Evaluation method for traffic safety

Abstract

The invention discloses an evaluation method for traffic safety, and belongs to the technical field of traffic safety, and the method comprises the steps: constructing a traffic safety evaluation index system which comprises a traffic safety target layer, a traffic safety dimension layer and a traffic safety index layer, the traffic safety dimension layer comprises a vehicle dimension, a traffic dimension, a human dimension, a social environment dimension, a social economic dimension, a natural environment dimension and a road dimension; the traffic safety index layer sets specific indexes based on each dimension; and calculating the score of each dimension, and calculating to obtain a comprehensive score based on the weight determined by the literature analysis method so as to evaluate the traffic safety condition. Through the traffic safety evaluation index system, the scientificity, comprehensiveness and effectiveness of traffic safety evaluation, the real-time data and static factor coupling analysis method and the flexible adjustment capability of the weight are improved, and reliable technical support is provided for short-term risk response and long-term planning construction of traffic safety management.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic safety, and more specifically, to an evaluation method for traffic safety. Background Art In recent years, with the rapid advancement of transportation infrastructure construction and the widespread application of intelligent transportation systems, traffic safety issues have received increasing attention from all walks of life. As an important tool to guide traffic planning, accident prevention and safety management, traffic safety evaluation methods are developing in the direction of data-driven and intelligent. Traditional traffic safety evaluation methods mainly rely on real-time data for evaluation. These data come from traffic flow monitoring systems, vehicle speed monitoring, weather sensors, and smart cameras, which can capture the dynamic changes of short-term traffic environments. Through these real-time data, the evaluation system can identify traffic safety hazards caused by short-term factors such as changes in vehicle density, reduced road friction coefficient, and abnormal driver behavior, providing a basis for rapid response for management departments. This traffic safety evaluation method based on real-time data has significant advantages in dynamics and timeliness, especially after a traffic accident occurs, it can quickly locate the risk source and formulate corresponding emergency plans.

[0002] However, although the application of real-time data in traffic safety assessment provides important technical support, its evaluation system also has obvious shortcomings. First, this evaluation method is highly dependent on short-term data and focuses on assessing whether there are safety hazards in the instantaneous traffic state, but ignores the comprehensive assessment of the safety and stability of the long-term traffic system. Traffic safety is not only the product of the dynamic changes of the traffic system, but also deeply affected by long-term static background factors. For example, static factors such as population density, regional economic development level, social environment, infrastructure construction level, and traffic planning and design, although their change cycle is long, directly or indirectly affect the traffic demand and traffic behavior patterns in the region. These background factors not only shape the long-term operating characteristics of the regional traffic system, but also play an important role in the distribution of accident risks, accident types and accident frequency.

[0003] Secondly, the current traffic safety evaluation method based on real-time data is difficult to fully reveal the interaction mechanism between dynamic and static factors. For example, a short-term surge in traffic volume may cause safety hazards, but this phenomenon may be related to long-standing unreasonable traffic planning or infrastructure defects. For another example, the frequent accidents on a specific road section may be related to its design defects or long-term flooding black spots, but these problems are usually not fully reflected by short-term dynamic data. In addition, the evaluation system that relies solely on dynamic data may also ignore the long-term trend of traffic safety, causing management departments to pay more attention to short-term risk response when formulating policies, while ignoring preventive construction and the construction of long-term mechanisms. This fragmented evaluation method may not provide effective guidance for the overall optimization and safety management of regional transportation systems. Summary of the invention

[0004] In order to overcome the shortcomings of existing technologies, an evaluation method for traffic safety is proposed. This method constructs a scientific and comprehensive traffic safety evaluation index system, adopts multi-dimensional, multi-algorithm and quantitative methods, improves the scientificity, comprehensiveness and effectiveness of traffic safety assessment, and the real-time data and static factor coupling analysis method and the flexible adjustment of weights provide reliable technical support for short-term risk response and long-term planning and construction of traffic safety management.

[0005] The technical solution adopted by the present invention to solve the technical problem is: a method for evaluating traffic safety, the improvement of which is that the method comprises the following steps: Step 1: Construct a traffic safety evaluation index system, which includes: Determine the traffic safety target layer, including the dual goals of road traffic safety assessment and regional traffic safety situation assessment; Determine the traffic safety dimension layer, and conduct collaborative analysis of the vehicle dimension, traffic dimension, human dimension, social environment dimension, socio-economic dimension, natural environment dimension and road dimension by coupling the real-time data dimension with the long-term static background dimension; Determine the traffic safety indicator layer and set differentiated indicators based on the types of each dimension, including: The vehicle dimension uses a dynamic difference indicator between the real-time vehicle speed and the road speed limit; The traffic dimension uses a composite index of dangerous vehicle rating and human-machine separation construction; The human dimension, socio-economic dimension, and natural environment dimension use month-on-month change indicators based on historical benchmarks; The road dimension adopts the spatial risk index of the distribution density of waterlogging points; The social environment dimension uses the dynamic impact index of the construction site safety score; Step 2: Based on the traffic safety evaluation index system, calculate the scores of each dimension respectively, including: For the human dimension, socio-economic dimension, and natural environment dimension, a recursive standardization method based on historical benchmark values ​​is used to calculate the month-on-month change score; For the traffic dimension, road dimension and social environment dimension, the spatial radiation impact analysis method is used to calculate the multi-dimensional cross-effect score; For the vehicle dimension, the CART regression tree algorithm is used to dynamically score the difference between the real-time vehicle speed and the road speed limit; Step 3: Determine the weight of each dimension based on the literature analysis method, dynamically adjust the weight of each dimension through the hierarchical analysis method (AHP), and obtain a comprehensive score to evaluate the traffic safety situation.

[0006] Furthermore, the specific indicators of each dimension further include: Vehicle dimension: The difference between the real-time vehicle speed and the road speed limit is combined with the CART regression tree algorithm for classification and prediction, and the vehicle dimension score is calculated to reflect the smoothness and safety of the road; Traffic dimension: Based on the level and quantity of dangerous vehicles and the construction of road separation between man and machine, the linear attenuation method is used to calculate the score of the dangerous vehicle impact range, and the latitude and longitude interpolation method is combined to calculate the score of the affected grid points. Finally, the comprehensive score of the traffic dimension is weighted to calculate the level and quantity of dangerous vehicles and the construction of road separation between man and machine; Human dimension: Based on population density, the dimension score is calculated by combining its month-on-month change with the benchmark value of a specific year through standardization; Social environment dimension: Combine the safety conditions of construction sites, dangerous houses and adjacent dangerous areas to calculate their potential impact on traffic safety, and then use the weighted sum to obtain the social environment dimension score; Socio-economic dimension: The score of the socio-economic dimension is calculated based on the quarterly regional GDP macroeconomic indicator and its historical trend; Natural environment dimension: The dimension score is calculated based on the seasonal climate warning coefficient and weather warning, combined with the impact of natural environment changes, after standardization; Road dimension: Based on the distribution of waterlogging spots and flooding black spots, the Gaussian diffusion model is used to calculate the safety status of the affected area and derive a safety assessment score for the road dimension.

[0007] Furthermore, the specific steps of calculating the scores of each dimension include: For the human dimension, socio-economic dimension, and natural environment dimension, the dimension scores are calculated based on their month-on-month changes, recursively based on a specific year, and standardized; For the dimensions of traffic, roads and social environment, the dimension scores are calculated based on the calculation rules of the radiation center grid points and the irradiated grid points, combined with the natural language processing (NLP) method and the Gaussian diffusion model algorithm; The CART regression tree method is used to classify and predict the average vehicle speed, road speed limit and traffic conditions in the vehicle dimension, and calculate the vehicle dimension score.

[0008] Furthermore, the radiation center grid point score further includes: For the records of flooding black spots and waterlogging spots, the original status is decomposed into multiple records year by year, and the NLP method is used to learn and filter keywords to obtain the correlation coefficient matrix, and the points are assigned according to the absolute value difference of the elements; For records of people who need to be relocated, including construction sites, dangerous houses and adjacent dangerous areas, the score of the original center point is calculated according to the scale of relocation; For dangerous vehicles, the raw center point score is determined based on the vehicle type.

[0009] Furthermore, the irradiated grid point score calculation rule further includes: Conduct regional impact radiation for the points assigned with the center score, and determine the radiation radius and radiation range according to the different situations of dangerous vehicles and dangerous places; The Gaussian diffusion model is used to calculate the transfer pressure score or dangerous vehicle index layer score at different distances from the event site to reflect its impact on the surrounding areas.

[0010] Furthermore, the evaluation index of the vehicle dimension includes real-time vehicle speed and road speed limit, and the scoring calculation method includes: By constructing a long-term panel data set, the difference between the average speed of vehicles and the road speed limit is calculated; Classify and predict the relationship between vehicle speed classification and road speed limit based on CART regression tree algorithm; The classification difference between vehicle speed and road speed limit is calculated, and the classification result is normalized into a vehicle dimension score to reflect road smoothness and safety.

[0011] In the classification and prediction of vehicle speed and road speed limit, the CART regression tree algorithm recursively divides the data set into multiple subsets to find the optimal feature segmentation point. Its objective function is defined as: ; in, Indicates that the variable feature value is Under the condition of data set About ( ) The Gini coefficient of the partition in the CART regression tree; The function representation counts the frequency of variables that meet the conditions in the variable array; It is represented as a certain threshold used to split the data set on the feature variable, which is determined endogenously by the model fitting; Represented as a set of samples under certain conditions; and Respectively represent the characteristic variable values Under the condition, the subsets that meet and do not meet the condition; is the characteristic variable of the Gini coefficient; is the proportion of samples of this category in the data set, is the total sample data count, that is, the number of all samples in the data set; and Respectively represent the number of samples belonging to one category and another category; and They respectively represent the proportion of samples belonging to one category and another category in the total sample size.

[0012] The beneficial effects of the present invention are: by constructing a scientific and comprehensive traffic safety evaluation index system and adopting a multi-dimensional, multi-algorithm and quantitative method, the scientificity, comprehensiveness and effectiveness of traffic safety assessment are improved. The real-time data and static factor coupling analysis method and the flexible adjustment ability of weights provide reliable technical support for short-term risk response and long-term planning and construction of traffic safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A flow chart of a method for evaluating traffic safety according to the present invention; Figure 2 This is a module diagram of a traffic safety evaluation index system of the present invention. DETAILED DESCRIPTION

[0014] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0015] The following will clearly and completely describe the concept, specific structure and technical effects of the present invention in combination with the embodiments and drawings, so as to fully understand the purpose, characteristics and effects of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, other embodiments obtained by technicians in this field without creative work are all within the scope of protection of the present invention. In addition, all the connection / connection relationships involved in the patent do not refer to the direct connection of components, but refer to the formation of a better connection structure by adding or reducing connection accessories according to the specific implementation situation. The various technical features in the invention can be combined interchangeably without conflicting with each other.

[0016] See also Figure 1-Figure 2 As shown, the present invention provides a method for evaluating traffic safety, which comprises the following steps: Constructing a traffic safety evaluation index system, which includes a traffic safety target layer, a traffic safety dimension layer and a traffic safety index layer, wherein the target layer sets an overall goal including the safety of road traffic and the safety of regional traffic; The traffic safety dimension layer is subdivided into a real-time dimension, a long-term background dimension, and a combined dimension to evaluate road traffic safety and regional traffic safety; The traffic safety index layer sets specific indicators based on each dimension, including but not limited to: Vehicle dimension: The difference between the real-time vehicle speed and the road speed limit is combined with the CART regression tree algorithm for classification and prediction, and the vehicle dimension score is calculated to reflect the smoothness and safety of the road; Traffic dimension: Based on the level and quantity of dangerous vehicles and the construction of road separation between man and machine, the linear attenuation method is used to calculate the score of the dangerous vehicle impact range, and the latitude and longitude interpolation method is combined to calculate the score of the affected grid points. Finally, the comprehensive score of the traffic dimension is weighted to calculate the level and quantity of dangerous vehicles and the construction of road separation between man and machine; Human dimension: Based on population density, the dimension score is calculated by combining its month-on-month change with the benchmark value of a specific year through standardization; Social environment dimension: Combine the safety conditions of construction sites, dangerous houses and adjacent dangerous areas to calculate their potential impact on traffic safety, and then use the weighted sum to obtain the social environment dimension score; Socio-economic dimension: The score of the socio-economic dimension is calculated based on the quarterly regional GDP macroeconomic indicator and its historical trend; Natural environment dimension: The dimension score is calculated based on the seasonal climate warning coefficient and weather warning, combined with the impact of natural environment changes, after standardization; Road dimension: Based on the distribution of waterlogging spots and flooding black spots, the Gaussian diffusion model is used to calculate the safety status of the affected area and derive a safety assessment score for the road dimension.

[0017] In the present invention, the traffic safety evaluation is divided into real-time dimension, long-term background dimension and combined dimension, and the real-time traffic data (such as vehicle speed, road speed limit, etc.) is combined with long-term background factors (such as population density, socio-economic background, natural environment and road construction, etc.), and the joint effect of dynamic and static factors is fully considered. Dynamic data can reflect short-term traffic safety hazards, and static background factors reveal the causes of long-term accident risks. The coupling analysis of the two makes up for the deficiency of traditional evaluation methods that only focus on real-time data and ignore long-term factors, thereby improving the comprehensiveness and scientificity of traffic safety evaluation. The present invention also constructs a three-level traffic safety evaluation index system, covering the target layer, dimension layer and indicator layer. The dimension layer is subdivided into seven dimensions, including vehicle driving conditions, traffic factors, population characteristics, social environment, socio-economic background, natural environment and road construction, which can fully cover the key factors affecting traffic safety. At the same time, specific indicators (such as real-time vehicle speed, dangerous vehicle level, road man-machine separation construction, seasonal climate warning, etc.) have achieved quantitative measurement, which can intuitively reflect the traffic safety situation and potential hazards, and ensure the scientificity and operability of the evaluation results.

[0018] Further, calculating the scores of each dimension includes the following steps; For the human dimension, socio-economic dimension, and natural environment dimension, the dimension scores are calculated based on their month-on-month changes, recursively based on a specific year, and standardized; For the dimensions of traffic, roads and social environment, the dimension scores are calculated based on the calculation rules of the radiation center grid points and the irradiated grid points, combined with the natural language processing (NLP) method and the Gaussian diffusion model algorithm; The CART regression tree method is used to classify and predict the average vehicle speed, road speed limit and traffic conditions in the vehicle dimension, and calculate the vehicle dimension score.

[0019] The Gaussian diffusion model is used to calculate the radiation impact range of traffic accidents or dangerous vehicles and the safety status of the irradiated area, and can accurately evaluate the risk spread of accidents to surrounding areas.

[0020] The natural language processing (NLP) method is used to analyze historical records (such as the status of flooding black spots or dangerous houses), extract key features and quantify scores to ensure effective use of historical information.

[0021] The CART regression tree algorithm: accurately models the classification and prediction of vehicle speed and road speed limit, normalizes the results, and scientifically evaluates the safety status of the vehicle dimension.

[0022] Through the application of multiple algorithms and models, the present invention can improve the accuracy of traffic safety evaluation and adapt to complex traffic environments in different regions and scenarios.

[0023] In this embodiment, the scores of the human, socio-economic and natural environment dimensions are calculated as follows: The scores of the human, social economy, and natural environment dimensions are affected by a single primary indicator, which is set as a background factor. Their month-on-month changes are considered in equal proportions, and are recursively calculated with 2015 as the initial benchmark year: Take equal weight ,in is the weight vector of human dimension; is the weight vector of socioeconomic dimensions; Natural dimension weight vector.

[0024] After standardization, take the original data I i They are population density PD, the inverse of the regional quarterly gross national product 1 / GDP, and the multi-year average precipitation seasonality coefficient S, which correspond to the three dimensions of people, social economy, and natural environment mentioned above. The original score of dimension i at time t is: .

[0025] This score is the corresponding dimension within the coverage area of ​​the region safety score.

[0026] The traffic, road and social environment dimension score calculation method is: Radiation center grid score calculation rules: Calculate the original score of the original center point at time t for the flood black spots and waterlogging spots with text records in the data category ,in They represent flooding black spots and water-prone spots respectively. Indicates Records: Divide the original state of each record in the two variables into (y-y0) records year by year, including the initial year y0 and the repair progress to the current year y Then use the NLP method to learn (y-y0) records, filter verbs and nouns, and obtain the upper triangular matrix U of correlation coefficients; Calculate the absolute value difference of the elements in U, ; Pick The quartiles and bins of are used as the standard to assign points to the verbs before the nouns in U. Additional processing includes: assigning the same score to synonyms, assigning negative scores to negative correlation coefficients, and assigning the score intervals consistent with or related to the final safety score level division. A set of word score correspondence tables is obtained, word L, score .

[0027] Finally, the original score of a record at time t is calculated. For a record K, there are L words in total. ; For indicators with a number of people to be relocated, including construction site conditions, dangerous housing conditions, and adjacent dangerous areas, since personnel transfer has general mobility and the pressure of personnel transfer is related to the degree of personnel concentration, the original score of the original center point at time t is calculated according to the scale of transfer ,in They represent construction sites, dangerous houses, and leading area indicators respectively. Indicates Records: The average area required per person is d, and the number of people who need to be transferred , Total area required: ; As an indicator of carrying pressure Score. For all data, we get the data set transport area sequence D. For D, we fit the normal distribution and get For higher than The risk score of this dimension is set to 6 points. , The corresponding segmentation standards are shown in Table 1 (5-6, 6-8, 8-10, 10 points).

[0028] This score It is the safety score of the place where the incident occurred within the coverage area of ​​the region, specifically: For the records of flooding black spots and waterlogging spots, the original status is decomposed into multiple records year by year, and the NLP method is used to learn and filter keywords to obtain the correlation coefficient matrix, and the points are assigned according to the absolute value difference of the elements; For records of people who need to be relocated, including construction sites, dangerous houses and adjacent dangerous areas, the score of the original center point is calculated according to the scale of relocation; For dangerous vehicles, the raw center point score is determined based on the vehicle type.

[0029] The irradiated grid point score calculation rule further includes: The points assigned with the central score are radiated with regional impact. According to different dangerous situations, the radiation radius and radiation range are determined. The Gaussian diffusion model is used to measure the attenuation law, and the water accumulation pressure and personnel transfer pressure at different distances from the incident point are calculated. According to historical conditions, the pressure score or the dangerous vehicle index layer score is given to reflect its impact on the surrounding area. The spatial summation of all event radiation scores is performed to obtain the corresponding index safety score. ,in They respectively represent flooding black spots, water-prone areas, construction sites, dangerous houses, and areas adjacent to dangerous areas.

[0030] By decomposing the records year by year and combining NLP methods to learn and filter keywords, key features can be extracted, a correlation coefficient matrix can be constructed, the main factors affecting traffic safety can be accurately identified, and the risk source can be accurately located. By conducting a radiation analysis of the regional impact of the center point, the scope and degree of impact of dangerous vehicles and dangerous locations on the surrounding areas can be clarified, avoiding focusing on a single risk point while ignoring its potential regional hazards. The determination of the radiation radius and range can dynamically adapt to different types of hazards (such as dangerous vehicles, construction sites, and water-prone points), and more accurately reflect the actual risk distribution. The Gaussian diffusion model can truly simulate the attenuation characteristics of risks in space, reflecting the intensity of the impact of risks at different locations from the center point. This calculation method based on a mathematical model is more scientific and avoids errors dominated by human experience. The gradual attenuation characteristics of the Gaussian model make the close-range impact more significant, while the long-range impact gradually weakens, which is highly consistent with the actual situation.

[0031] The evaluation indicators of the traffic dimension include the level and number of dangerous vehicles and the construction of road human-machine separation. The scoring calculation method includes: Dangerous vehicles: Classify them into Class I dangerous vehicles and Class II dangerous vehicles according to their types, set their real-time longitude and latitude as the center point, and assign fixed center scores to them. ,in , Indicates records; The influence range of dangerous vehicles is linearly weakened and the scores of affected grid points are determined by the linear spatial interpolation method of longitude and latitude. The safety scores of all radiation grid points with or without dangerous vehicles are summed up to obtain the time , safety score of each longitude and latitude grid point ,in ; The construction of road human-machine separation is matched, graded, and scored according to the national legal standards and the final safety level classification of the implementation example. The score is scored according to the actual road conditions to obtain the safety score at a certain latitude and longitude grid point on the road. ,in , which is part of the comprehensive score of the road traffic dimension.

[0032] In addition, for a road , divided into There are different road sections with different road separation construction conditions, and the longitude and latitude grid points of the road sections are given safety , according to its road length ,according to ; Calculate the traffic dimension score based on the road.

[0033] In this embodiment, the original center point score is determined for the dangerous vehicle according to the type of the dangerous vehicle. , indicating dangerous vehicle indicators, defining first-level dangerous vehicles, including dangerous goods transport vehicles, ordinary freight trucks, concrete trucks and mud trucks, original center point score Secondary dangerous vehicles, including long-distance passenger vehicles, tourist chartered buses, school buses and public buses, original center point score point.

[0034] Assigning a center score Event, considering the regional radiation impact, there are two situations: General assumption: The presence of a hazardous vehicle or hazardous location affects all azimuths.

[0035] Radiation radius determination: including dangerous vehicles ( ), the positioning is known, it is clearly known that the impact is small-scale, and the impact range is fixed (0.002 km), the longitude and latitude accuracy of this invention is is 0.0001*0.0001, corresponding to a ground distance of 0.01 km and a vehicle length of , the danger level is in linear form. Set the critical danger score to , is a small constant, then There are scores for each azimuth angle, and the distance to the vehicle positioning within its influence range km away, the score satisfies: ; Note: Refers to The records indicate the safety score where there are dangerous vehicles, that is, the score of the radiation center point; Refers to the scores of each longitude and latitude grid point within the influence range of the dangerous vehicle after linear attenuation; use It represents the dangerous vehicle index layer score after all scores on the latitude and longitude grid points are superimposed; among all variables, it is the indicative dangerous vehicle index layer. .

[0036] Uncertain radiation radius: including flooding black spots, water-prone spots, construction sites, dangerous houses, and adjacent dangerous areas (with ), calculate the corresponding area. Assume that the transfer pressure is centered at the incident location and spreads outward evenly. For flooding black spots and water-prone spots, directly use the safety score of the center point. As an impact pressure. For construction sites, dangerous houses, leading areas, in either direction, a single person occupies an area of , the single transfer area pressure is , the corresponding safety score is The maximum number of people a bus can carry is , then the rate at which the transport pressure decreases with the distance from the event center is .

[0037] The transport process is analogized to the Gaussian diffusion model of the uniform diffusion process of stable ambient gas, then the distance from the event point km safety pressure is: ; ; in, Indicates the event The score is centered on the place of occurrence. The attenuation function relationship in space.

[0038] The longitude and latitude of the center point of the longitude and latitude Gaussian spatial interpolation method is , the latitude and longitude accuracy of the regional model is , each unit change in longitude and latitude corresponds to the distance on Earth ,by Centered on The fraction is fractionally decayed, and its decay function is , the longitude and latitude of the determined point The distance to the center point is calculated as ,satisfy

[0039] , decays until: ; When it was first established, is the distance when the condition is met, where is the cutoff score, taking the smallest constant. Determine the latitude and longitude of the point Selection: Take polar coordinate interpolation, For the farthest radius that meets the above relationship, the maximum number of interpolation steps along the radius is : ,right , Take it as 1, then each group for: ; Sum of area scores and road segment matching: by Matching sum, a point The moment score is: ; in To determine whether there is a corresponding indicator at this point The score density function is 1 when it exists and 0 when it does not exist. The road range is matched in units of longitude and latitude, and the sum is the total score of the road segment.

[0040] Further, the literature analysis method comprises the following steps: According to the correlation between the research object and the indicator layer, clarify the type of literature to be collected, select an authoritative literature database, and combine the evaluation topic and secondary indicators to generate relevant keywords: Quickly browse the titles and abstracts of the literature to determine their relevance to the research object and indicator level, eliminate irrelevant literature, evaluate the quality of the literature based on the author background, publication journal, publisher, and number of citations, and select high-quality literature for in-depth research; Read the selected literature in detail, extract information related to the indicator layer weights, record the source of each weight data in detail, including the literature title, author and publication time information, summarize the extracted weight data, calculate the weight according to the relative proportion of each indicator, adjust and normalize according to the actual situation; According to the processing results, the weights of each first-level indicator and second-level indicator are determined.

[0041] In this embodiment, the scope of the literature is first determined. According to the correlation between the research object and the indicator layer, the type of literature to be collected is determined to be journal articles. Then, authoritative literature databases such as China National Knowledge Infrastructure, Scopus, Web of Science, etc. are selected for retrieval to ensure the quality and reliability of the literature. In combination with the research topic and the secondary indicators, search keywords are formulated. For example, "traffic safety evaluation", "AHP hierarchical analysis method", "flooding black spots", etc.

[0042] Quickly browse the title and abstract of the literature to determine whether it is relevant to the research object and indicator level, and exclude obviously irrelevant literature. Then evaluate the quality of the literature from aspects such as author background, publication journal or publisher, and number of citations, and select high-quality literature.

[0043] Read the selected literature and extract information related to the indicator layer weights, such as AHP hierarchical analysis method, statistical analysis results, etc. Record the source literature of each weight data in detail, including the title, author, publication time and other information, for subsequent tracing and verification. Summarize the extracted weight data, calculate the weight according to the relative proportion of each indicator, adjust it according to the actual situation, and perform normalization.

[0044] It was finally determined that the first-level indicator road dimension includes flooding black spots and waterlogging spots, with weights of [0.5, 0.5]; the first-level indicator traffic dimension includes dangerous vehicles and human-machine separation, with weights of [0.3, 0.7]; the first-level indicator social environment dimension includes construction sites, dangerous areas and dangerous houses, with weights of [0.4572, 0.2686, 0.2742].

[0045] The direct influencing factors of traffic safety include average vehicle speed, road speed limit and traffic conditions. Since there is a restrictive relationship between road speed limit and average vehicle speed events, the CART regression tree method is used to classify and predict the variables and indicator changes of the two. The process is as follows: 1) Construct a long-term panel data set for each road section, including average vehicle speed, road speed limit, and the difference between vehicle speed and speed limit. The difference between vehicle speed and speed limit y is calculated as: ; in is the vehicle speed, Speed ​​limit for the road.

[0046] 2) Panel data solution feature segmentation and feature value selection: Dataset segmentation: , split into two sets of numbers, is the feature value of segmentation data, , They represent the two data sets after g is divided according to the eigenvalues, and var represents each value. The means of the two data sets are ; 3) Continue recursively to traverse all possible solutions , find the optimal solution so that the objective function is minimized and the screening conditions are met. Objective function: , Set the filter conditions to determine whether the vehicle speed classification matches the road speed limit: Standard road speed limit classifications include (km / h):

[0047] For solution Middle Pair The characteristic value of is taken, and the minimum difference between it and the standard road speed limit and the average value of is calculated: ; in for The number of eigenvalues ​​of in the solution, It is recorded as the mean of ping, and min means the minimum value. The limiting condition 5 is the minimum speed interval in the road speed limit classification. Since 5km / h is often associated with congestion events, this right uses this value as the standard to define the speed interval.

[0048] 4) After obtaining the optimal solution, use The coefficient is normalized and used as the indicator weight.

[0049] Coefficient, for the data set , for a classification standard ,have: ; in, Indicates that the variable feature value is Under the condition of data set About ( ) The Gini coefficient of the partition in the CART regression tree; The function representation counts the frequency of variables that meet the conditions in the variable array; It is represented as a certain threshold used to split the data set on the feature variable, which is determined endogenously by the model fitting; Represented as a set of samples under certain conditions; and Respectively represent the characteristic variable values Under the condition, the subsets that meet and do not meet the condition; is the characteristic variable of the Gini coefficient; is the proportion of samples of this category in the data set, is the total sample data count, that is, the number of all samples in the data set; and Respectively represent the number of samples belonging to one category and another category; and They respectively represent the proportion of samples belonging to one category and another category in the total sample size.

[0050] This proposal proposes the following method to incorporate road access restrictions into traffic safety indicators, which is the best way to match data: (1) Adjust the road speed limit to the critical value in cases where there are mandatory traffic restrictions . In addition: according to relevant traffic and meteorological regulations, the speed limit is 5km / h in situations where traffic is prohibited (such as typhoon red warning, road closure); other situations are determined according to real-time regulations.

[0051] (2) Construct a traffic safety index to score the traffic dimension : ; Among them, v is the average vehicle speed, vl is the average speed limit of the road, f(v) is the safety score relationship between the average vehicle speed and the road speed limit, which is a non-parametric relationship obtained based on historical data; f((v-vl) / vl) is the safety score relationship between the difference between the vehicle speed and the road speed limit and the road speed limit, which is a non-parametric relationship obtained based on historical data.

[0052] This indicator variable form best reflects the speed driving conditions and road restriction conditions, and the feedback information measures abnormal speed conditions such as congestion and speeding during vehicle driving.

[0053] The normal distribution is used to fit the dimensional hierarchical safety scores, and nonlinear mapping is performed to each dimension to form a consistent safety score classification. The above method is also used to obtain the weights of each dimensional level, and the target layer regional traffic safety score and road traffic safety score are calculated. The regional traffic safety includes five dimensions: roads, social environment, natural environment, people and social economy, and their weights are 0.1355, 0.1235, 0.1113, 0.3261, and 0.0722 respectively; road traffic safety includes seven dimensions: vehicle dimension, traffic dimension, road dimension, social environment dimension, natural environment dimension, people dimension and social economy dimension, and their weights are 0.1883, 0.0431, 0.1355, 0.1235, 0.1113, 0.3261 and 0.0722 respectively.

[0054] In practical applications, the weights can be adjusted in a targeted manner using the analytic hierarchy process (AHP) according to the requirements of specific time and place. The specific safety score levels are as follows: Table 1

[0055] The preferred implementation of the present invention has been specifically described, but the invention is not limited to the described embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for evaluating traffic safety, characterized in that: The method comprises the following steps: Step 1: Construct a traffic safety evaluation index system, which includes: Determine the traffic safety target layer, including the dual goals of road traffic safety assessment and regional traffic safety situation assessment; Determine the traffic safety dimension layer, and conduct collaborative analysis of the vehicle dimension, traffic dimension, human dimension, social environment dimension, socio-economic dimension, natural environment dimension and road dimension by coupling the real-time data dimension with the long-term static background dimension; Determine the traffic safety indicator layer and set differentiated indicators based on the types of each dimension, including: The vehicle dimension uses a dynamic difference indicator between the real-time vehicle speed and the road speed limit; The traffic dimension uses a composite index of dangerous vehicle rating and human-machine separation construction; The human dimension, socio-economic dimension, and natural environment dimension use month-on-month change indicators based on historical benchmarks; The road dimension adopts the spatial risk index of the distribution density of waterlogging points; The social environment dimension uses the dynamic impact index of the construction site safety score; Step 2: Based on the traffic safety evaluation index system, calculate the scores of each dimension respectively, including: For the human dimension, socio-economic dimension, and natural environment dimension, a recursive standardization method based on historical benchmark values ​​is used to calculate the month-on-month change score; For the traffic dimension, road dimension and social environment dimension, the spatial radiation impact analysis method is used to calculate the multi-dimensional cross-effect score; For the vehicle dimension, the CART regression tree algorithm is used to dynamically score the difference between the real-time vehicle speed and the road speed limit; Step 3: Determine the weight of each dimension based on the literature analysis method, dynamically adjust the weight of each dimension through the hierarchical analysis method (AHP), and obtain a comprehensive score to evaluate traffic safety conditions.

2. A method for evaluating traffic safety according to claim 1, characterized in that: Specific indicators for each dimension further include: Vehicle dimension: The difference between the real-time vehicle speed and the road speed limit is combined with the CART regression tree algorithm for classification and prediction, and the vehicle dimension score is calculated to reflect the smoothness and safety of the road; Traffic dimension: Based on the level and number of dangerous vehicles and the construction of road separation between man and machine, the linear attenuation method is used to calculate the score of the dangerous vehicle impact range, and the latitude and longitude interpolation method is combined to calculate the score of the affected grid points, and finally the comprehensive score of the traffic dimension is calculated by weighted calculation; Human dimension: Based on population density, the dimension score is calculated by combining its month-on-month change with the benchmark value of a specific year through standardization; Social environment dimension: Combine the safety conditions of construction sites, dangerous houses and adjacent dangerous areas to calculate their potential impact on traffic safety, and then use the weighted sum to obtain the social environment dimension score; Socio-economic dimension: The score of the socio-economic dimension is calculated based on the quarterly regional GDP macroeconomic indicator and its historical trend; Natural environment dimension: The dimension score is calculated based on the seasonal climate warning coefficient and weather warning, combined with the impact of natural environment changes, after standardization; Road dimension: Based on the distribution of waterlogging spots and flooding black spots, the Gaussian diffusion model is used to calculate the safety status of the affected area and derive a safety assessment score for the road dimension.

3. The method for evaluating traffic safety according to claim 1, characterized in that: The specific steps of calculating the scores of each dimension include: For the human dimension, socio-economic dimension, and natural environment dimension, the dimension scores are calculated based on their month-on-month changes, recursively based on a specific year, and standardized; For the dimensions of traffic, roads and social environment, the dimension scores are calculated based on the calculation rules of the radiation center grid points and the irradiated grid points, combined with the natural language processing (NLP) method and the Gaussian diffusion model algorithm; The CART regression tree method is used to classify and predict the average vehicle speed, road speed limit and traffic conditions in the vehicle dimension, and calculate the vehicle dimension score.

4. The method for evaluating traffic safety according to claim 3, characterized in that: The radiation center grid score further includes: For the records of flooding black spots and waterlogging spots, the original status is decomposed into multiple records year by year, and the natural language processing (NLP) method is used to learn and filter keywords to obtain the correlation coefficient matrix, and the scores are assigned according to the absolute value difference of the elements; For records of people who need to be relocated, including construction sites, dangerous houses and adjacent dangerous areas, the score of the original center point is calculated according to the scale of relocation; For dangerous vehicles, the raw center point score is determined based on the vehicle type.

5. The method for evaluating traffic safety according to claim 3, characterized in that: The irradiated grid point score calculation rule further includes: Conduct regional impact radiation for the points assigned with the center score, and determine the radiation radius and radiation range according to the different situations of dangerous vehicles and dangerous places; The Gaussian diffusion model is used to calculate the transfer pressure score or dangerous vehicle index layer score at different distances from the incident point to reflect its impact on the surrounding area; The evaluation index of the vehicle dimension includes real-time vehicle speed and road speed limit, and the scoring calculation method includes: By constructing a long-term panel data set, the difference between the average speed of vehicles and the road speed limit is calculated; Classify and predict the relationship between vehicle speed classification and road speed limit based on CART regression tree algorithm; The classification difference between vehicle speed and road speed limit is calculated, and the classification result is normalized into a vehicle dimension score to reflect road smoothness and safety.

6. A method for evaluating traffic safety according to claim 5, characterized in that: In the classification and prediction of vehicle speed and road speed limit, the CART regression tree algorithm recursively divides the data set into multiple subsets to find the optimal feature segmentation point. Its objective function is defined as: ； in, Indicates that the variable feature value is Under the condition of data set About ( ) The Gini coefficient of the partition in the CART regression tree; The function representation counts the frequency of variables that meet the conditions in the variable array; It is represented as a certain threshold used to split the data set on the feature variable, which is determined endogenously by the model fitting; Represented as a set of samples under certain conditions; and Respectively represent the characteristic variable values Under the condition, the subsets that meet and do not meet the condition; is the characteristic variable of the Gini coefficient; is the proportion of samples of this category in the data set, is the total sample data count, that is, the number of all samples in the data set; and Respectively represent the number of samples belonging to one category and another category; and They respectively represent the proportion of samples belonging to one category and another category in the total sample size.

Images