linked to the same accident. The larger the minimum distance threshold, the more instances of duplicate associations arise between traffic violations and accidents, as indicated in Table 3. When the distance threshold exceeds 300 m, the expansion of the spatial range leads to more traffic accidents and violations in the same or adjacent areas being treated as associated events by the system, a phenomenon that is particularly pronounced in high-risk locations. Additionally, as the threshold increases, events that were originally unrelated may be incorrectly treated as associated due to the relaxed distance constraints, leading to a substantial rise in duplicate data.

Therefore, this study set the minimum time threshold at 30 min and the minimum distance threshold at 300 m, yielding 2126 traffic violation and accident association records. The complete data processing procedure is illustrated in Figure 1.

3. Results

3.1. Results of the Association Data Distribution Based on Spatiotemporal Constraints

The distribution trend graphs for the 2126 key data points under spatial and temporal constraints were plotted separately. According to Figure 2a, the associated data under time constraints is primarily concentrated between 7:30 a.m. and 10:30 p.m., with noticeable morning (8:00–9:00) and evening (5:00–6:00 p.m.) peaks. According to Figure 2b, under spatial constraints, the distance differences of the associated data show an almost uniform increasing trend. To further analyze the cumulative distribution and probability density of the associated data under time and space constraints, cumulative distribution curves and probability density curves were plotted separately for the data under spatial and temporal constraints, as illustrated in Figure 3. According to Figure 3a, the associated data under time constraints is evenly distributed, and the probability of occurrence is higher for data with a time difference of 4 to 5 min. This also highlights that during peak hours or in traffic hotspots, factors such as signal timing, road conditions, traffic flow variability, and driver behavior patterns collectively contribute to the concentration, periodicity, and chain reactions of traffic violations and accidents. According to Figure 3b, when the distance difference is 150 m, the cumulative value of the associated data approaches 0.5, and the curve shows an approximately uniform distribution. The probability density curve reveals that the occurrence probability is higher when the distance difference is 150 m and 225 m. This could be due to the fact that when traffic violations or accidents occur, the spacing between vehicles typically falls within 150 m. This distance might represent a typical vehicle gap, reflecting the common distribution of vehicle distances in traffic flow.

3.2. Results of Association Rules

The 2126 associated data points were input into the FP-growth algorithm for association rule mining of traffic accident and violation data. Each itemset contains traffic accident types, violation types, the time of occurrence (temporal attributes), surrounding land use (spatial attributes), weather conditions, and congestion levels. The encoding and support for each item type are presented in Table 4.

To further illustrate the fundamental relationship between traffic data, accident types, and violations, a network diagram depicting the relationship between accident and violation types is shown in Figure 4. The numbers in the figure represent the count of each type, while the arrows denote the causal relationship between the violation type (denoted as b) and the accident type (denoted as a). As shown in Figure 4, motor vehicle accidents (a1) occur most frequently, with a total of 1256 incidents, indicating that it is the most common accident type. Motor vehicle and non-motor vehicle accidents (a2) are significantly fewer than a1, totaling only 398 incidents. However, according to the correlation with violations, violations were responsible for 374 of these accidents, resulting in a violation-induced accident rate of 94%. The other accident types occur less frequently. From the perspective of violation types, stopping at or within the intersection when encountering a stop signal (b1) occurs most frequently, with the highest associated accident rate of 80.8%. The remaining three types of violations occur at similar frequencies, with the associated accident rates being relatively low, at 57.7%, 34.4%, and 32.4%, respectively.

Setting the minimum support to 0.002, the associated data were inputted into the FP-growth algorithm to mine all frequent itemsets. To focus on analyzing the association rules between traffic violations and accidents, frequent itemsets that contain both traffic violations and accidents were considered in the association rule expressions. A confidence threshold of 0.4 was applied, and the filtered association rule expressions are presented in Table 5. The choice of a 0.4 threshold was aimed at filtering out events with low frequency and weak correlation, allowing for a focus on more significant associative phenomena. Setting the threshold at 0.4 is aimed at filtering out low-frequency, weakly correlated events, allowing a focus on more significant associations. It ensured that the selected rules had practical value in traffic management and prevention, preventing both excessive filtering that might have missed key associations and misleading results caused by overly low confidence levels.

In total, there are 18 strong association rules between traffic accidents and violations. Most of these rules involve motor vehicle-to-motor vehicle (a1) accidents, but only seven of them are classified as strong association rules. There are six association rules for motor vehicle-to-non-motor vehicle (a2) accidents, and all of them are strong association rules. There are three association rules for motor vehicle-to-pedestrian (a3) accidents, and all of them are strong association rules. There is only one association rule for single-vehicle accidents (a4) and other traffic accidents (a7), which is also a strong association rule, as shown in Table 6. As the electronic enforcement system currently focuses primarily on motor vehicle violations and does not yet include data on violations by non-motor vehicles and pedestrians, it is not possible to derive association rules for non-motor vehicle and pedestrian accidents.

4. Discussion

The data selected spans the period from 2019 to 2023, encompassing the COVID-19 pandemic (primarily from 2020 to 2022 in Beijing), during which global traffic patterns underwent significant changes. According to Inada’s study [38], the lockdown measures implemented during the COVID-19 pandemic led to a dramatic reduction in traffic flow, accompanied by notable shifts in driver behavior, such as an increase in speeding. To examine whether the pandemic influenced the association rules between traffic accidents and violations, we further compared the average annual traffic data from the pandemic period (2020–2022) with that from the non-pandemic period (2019 and 2023), as presented in Table 7. The results indicate that the frequency of traffic violations during the pandemic was lower than that in the non-pandemic period, with the probability of accidents caused by violations being under 10%. The likely explanation for this phenomenon was that, due to the COVID-19 control measures, road traffic and pedestrian flow were significantly reduced, leading to a substantial decrease in accidents caused by violations. As a result, stable association rules could not be established during the pandemic. Although the data used in this study spanned both the pandemic and non-pandemic periods, the association rules generated were almost exclusively derived from the non-pandemic period, with the pandemic data exerting minimal impact on the associations.

The association rules derived from the above data results are as follows:

(1) Association rules for motor vehicle-to-motor vehicle traffic accidents and violations:

To more effectively analyze the impact of different violation types on motor vehicle and non-motor vehicle accidents under specific conditions, the same dependent variable is controlled, and various correlation tables (Table 8) are generated. As shown in Table 8, violations such as “illegal parking” (b2), “crossing the red light” (b1), and “violating the prohibition” (b4) all contribute to motor vehicle and non-motor vehicle traffic accidents. Furthermore, under specific conditions (“rain” (d1) and “slow traffic” (c1)), the correlation between violations and traffic accidents is strong, with lift values exceeding 1. According to Mostafa’s study, in complex urban environments with high traffic density, such as approaching intersections, drivers may exhibit more aggressive driving behaviors [39], making them more prone to violations such as running red lights and crossing the stop line. Running red lights disrupts the priority traffic flow at intersections, while crossing the stop line occupies critical positions, increasing the potential for conflict points as vehicles pass through. Moreover, rain impairs visibility and extends braking distances, while in slow traffic conditions, the reduced following distance and limited maneuverability significantly heighten the risk associated with these violations, ultimately creating a strong correlation and posing a serious threat to traffic safety. Under the conditions of “rainy days” (d1) and “Morning Peak” (e1), the occurrence of the “running red lights” violation (b3) leads to motor vehicle collisions, with a 100% occurrence probability and a lift value of 1.700, demonstrating a strong correlation between the two. This may be due to the fact that the morning rush hour is a peak commuting period, characterized by heavy traffic flow, numerous conflict points at intersections, and a significant reduction in spatial and temporal decision-making windows. Under time pressure and driven by impatience, drivers are more likely to engage in risky behavior, which aligns with Bucchi and Mostafa’s study [39,40]. Wet and slippery road surfaces during rainy weather significantly extend vehicle braking distances and reduce handling stability, while obstructed visibility and blurred signage further impair drivers’ ability to judge traffic signals and intersection dynamics. Additionally, the mixture of high- and low-speed traffic, combined with the uncertainty of non-motorized vehicles and pedestrians’ movements during rainy weather, intensifies the complexity of intersections, significantly increasing the risk of collisions between red light running vehicles and vehicles traveling normally. These underlying factors are compounded, resulting in a 100% association and a significant lift value of 1.700, indicating that red light running behavior poses a particularly severe threat to traffic safety under such conditions. After removing the “Morning Peak” factor, the confidence level of the association drops to 71.4%, with a lift value of 1.213, still indicating a strong correlation between them. The possible reason is that during the morning rush hour, dense traffic and frequent intersection conflicts lead to a higher absolute frequency of accidents caused by red light running. After removing the morning peak factor, the data distribution extends to non-peak hours, with reduced traffic volume and fewer intersection conflicts, which dilutes the absolute strength of the association, thus lowering the confidence level. However, the persistent impact of wet and slippery road surfaces and obstructed visibility during rainy days results in a significant relative risk for red light running behavior at any time of day. At the same time, running red lights, as a high-risk behavior, retains its intrinsic association with accidents, which does not diminish with changes in time periods. Therefore, although the confidence level has decreased, the lift value still indicates that the impact of red light running behavior on traffic accidents remains significant and consistent. Similarly, high traffic density and short vehicle spacing in congested environments significantly increase the probability of spatial conflicts between vehicles. Red light running behavior further disrupts the already strained traffic order, forcing vehicles traveling normally to take measures such as emergency braking or evasive actions, significantly increasing the risk of collisions. Additionally, drivers’ impatience and distracted attention in congested environments increase the occurrence rate of red light running behavior [41], while the dynamic complexity and speed differences in such environments amplify the severity of accidents. Ultimately, red light running behavior in congested environments has a persistent and significant impact on traffic accidents. Therefore, under the “road congestion” (c2) condition, the “running red lights” violation (b3) is also strongly associated with motor vehicle collisions, with a lift value of 1.018. Similar to the “Morning Peak” (e1), the “Evening Peak” (e2) period further amplifies the association between the two (with a lift value of 1.072). In addition, violations such as “illegal parking” (b2), “stopping beyond the stop line” (b1), and “violating prohibitions” (b4) all contribute to motor vehicle collisions, with a strong association between violations and accidents in certain environments (such as “rainy weather” and “slow traffic”). This is because illegally parked vehicles occupy road space and obstruct visibility, forcing normal vehicles to maneuver around them, thereby increasing the risk of collisions. Stopping beyond the stop line disrupts the priority order of traffic signals, occupying critical positions at intersections and increasing conflict points when vehicles pass through. Violating prohibitions (such as illegal left turns or U-turns) alters traffic flow rules, further amplifying the disruptive effects of sudden dynamics in slow-moving traffic environments. Additionally, rain obstructs visibility and increases braking distances, while in slow-moving traffic, short following distances and limited maneuvering space significantly elevate the danger of these violations, ultimately forming a strong association and posing a serious threat to traffic safety.

(2) Association rules for motor vehicle-to-non-motor vehicle traffic accidents and violations:

In environments such as “morning rush hour” (e1), “slow traffic” (c1), “congestion” (c2), and “education zones” (f4), “illegal parking” (b2) exhibits a strong correlation with both motor vehicle and non-motor vehicle accidents, with lift values exceeding 2. Based on the actual road safety conditions, it can be observed that illegally parked vehicles occupying bicycle lanes may cause collisions between non-motorized vehicles and motor vehicles. Additionally, non-motorized vehicles may be forced to “borrow” motor vehicle lanes, leading to accidents with vehicles traveling normally. It is noteworthy that in the “Educational Area”, accidents involving “motor vehicles and non-motorized vehicles” caused by “illegal parking” primarily occur during school starting and dismissal times at primary and secondary schools and kindergartens. During these periods, a large number of motor vehicles and non-motorized vehicles, which are used for picking up and dropping off students, converge in the area, leading to frequent traffic accidents and posing significant safety risks. Additionally, during the “evening rush hour”, the violation of “running red lights” and “motor vehicle and non-motor vehicle” traffic accidents exhibit a strong correlation, with a lift value of 2.314. During the “Evening Peak” period, the demand for both motor vehicles and non-motorized vehicles reaches its peak, leading to dense traffic flow at intersections and increased conflict points, making red light running behavior more likely to trigger accidents. Motor vehicles running red lights typically pass through intersections at higher speeds, while non-motorized vehicles (such as electric bicycles), due to their higher maneuverability, often attempt to cross intersections without signal protection. The low-illumination environment during the “Evening Peak” period further reduces motor vehicle drivers’ ability to perceive the dynamics of non-motorized vehicles, preventing timely evasive actions and thereby increasing the risk of accidents. Additionally, during the “Evening Peak” period, drivers often experience reduced reaction times due to fatigue, while some non-motorized vehicle drivers may engage in aggressive behavior, such as sudden starts or bypassing normal traffic paths, to save time. This significantly increases the likelihood of dynamic conflicts with motor vehicles running red lights. In such cases, red light running behavior not only directly disrupts the priority rules of traffic signals but also intensifies conflicts between motor vehicles and non-motorized vehicles in the dynamic traffic flow, reflecting the amplifying effect of the complex “Evening Peak” environment on accident associations. Strengthening traffic signal management and law enforcement against red light running during the “Evening Peak” period, as well as regulating non-motorized vehicle behavior, are crucial measures to reduce such accidents. In the “Residential Area”, vehicles “stopping beyond the stop line” can lead to motor vehicle and non-motorized vehicle collisions, with a lift value of 2.217. In the actual traffic environment in China, non-motorized vehicle drivers (mainly electric bicycle riders) often start before the pedestrian green light phase begins, while the traffic signal for motor vehicles is still in the yellow phase. Motor vehicles stopping beyond the stop line, due to their proximity to the intersection core area or encroachment onto non-motorized vehicle lanes, further narrow the already limited traffic space. Additionally, because the yellow phase is short, drivers often accelerate to pass through the intersection before the signal changes. As non-motorized vehicles start early, their paths overlap with those of the fast-moving motor vehicles, increasing the risk of collisions. Additionally, the complex structure of traffic participants in residential areas (such as the mixing of pedestrians and non-motorized vehicles) and the low level of adherence to traffic rules further amplify this association. Non-motorized vehicle drivers typically ignore traffic signal rules, and red light running behavior exacerbates this disruption, making dynamic conflicts between motor vehicles and non-motorized vehicles harder to avoid, ultimately significantly increasing the risk of accidents. Strengthening signal management in residential areas, regulating motor vehicle parking line positions, and enhancing non-motorized vehicle awareness are key measures to reduce such accidents.

(3) Association rules for motor vehicle-to-pedestrian traffic accidents and violations:

In “motor vehicle and pedestrian” traffic accidents, during the “Evening Peak” (e2) period, the high pedestrian traffic at intersections significantly increases the safety risks to pedestrians caused by “stopping beyond the stop line” violations (b1), with a lift value of 4.576. In China, pedestrians heavily rely on traffic signals and often perceive the green light as providing absolute safety. This overconfidence makes pedestrians more likely to overlook the threat posed by vehicles stopping beyond the stop line during red lights, further increasing the probability of accidents [42]. Therefore, strengthening traffic signal management at intersections during the “Evening Peak” period, strictly enforcing laws to limit red light running behavior, and enhancing both drivers’ and pedestrians’ awareness of traffic safety are crucial measures to reduce motor vehicle and pedestrian accidents. Due to the relatively complex traffic environment in the “Residential Area” (f1), violations of prohibitions near residential areas (such as driving in the wrong direction or entering restricted zones) (b4) pose a threat to pedestrians’ normal passage, ultimately leading to traffic accidents, with a lift value of 4.524. This is because residential areas have a high number of pedestrians, including vulnerable groups such as the elderly and children, whose limited reaction abilities make it difficult for them to cope with sudden vehicle maneuvers. Furthermore, many old neighborhoods in Beijing lack adequate pedestrian–vehicle separation designs, and sidewalks are often occupied by illegally parked vehicles or street vendors, forcing pedestrians to walk alongside motor vehicles, which significantly heightens the threat posed by vehicles traveling in the wrong direction or violating traffic rules. Additionally, drivers often act on a sense of overconfidence or to save time, choosing to drive against traffic or make illegal U-turns near residential areas. These behaviors disrupt the fundamental order of traffic rules, and especially in situations with narrow roads and obstructed visibility, they increase the likelihood of dynamic conflicts between pedestrians and vehicles. At the same time, traffic management around residential areas is relatively lax, with a lack of monitoring and law enforcement coverage. The low cost of violations further encourages the occurrence of illegal behaviors. The combination of weak awareness of traffic rules and insufficient regulation makes the traffic environment in residential areas more complex and poses higher risks. Therefore, addressing this issue requires optimizing road design in residential areas, strengthening electronic surveillance and law enforcement, regulating driver behavior, restoring the functionality of sidewalks, and enhancing residents’ traffic safety awareness through community collaboration to effectively reduce pedestrian traffic accidents caused by vehicle violations of traffic prohibitions. Similar to motor vehicle and non-motorized vehicle collisions, “illegal parking” (b2) violations can also occupy pedestrian pathways, leading to traffic accidents between pedestrians and motor vehicles, with a lift value of 4.203.

(4) Association rules for single-vehicle accidents and traffic violations:

Single-vehicle accidents and the violation of “illegal parking” (b2) primarily occur due to illegally parked vehicles occupying regular road space, which leads to damage to the vehicles during parking, with a lift value of 9.095. Illegally parked vehicles are often placed in irregular areas, such as narrow roads, blind spots, near curves, or non-motorized vehicle lanes. These locations not only obstruct the normal flow of traffic but also create traffic bottlenecks, forcing passing vehicles to take emergency evasive actions or engage in illegal maneuvers (such as crossing the line or emergency braking), leading to contact with the illegally parked vehicles. Additionally, when illegally parked vehicles are located at road corners or areas with obstructed visibility, other drivers are more likely to experience scrapes or collisions due to poor sightlines, as they may fail to notice the vehicles in time. Illegal parking may also trigger a chain reaction. For example, when non-motorized vehicle lanes are occupied, non-motorized vehicles may enter motor vehicle lanes, causing normal vehicles to change lanes to avoid collisions. This further intensifies dynamic conflicts on the road, ultimately resulting in damage to the illegally parked vehicle. In summary, illegal parking disrupts normal road resources and existing traffic order, not only affecting traffic flow but also making vehicles more vulnerable to becoming accident victims during the parking period. Addressing this issue requires enhancing the regulation and enforcement of illegal parking, properly planning parking areas, and optimizing road design to reduce the threat posed by illegally parked vehicles to traffic safety.

(5) Association rules for other types of accidents and traffic violations:

“Commercial areas” (f2) are typically high-incidence zones for “illegal parking” (b2), exhibiting a lift value of 19.536. Illegally parked vehicles may block traffic and occupy fire lanes, thus disrupting the normal flow of traffic. Due to the concentration of shopping, dining, and office activities in the “Commercial Area”, there is a high demand for short-term motor vehicle stops. However, the area often lacks sufficient parking facilities or designated temporary parking zones, leading to some drivers parking illegally. Illegally parked vehicles typically occupy main roads, non-motorized vehicle lanes, or roadside areas, directly reducing the available passage space and creating traffic bottlenecks. This situation, particularly during peak hours, further exacerbates road congestion. Additionally, illegal parking in the “Commercial Area” often occupies emergency lanes, such as fire lanes and emergency lanes, which not only disrupts traffic flow but also threatens the emergency response capability in case of incidents. The presence of illegally parked vehicles forces other vehicles to detour or frequently change lanes, increasing the complexity of traffic flow and the potential risk of accidents. Due to the mixed traffic environment in the “Commercial Area”, conflicts between the activity zones of non-motorized vehicles, pedestrians, and illegally parked vehicles may lead to more dynamic disruptions. For example, non-motorized vehicles may be forced to use motor vehicle lanes, or pedestrians may be compelled to walk in traffic lanes, further increasing safety risks. In summary, the issue of illegal parking in the “Commercial Area” stems from a lack of parking resources and inadequate management. This not only obstructs traffic flow but also poses a threat to fire safety and overall traffic order. Addressing this issue requires increasing the supply of parking facilities in the Commercial Area, strictly enforcing laws to remove illegally parked vehicles, optimizing road design, and enhancing the role of public transportation in guiding traffic. These measures will help alleviate traffic pressure in the Commercial Area and improve safety levels.

In summary, the FP-growth algorithm can generate various association rules between traffic violations and traffic accidents. Among these, some association rules align with the actual traffic conditions in China [43,44,45]. For example, “running red lights” violations can lead to traffic accidents (between motor vehicles and motor vehicles, or between motor vehicles and non-motorized vehicles). This is especially true when external traffic conditions are adverse (such as during rain or traffic congestion), as these conditions exacerbate the likelihood of the violation resulting in an accident. Furthermore, association rules can uncover potential and often overlooked traffic safety issues. For example, some drivers believe that “illegal parking” only disrupts normal traffic order and has little impact on road safety. While this violation is penalized with a fine in China rather than points deductions, association rules reveal that “illegal parking” is associated with all types of traffic accidents. Notably, the lift values for motor vehicle and non-motorized vehicle accidents exceed 2, indicating that although the safety risk directly caused by the illegally parked vehicle may be low, the act of “illegal parking” significantly increases the surrounding traffic safety risks. Traffic management personnel should pay special attention to “illegal parking” behaviors occurring during the “Morning Peak”, on “Congestion” segments, and in the “Educational Area.” These issues directly impact the goals of sustainable urban transportation development, especially in the context of the increasingly prominent conflict between rising travel demand and limited resource supply. To achieve sustainable development of the transportation system, it is necessary to improve road design, optimize parking management, strengthen law enforcement against violations, and enhance the awareness of traffic rules among participants. These measures will reduce the frequency of traffic violations and their impact on safety and efficiency. Additionally, increasing investment in public transportation and non-motorized vehicle-friendly infrastructure, while promoting green travel options, will help reduce the environmental burden of traffic and ensure the coordinated development of urban transportation that is safe, convenient, and environmentally friendly.

5. Conclusions

This study primarily focused on the association rules between traffic accidents and traffic violations. A dataset consisting of 3264 traffic accident records and 147,876 traffic violation records from urban Beijing (2019–2023) was selected for analysis. Using a data association method constrained by both temporal and spatial dimensions, 2126 violations linked to traffic accidents were identified. The FP-growth algorithm was applied to derive 18 strong association rules, encompassing 5 categories of traffic accidents and 4 categories of traffic violations. The spatial distribution of the associated data is uniform, with the occurrence of traffic accident and violation data concentrated between 7:30 a.m. and 10:30 p.m. daily. There are distinct morning (8:00–9:00) and evening (5:00–6:00 p.m.) peak periods. By incorporating the temporal, spatial, traffic, and weather attributes of the data, the FP-growth algorithm was used to derive association rules for 5 types of traffic accidents and 4 types of traffic violations. These rules not only include common safety issues that are likely to lead to accidents with casualties, such as running red lights, which may cause accidents at the next intersection, and speeding, which increases the severity of accidents. They also uncover potential and often overlooked traffic safety issues, such as illegal parking, which can trigger various types of traffic accidents and increase the accident risk in the surrounding environment. A reasonable analysis of the causes has been provided, based on the actual conditions in Beijing. Using a focused approach, the results of the association rules can assist urban traffic managers in formulating more effective strategies to reduce related traffic violations. By strengthening the supervision of violations, optimizing road design, improving parking management, and enhancing both drivers’ and pedestrians’ awareness of traffic rules, the frequency of traffic accidents can be effectively reduced. Building an urban transportation system centered on safety, convenience, and environmental sustainability not only alleviates traffic pressure and reduces traffic accidents but also decreases the burden of transportation on resources and the environment, thereby achieving the coordinated and sustainable development of society and the transportation system. In the future, a transportation system centered on safety, convenience, and environmental sustainability should be established. By integrating more comprehensive data resources and introducing multidimensional analytical methods, the depth and practical value of traffic safety research can be further enhanced. This will not only reduce the occurrence of traffic accidents at their source and improve the safety level of urban transportation but also decrease the burden of transportation on resources and the environment, thereby achieving the coordinated and sustainable development of society and the transportation system.

However, this study has certain limitations. On the one hand, the mining of association rules relies heavily on the quality of the data itself. Although this paper proposes a data association method based on temporal and spatial constraints, the shortest distance and minimum time thresholds still require further calibration with a larger dataset. On the other hand, the traffic violation data used in this study lacks data on non-motorized vehicles and pedestrians, which prevents the mining of association rules for traffic accidents involving non-motorized vehicles and pedestrians. Future research will consider incorporating more comprehensive traffic violation data to further explore the association rules between traffic accidents and violations.

Author Contributions

Conceptualization, H.W. and G.L.; Methodology, H.W.; Validation, H.W.; Formal analysis, H.W.; Resources, G.L.; Data curation, H.W.; Writing—original draft, H.W.; Writing—review & editing, G.L.; Supervision, G.L.; Funding acquisition, G.L. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data presented in this study are available on request from the corresponding author.

Conflicts of Interest

The authors declare no conflict of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1. Data processing and analysis framework.

Figure 2. Distribution of traffic violation and accident association data based on time and space dimensions.

Figure 3. Cumulative distribution curve and probability density curve of traffic accident and violation association data.

Figure 4. Relationship diagram between accident types and violation types.

Table 1

Examples of standardized traffic accident information and traffic violation information.

Accident Time	Accident Location (Coordinates)	Accident Type	Brief Description	Violation Time	Violation Location (Coordinates)	Violation Type
2022.1.3 08:43	No. 12, Chengshousi Middle Road, Fengtai District, Beijing (116.4, 39.8)	Vehicle-to-Vehicle	Collision between a Mercedes-Benz and a Chevrolet sedan	2022.1.3 08:43	No. 12, Chengshousi Middle Road, Fengtai District, Beijing (116.4, 39.8)	Violation of traffic prohibition signs
2022.1.3 11:23	No. 6–10, Beixiaojie, Dongcheng District, Beijing (116.7, 40.1)	Vehicle-to-Non-Motorized Vehicle	Collision between a Toyota sedan and a bicycle	2022.1.3 10:43	No. 6–10, Beixiaojie, Dongcheng District, Beijing (116.7, 40.1)	Illegal parking violation

Table 2

Traffic accident and violation association data under different spatiotemporal thresholds.

Distance Threshold (m)Associations NumberTime Threshold (min)	5	10	15	30	45
50	62	120	179	255	312
100	123	239	369	489	596
150	206	393	639	1107	1457
300	453	845	1335	2126	2604
450	563	1056	1523	2691	3147

Table 3

Duplicate association data between traffic accidents and violations at different distance thresholds with a time threshold of 30 min.

Distance Threshold (m)	50	100	150	300	450
Number of Duplicate Associations	323	630	2079	4635	15,357

Table 4

Item coding and support values for the FP-growth algorithm.

Data Source	Item Code	Item Type	Support
Accident Type	a1	Vehicle-to-Vehicle	0.591
	a2	Vehicle-to-Non-Motorized Vehicle	0.187
	a3	Vehicle-to-Pedestrian	0.103
	a4	Single-Vehicle Accident	0.045
	a5	Non-Motorized Vehicle-to-Non-Motorized Vehicle	0.037
	a6	Non-Motorized Vehicle-to-Pedestrian	0.007
	a7	Other Accidents	0.030
Violation Type	b1	Stopping within the stop line or intersection when encountering a stop signal	0.367
	b2	Illegal Parking by a Motor Vehicle	0.199
	b3	Driving a Motor Vehicle Against Traffic Signals	0.232
	b4	Violation of Prohibition Signs by a Motor Vehicle	0.202
Road Condition	c1	Slow Traffic	0.061
	c2	Congestion	0.111
Time Period	e1	Morning Peak (8:00~9:00)	0.192
	e2	Evening Peak (17:00~18:00)	0.211
Weather	d1	Rainy Day	0.067
Land Use Type	f1	Residential Area	0.269
	f2	Commercial Area	0.246
	f3	Government/Factory	0.106
	f4	Educational Area	0.215
	f5	Medical Area	0.150
	f6	Recreational Area	0.014

Table 5

Association rules between traffic accidents and traffic violations.

Antecedent	Consequent	Support	Confidence	Lift
d1, b3, e1	a1	0.0030	1.000	1.700
d1, b2		0.0042	0.858	1.456
c1, b1		0.0069	0.749	1.272
b3, d1		0.0040	0.715	1.213
b3, c2, e2		0.0106	0.633	1.072
c1, b4		0.0048	0.626	1.062
b3, c2		0.0057	0.598	1.018
b4, c2		0.0112	0.580	0.982
f6, b2		0.0051	0.557	0.943
d1, e1, b1		0.0051	0.557	0.943
f4, b2, e1		0.0081	0.557	0.943
d1, b4		0.0064	0.544	0.923
d1, b1		0.0075	0.537	0.912
b3, e1		0.0164	0.536	0.911
c2, b2, f2		0.0088	0.500	0.850
c2, b2		0.0191	0.500	0.850
b2, e1		0.0281	0.500	0.850
b1, c2, e1		0.0081	0.500	0.850
b1, f2		0.0106	0.500	0.850
b4, e1		0.0240	0.465	0.789
b4, f5		0.0116	0.449	0.763
b1, c2		0.0170	0.449	0.763
f4, b4		0.0153	0.445	0.755
e2, b4		0.0156	0.445	0.754
b3, f3		0.0118	0.430	0.729
c1, b3		0.0083	0.430	0.729
f4, b2		0.0217	0.424	0.718
b1, e1		0.0222	0.422	0.716
b3, f4		0.0182	0.410	0.696
b1, f5		0.0085	0.400	0.680
e2, b1		0.0287	0.400	0.680
b2, e1	a2	0.0030	0.500	2.661
b3, e2		0.0135	0.436	2.314
b2, c1		0.0069	0.415	2.215
b2, c2		0.0030	0.415	2.216
b2, f4		0.0033	0.415	2.216
f1, b1		0.123	0.400	2.217
b1, e2	a3	0.0088	0.468	4.576
f1, b4		0.0075	0.461	4.524
b2		0.0036	0.430	4.203
b2	a4	0.0086	0.400	9.095
b2, f2	a7	0.0130	0.627	19.536

Table 6

Strong association rules between traffic accidents and traffic violations.

Accident Type	Violation Type	Other Factors	Lift
a1	b3	d1, e1	1.700
	b2	d1	1.456
	b1	c1	1.272
	b3	d1	1.213
	b3	c2, e2	1.072
	b4	c1	1.062
	b3	c2	1.018
a2	b2	e1	2.661
	b3	e2	2.314
	b2	c1	2.215
	b2	c2	2.216
	b2	f1	2.216
	b1	f4	2.217
a3	b1	e2	4.576
	b4	f1	4.524
	b2		4.203
a4	b2		9.095
a7	b2	f2	19.536

Table 7

Average annual traffic accident and violation data during the pandemic and non-pandemic periods.

Period	Pandemic Period				Non-Pandemic Period
Traffic Violations	Line-Crossing at a Red Light	Illegal Parking	Running a Red Light	Speeding	Line-Crossing at a Red Light	Illegal Parking	Running a Red Light	Speeding
Number of violations	32	76	35	223	1831	1522	863	938
Number of accidents caused	2	6	3	21	321	423	226	365

Table 8

Analysis of the correlation between different factors and motor vehicle and non-motor vehicle accidents.

Violation Type	Other Factors		Support	Confidence	Lift
b1	c1		0.0069	0.749	1.272
b4			0.0048	0.626	1.062
b2	d1		0.0042	0.858	1.456
b3			0.0040	0.715	1.213
		e1	0.0030	1.000	1.700
	c2		0.0057	0.598	1.018
		e2	0.0106	0.633	1.072

References

1. Zeng, Y.; Qiang, Y.; Zhang, N.; Yang, X.; Zhao, Z.; Wang, X. An Influencing Factors Analysis of Road Traffic Accidents Based on the Analytic Hierarchy Process and the Minimum Discrimination Information Principle. Sustainability; 2024; 16, 6767. [DOI: https://dx.doi.org/10.3390/su16166767]

2. Erenler, A.K.; Gümüş, B. Analysis of road traffic accidents in Turkey between 2013 and 2017. Medicina; 2019; 55, 679. [DOI: https://dx.doi.org/10.3390/medicina55100679] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31600894]

3. Nyamuame, G.Y.; Aglina, M.K.; Akple, M.S.; Philip, A.; Klomegah, W. Analysis of road traffic accidents trend in Ghana: Causing factors and preventive measures. Int. J. Eng. Sci. Manag. Res.; 2015; 2, pp. 127-132.

4. Cai, Q. Cause analysis of traffic accidents on urban roads based on an improved association rule mining algorithm. IEEE Access; 2020; 8, pp. 75607-75615. [DOI: https://dx.doi.org/10.1109/ACCESS.2020.2988288]

5. Kakkar, R.; Aggarwal, P.; Kakkar, M.; Deshpande, K.; Gupta, D. Road traffic accident: Retrospective study. Indian J. Sci. Res.; 2014; 5, pp. 59-62.

6. Janani, G.; Devi, N.R. Road traffic accidents analysis using data mining techniques. JITA-APEIRON; 2017; 14, pp. 84-91. [DOI: https://dx.doi.org/10.7251/JIT1702084J]

7. Liu, Z.; Wu, H.; Li, R. Effects of the penalty mechanism against traffic violations in China: A joint frailty model of recurrent violations and a terminal accident. Accid. Anal. Prev.; 2020; 141, 105547. [DOI: https://dx.doi.org/10.1016/j.aap.2020.105547]

8. Kotsiantis, S.; Kanellopoulos, D. Association rules mining: A recent overview. GESTS Int. Trans. Comput. Sci. Eng.; 2006; 32, pp. 71-82.

9. Agrawal, R.; Mannila, H.; Srikant, R.; Toivonen, H.; Verkamo, A.I. Fast discovery of association rules. Adv. Knowl. Discov. Data Min.; 1996; 12, pp. 307-328.

10. Telikani, A.; Gandomi, A.H.; Shahbahrami, A. A survey of evolutionary computation for association rule mining. Inf. Sci.; 2020; 524, pp. 318-352. [DOI: https://dx.doi.org/10.1016/j.ins.2020.02.073]

11. Ashraf, M.T.; Dey, K.; Mishra, S.; Rahman, M.T. Extracting rules from autonomous-vehicle-involved crashes by applying decision tree and association rule methods. Transp. Res. Rec.; 2021; 2675, pp. 522-533. [DOI: https://dx.doi.org/10.1177/03611981211018461]

12. Yang, Y.; Tian, N.; Wang, Y.; Yuan, Z. A parallel FP-growth mining algorithm with load balancing constraints for traffic crash data. Int. J. Comput. Commun. Control; 2022; 17, 4806. [DOI: https://dx.doi.org/10.15837/ijccc.2022.4.4806]

13. Kong, X.; Das, S.; Jha, K.; Zhang, Y. Understanding speeding behavior from naturalistic driving data: Applying classification based association rule mining. Accid. Anal. Prev.; 2020; 144, 105620. [DOI: https://dx.doi.org/10.1016/j.aap.2020.105620] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32570086]

14. Ayuso, M.; Guillén, M.; Alcañiz, M. The impact of traffic violations on the estimated cost of traffic accidents with victims. Accid. Anal. Prev.; 2010; 42, pp. 709-717. [DOI: https://dx.doi.org/10.1016/j.aap.2009.10.020]

15. Shawky, M.; Al-Badi, Y.; Sahnoon, I.; Al-Harthi, H. The relationship between traffic rule violations and accident involvement records of drivers. Advances in Human Aspects of Transportation: Proceedings of the AHFE 2016 International Conference on Human Factors in Transportation, July 27–31, 2016, Walt Disney World^®, Florida, USA; Springer: Berlin/Heidelberg, Germany, 2017; pp. 745-755.

16. Factor, R. The effect of traffic tickets on road traffic crashes. Accid. Anal. Prev.; 2014; 64, pp. 86-91. [DOI: https://dx.doi.org/10.1016/j.aap.2013.11.010]

17. Jiang, Z.-H.; Yang, X.-G.; Sun, T.; Wang, T.; Yang, Z. Investigating the relationship between traffic violations and crashes at signalized intersections: An empirical study in China. J. Adv. Transp.; 2021; 2021, 4317214. [DOI: https://dx.doi.org/10.1155/2021/4317214]

18. Zhang, G.; Yau, K.K.; Chen, G. Risk factors associated with traffic violations and accident severity in China. Accid. Anal. Prev.; 2013; 59, pp. 18-25. [DOI: https://dx.doi.org/10.1016/j.aap.2013.05.004]

19. Kaufmann, M.; Egbert, S.; Leese, M. Predictive policing and the politics of patterns. Br. J. Criminol.; 2019; 59, pp. 674-692. [DOI: https://dx.doi.org/10.1093/bjc/azy060]

20. Stylianou, K.; Dimitriou, L.; Abdel-Aty, M. Big data and road safety: A comprehensive review. Mobility Patterns, Big Data and Transport Analytics; Elsevier: Amsterdam, The Netherlands, 2019; pp. 297-343.

21. Montella, A.; Andreassen, D.; Tarko, A.P.; Turner, S.; Mauriello, F.; Imbriani, L.L.; Romero, M.A. Crash databases in Australasia, the European Union, and the United States: Review and prospects for improvement. Transp. Res. Rec.; 2013; 2386, pp. 128-136. [DOI: https://dx.doi.org/10.3141/2386-15]

22. Li, Y.; Abdel-Aty, M.; Yuan, J.; Cheng, Z.; Lu, J. Analyzing traffic violation behavior at urban intersections: A spatio-temporal kernel density estimation approach using automated enforcement system data. Accid. Anal. Prev.; 2020; 141, 105509. [DOI: https://dx.doi.org/10.1016/j.aap.2020.105509]

23. Mannering, F.; Bhat, C.R.; Shankar, V.; Abdel-Aty, M. Big data, traditional data and the tradeoffs between prediction and causality in highway-safety analysis. Anal. Methods Accid. Res.; 2020; 25, 100113. [DOI: https://dx.doi.org/10.1016/j.amar.2020.100113]

24. Martínez-Ruíz, D.M.; Fandiño-Losada, A.; de Leon, A.P.; Arango-Londoño, D.; Mateus, J.C.; Jaramillo-Molina, C.; Bonilla-Escobar, F.J.; Vivas, H.; Vanlaar, W.; Gutiérrez-Martínez, M.I. Impact evaluation of camera enforcement for traffic violations in Cali, Colombia, 2008–2014. Accid. Anal. Prev.; 2019; 125, pp. 267-274. [DOI: https://dx.doi.org/10.1016/j.aap.2019.02.002]

25. Moore, J.E.; Giuliano, G.; Cho, S. Secondary accident rates on Los Angeles freeways. J. Transp. Eng.; 2004; 130, pp. 280-285. [DOI: https://dx.doi.org/10.1061/(ASCE)0733-947X(2004)130:3(280)]

26. Alonso, F.; Esteban, C.; Calatayud, C.; Sanmartín, J. Speed and road accidents: Behaviors, motives, and assessment of the effectiveness of penalties for speeding. Am. J. Appl. Psychol.; 2013; 1, pp. 58-64.

27. Wright, N.A.; Lee, L.-T. Alcohol-related traffic laws and drunk-driving fatal accidents. Accid. Anal. Prev.; 2021; 161, 106358. [DOI: https://dx.doi.org/10.1016/j.aap.2021.106358]

28. Zhang, G.; Tan, Y.; Zhong, Q.; Hu, R. Analysis of traffic crashes caused by motorcyclists running red lights in Guangdong Province of China. Int. J. Environ. Res. Public Health; 2021; 18, 553. [DOI: https://dx.doi.org/10.3390/ijerph18020553]

29. Li, X.; Zhuge, C.; Yu, B. Analysis on the impact of illegal driver behaviors on road traffic accidents case study on China. Proceedings of the 2019 11th International Conference on Intelligent Human-Machine Systems and Cybernetics (IHMSC); Hangzhou, China, 24–25 August 2019; IEEE: new York, NY, USA: pp. 250-253.

30. Goldenbeld, C.; Daniels, S.; Schermers, G. Red light cameras revisited. Recent evidence on red light camera safety effects. Accid. Anal. Prev.; 2019; 128, pp. 139-147.

31. Bagasatwika, A. Electronic traffic law enforcement: Is it able to reduce traffic violations?. Unnes Law J.; 2020; 6, pp. 73-96. [DOI: https://dx.doi.org/10.15294/ulj.v5i1.28642]

32. Scott, A.; Balthrop, A.; Miller, J. Did the Electronic Logging Device Mandate Reduce Accidents?. 2019; Available online: https://ssrn.com/abstract=3314308 (accessed on 17 January 2025).

33. Maria, E.; Budiman, E.; Taruk, M. Measure distance locating nearest public facilities using Haversine and Euclidean Methods. J. Phys. Conf. Ser.; 2020; 1450, 012080. [DOI: https://dx.doi.org/10.1088/1742-6596/1450/1/012080]

34. Azdy, R.A.; Darnis, F. Use of haversine formula in finding distance between temporary shelter and waste end processing sites. J. Phys. Conf. Ser.; 2020; 1500, 012104. [DOI: https://dx.doi.org/10.1088/1742-6596/1500/1/012104]

35. Lin, K.-C.; Liao, I.-E.; Chen, Z.-S. An improved frequent pattern growth method for mining association rules. Expert Syst. Appl.; 2011; 38, pp. 5154-5161. [DOI: https://dx.doi.org/10.1016/j.eswa.2010.10.047]

36. Siahaan, A.P.U.; Ikhwan, A.; Aryza, S. A novelty of data mining for promoting education based on FP-growth algorithm. Int. J. Civ. Eng. Technol.; 2018; 9, pp. 1660-1669.

37. Zeng, Y.; Yin, S.; Liu, J.; Zhang, M. Research of Improved FP-Growth Algorithm in Association Rules Mining. Sci. Program.; 2015; 2015, 910281. [DOI: https://dx.doi.org/10.1155/2015/910281]

38. Inada, H.; Ashraf, L.; Campbell, S. COVID-19 lockdown and fatal motor vehicle collisions due to speed-related traffic violations in Japan: A time-series study. Inj. Prev.; 2021; 27, pp. 98-100. [DOI: https://dx.doi.org/10.1136/injuryprev-2020-043947]

39. Tawfeek, M.H.; El-Basyouny, K. A context identification layer to the reasoning subsystem of context-aware driver assistance systems based on proximity to intersections. Transp. Res. Part C Emerg. Technol.; 2020; 117, 102703. [DOI: https://dx.doi.org/10.1016/j.trc.2020.102703]

40. Bucchi, A.; Sangiorgi, C.; Vignali, V. Traffic psychology and driver behavior. Procedia-Soc. Behav. Sci.; 2012; 53, pp. 972-979. [DOI: https://dx.doi.org/10.1016/j.sbspro.2012.09.946]

41. Hussain, Q.; Alhajyaseen, W.K.; Brijs, K.; Pirdavani, A.; Brijs, T. Innovative countermeasures for red light running prevention at signalized intersections: A driving simulator study. Accid. Anal. Prev.; 2020; 134, 105349. [DOI: https://dx.doi.org/10.1016/j.aap.2019.105349]

42. Zhu, D.; Sze, N.; Bai, L. Roles of personal and environmental factors in the red light running propensity of pedestrian: Case study at the urban crosswalks. Transp. Res. Part F Traffic Psychol. Behav.; 2021; 76, pp. 47-58. [DOI: https://dx.doi.org/10.1016/j.trf.2020.11.001]

43. Zhou, R.; Chen, H.; Chen, H.; Liu, E.; Jiang, S. Research on traffic situation analysis for urban road network through spatiotemporal data mining: A case study of Xi’an, China. IEEE Access; 2021; 9, pp. 75553-75567. [DOI: https://dx.doi.org/10.1109/ACCESS.2021.3082188]

44. Ruan, Z.; Song, C.; Yang, X.-h.; Shen, G.; Liu, Z. Empirical analysis of urban road traffic network: A case study in Hangzhou city, China. Phys. A Stat. Mech. Its Appl.; 2019; 527, 121287. [DOI: https://dx.doi.org/10.1016/j.physa.2019.121287]

45. Wang, Z.; Yue, Y.; Li, Q.; Nie, K.; Tu, W.; Liang, S. Analyzing risk factors for fatality in urban traffic crashes: A case study of Wuhan, China. Sustainability; 2017; 9, 897. [DOI: https://dx.doi.org/10.3390/su9060897]