Road safety comprehensive assessment method and system for overcoming interaction effects of risk factors

Abstract

The present application relates to the technical field of road safety evaluation, in particular to a road safety comprehensive evaluation method and system for overcoming the interaction effect of risk factors, which comprises the following steps: based on constructing a multi-level road risk evaluation index system, a multi-dimensional risk factor set is obtained, a subjective weight is calculated by using a least square optimization model, and an objective entropy weight is obtained by combining historical data to form a comprehensive weight; the comprehensive weight is taken as a fuzzy density, a risk factor interaction coefficient is solved based on a fuzzy measure equation, and a fuzzy measure function is constructed; real-time risk factor data is collected, a comprehensive road risk value is calculated by using Choquet fuzzy integration, random disturbance is carried out by combining Monte Carlo sampling, fuzzy integration calculation is repeated, and the probability distribution and confidence interval of the comprehensive risk value are obtained; and the present application realizes dynamic quantitative evaluation and decision support under the interaction of multi-dimensional risk factors during road risk evaluation.

Description

Technical Field

[0001] This invention relates to the field of road safety assessment technology, and more specifically, to a comprehensive road safety assessment method and system that overcomes the interactive effects of risk factors. Background Technology

[0002] Road risk assessment is a core foundational task in road traffic safety management and infrastructure operation and maintenance. Its core task is to systematically identify, quantify, and predict potential safety risks caused by inherent physical defects in road alignment, roadbed and pavement, traffic engineering facilities, and the surrounding environment, thereby providing a scientific basis for road maintenance management, safety hazard rectification, and traffic safety improvement.

[0003] Traditional road risk assessments primarily rely on manual inspections and specialized testing vehicles. The former, based on on-site experience and subjective judgment, is inefficient, lacks standardized procedures, has limited coverage, and poses safety hazards. The latter, while utilizing high-precision sensors to objectively collect road surface technical conditions, suffers from expensive equipment, long inspection cycles, and delayed data updates. Furthermore, its assessment scope is limited to road surface indicators, failing to cover multi-dimensional risk factors such as linear geometry, traffic facilities, and the roadside environment. Secondly, existing risk assessment models often target single factors, lacking a unified framework to quantify the relative importance of heterogeneous risk factors. This results in a fragmented risk weighting system and results that fail to reflect the overall road safety level. In addition, mainstream comprehensive risk calculations are still based on linear weighting, assuming risks are independent. This fails to express the non-linear interaction effects between risk factors such as sharp bends and water accumulation, leading to a significant deviation between the final assessment results and the actual risk situation.

[0004] In summary, most existing road risk assessment methods remain at the stage of linear weighting, which has problems such as heterogeneity of multidimensional risk factors and distortion of comprehensive road risk assessment results. There is an urgent need for a method that can take the basic weights output from stage one as input, perform deeper nonlinear integration, and finally obtain a comprehensive risk value that conforms to the actual situation. Summary of the Invention

[0005] In view of this, the present invention proposes a comprehensive road safety assessment method and system to overcome the interaction effect of risk factors. It aims to solve the problem that the existing road risk assessment technology has limitations in data acquisition methods, risk factor weighting system and risk fusion calculation. In particular, it cannot uniformly quantify the relative importance of multidimensional heterogeneous risk factors, and the fusion method based on the linear independence assumption cannot reflect the dynamic interaction effect between risks, resulting in a serious deviation between the comprehensive risk assessment results and the actual road risks.

[0006] This invention proposes a comprehensive road safety assessment method to overcome the interaction effects of risk factors, comprising: A road risk assessment index system is constructed based on the target layer, criterion layer, and indicator layer to obtain a set of risk factors. An interval judgment matrix is ​​constructed according to the relative importance of each risk factor in the risk factor set, and the interval judgment matrix is ​​converted into a deterministic judgment matrix using the midpoint of the interval. A least squares optimization model is established based on the deterministic judgment matrix, subjective weights are obtained, and the consistency of the deterministic judgment matrix is ​​checked based on the subjective weights. The deterministic judgment matrix is ​​then adjusted based on the check results. Historical road risk data is acquired, and objective entropy weights are obtained based on the historical road risk data. The fusion weights are obtained by combining subjective weights and objective entropy weights according to preset weight coefficients. The fusion weights are used as fuzzy density, and the risk factor interaction coefficients of the fuzzy density are solved based on the fuzzy measure equation. The fuzzy measure function of the fuzzy density is constructed based on the interaction coefficients. Obtain real-time risk factor data of the road to be evaluated, obtain the risk factor vector of the road to be evaluated based on the real-time risk factor data, and obtain the comprehensive road risk value of the road to be evaluated based on the fuzzy measure function and the risk factor vector of the road to be evaluated. The statistical distribution between the comprehensive road risk value and the fusion weight is obtained. Based on the statistical distribution and Monte Carlo sampling, the real-time risk factor data and the fusion weight are subjected to multiple random perturbations and repeated fuzzy integral calculations to obtain the probability distribution of the comprehensive road risk value of the road to be evaluated and the confidence interval of the probability distribution of the comprehensive road risk value.

[0007] Furthermore, based on the target layer, criterion layer, and indicator layer, a road risk assessment indicator system is constructed. When obtaining the risk factor set, it includes: Road risk is modeled hierarchically based on the hierarchical analysis method structure of target layer, criterion layer and indicator layer; Obtain the road structure, traffic facilities, alignment features, and road environment from the preset road configuration, and extract the evaluation dimensions from the preset road configuration based on the road structure, traffic facilities, alignment features, and road environment; Based on the mechanism of road safety impact, the risk indicators of each dimension in the assessment dimension are evaluated, and the key risk indicators of each dimension in the assessment dimension are identified. Key risk indicators from each dimension are standardized and aggregated, and the aggregate is determined as a risk factor set.

[0008] Furthermore, when constructing an interval judgment matrix based on the relative importance of each risk factor in the risk factor set, and converting this interval judgment matrix into a deterministic judgment matrix using the midpoint of the interval, the following steps are taken: Based on the interval number, each risk factor in the risk factor set is compared pairwise, the comparison results are determined as interval elements, and an interval judgment matrix is ​​constructed based on each interval element. Based on the relationship between the upper and lower bounds of each interval element, obtain the interval midpoint of each interval element; Replace each interval element in the interval judgment matrix with the midpoint of each interval element, and then determine the interval judgment matrix after replacing each interval element as the deterministic judgment matrix.

[0009] Furthermore, when establishing a least-squares optimization model based on the deterministic judgment matrix, obtaining subjective weights, and performing consistency verification on the deterministic judgment matrix based on the subjective weights, the process includes: Obtain the deviation values ​​of each element in the deterministic judgment matrix between each element and its corresponding weight ratio, and establish the objective function of the sum of squared errors of each element based on the deviation values ​​of each element; A least squares optimization model is established based on the squared error of each element, the objective function, the non-negativity constraint of the weights, and the normalization constraint of the weights. Based on the fitting solution of the least squares optimization model, obtain the subjective weight vector that minimizes the bias of the judgment matrix in the deterministic judgment matrix; Based on the relationship between the subjective weight vector and the deterministic judgment matrix, the dynamic consistency index of the deterministic judgment matrix is ​​determined, and based on the relationship between the dynamic consistency index and the configured preset consistency index, it is determined whether the deterministic judgment matrix should be adjusted. When the dynamic consistency index is greater than or equal to the preset consistency index, it is determined that the deterministic judgment matrix will not be adjusted. When the dynamic consistency index is less than the preset consistency index, the deterministic judgment matrix will be adjusted.

[0010] Furthermore, when the dynamic consistency index is less than the preset consistency index, the deterministic judgment matrix is ​​adjusted, including: Obtain the ideal ratio matrix of the deterministic judgment matrix based on subjective weights; Obtain the element-wise difference between each element of the deterministic judgment matrix and the corresponding ideal ratio in the ideal ratio matrix, map the element-wise difference to the corresponding adjustment coefficient, and adjust the element according to the adjustment coefficient, where: The adjustment coefficient is configured such that when the element difference is greater than the preset element difference threshold, the adjustment coefficient is determined to be less than 1; when the element difference is less than the preset element difference threshold, the adjustment coefficient is determined to be greater than 1; and when the element difference is equal to the preset element difference threshold, the adjustment coefficient is determined to be 1. The deterministic judgment matrix is ​​updated based on the adjusted elements, and the inverse elements of each adjusted element in the deterministic judgment matrix are updated simultaneously. Based on the adjusted deterministic judgment matrix, a least squares optimization model is constructed for the second time, and subjective weights are obtained for the second time based on the constructed least squares optimization model; The adjusted deterministic judgment matrix is ​​subjected to a second consistency index verification based on the subjective weights obtained in the second step, until the dynamic consistency index is greater than or equal to the preset consistency index.

[0011] Furthermore, historical road risk data is acquired, and objective entropy weights are obtained based on this data. When obtaining the fusion weights according to the subjective weights and objective entropy weights using preset weighting coefficients, the following steps are taken: Historical road risk data for each time period is obtained, and the historical data is normalized. Based on the normalized historical data, a standardized matrix is ​​constructed. The probability distribution of each risk factor is obtained based on the standardized matrix, and the entropy value of each risk factor is obtained based on the probability distribution. The objective entropy weight of each risk factor is obtained based on the entropy value, and the subjective weight and objective entropy weight are linearly weighted according to the preset weight fusion coefficient to obtain the weight vector. The weight fusion coefficient is configured to be set based on the relative contribution of the subjective weight and the objective entropy weight. The weight vector is processed by normalization to obtain the fusion weights.

[0012] Furthermore, when constructing the fuzzy measure function of the fuzzy density based on the fusion weights as the fuzzy density, and solving for the risk factor interaction coefficients of the fuzzy density based on the fuzzy measure equation, the following steps are included: The fusion weights are used as the marginal fuzzy densities of each risk factor to construct an initial fuzzy density vector; A fuzzy measure interaction equation is constructed based on the marginal fuzzy density of each risk factor. The fuzzy measure parameters are then solved based on the fuzzy measure interaction equation to obtain the interaction coefficients between each risk factor. Based on the marginal fuzzy density and interaction coefficient, and according to the fuzzy measure generation formula, the fuzzy density value of any subset of risk factors is obtained, and the fuzzy measure function is constructed.

[0013] Furthermore, when obtaining real-time risk factor data for the road to be evaluated, obtaining the risk factor vector of the road to be evaluated based on the real-time risk factor data, and obtaining the comprehensive road risk value of the road to be evaluated based on the fuzzy measure function and the risk factor vector of the road to be evaluated, the process includes: Missing values ​​are imputed, noise is filtered out, and outliers are removed from the real-time risk factor data of the road to be evaluated to obtain a risk factor dataset; The risk factor dataset is normalized, and the normalized risk factors are combined according to a preset index order to form a real-time risk factor vector for the road to be evaluated. The incremental contribution value of each item in the real-time risk factor vector is obtained based on the fuzzy measure function and the Choquet fuzzy integral calculation formula. The incremental contribution values ​​are summed to obtain the comprehensive road risk value of the road to be evaluated.

[0014] Furthermore, when obtaining the statistical distribution between the comprehensive road risk value and the fusion weight, the following steps are included: Obtain sample pairs of fusion weights and comprehensive road risk values ​​from several historical assessment periods, and clean and align the sample pairs in chronological order to construct a joint sample matrix composed of fusion weights and comprehensive road risk values; Based on empirical distribution estimation, kernel density estimation, or Gaussian mixture model, the marginal distribution and joint distribution between the fusion weights and the comprehensive road risk value in the joint sample matrix are analyzed. A statistical model is then established based on the marginal distribution and joint distribution between the fusion weights and the comprehensive road risk value. Based on the statistical model, the mean, variance, covariance, and correlation structure of the fusion weights and the comprehensive road risk value are obtained. Based on the mean, variance, covariance, and correlation structure of the fusion weights and the comprehensive road risk value, a statistical distribution between the comprehensive road risk value and the fusion weights is constructed.

[0015] Compared with existing technologies, the advantages of this invention are as follows: By utilizing the subjective weights obtained from the interval judgment matrix and the least squares optimization model, expert experience can participate in weight construction in a structured and verifiable manner, and the reliability of the judgment matrix is ​​improved through consistency verification and adjustment. Simultaneously, the objective entropy weights extracted from historical risk data can truly reflect the differences in information content of each risk factor in historical samples, preventing excessive reliance on human experience in weight allocation, thus constructing a fusion weight system that balances subjective perception and objective data. Secondly, by introducing fuzzy density and fuzzy measure functions to solve for the interaction coefficients between risk factors, the limitation of traditional linear weighting in expressing risk coupling relationships is overcome. It can effectively characterize risk superposition effects such as "sharp bend + water accumulation," enabling the comprehensive risk value to truly reflect the nonlinear gain or inhibition relationship between multiple factors, improving the authenticity and sensitivity of road risk assessment results. Finally, by constructing statistical distributions and performing perturbation analysis on real-time risk factors and fusion weights using Monte Carlo sampling, the probability distribution and confidence interval of the comprehensive road risk value can be obtained, thereby elevating risk assessment from a "single value" to a "probabilistic result." This mechanism not only enhances the robustness and credibility of the assessment results, but also enables the quantitative assessment of the sources of uncertainty, providing decision-makers with a more comprehensive and reliable basis for risk warning.

[0016] On the other hand, this application also provides a comprehensive road safety assessment system to overcome the interaction effects of risk factors, including: The matrix building module is configured to construct a road risk assessment indicator system based on the target layer, criterion layer, and indicator layer, thereby obtaining a set of risk factors. The module is also configured to construct an interval judgment matrix based on the relative importance of each risk factor in the risk factor set, and convert this interval judgment matrix into a deterministic judgment matrix using the interval median. Furthermore, the module is configured to build a least-squares optimization model based on the deterministic judgment matrix, obtain subjective weights, perform consistency verification on the deterministic judgment matrix based on the subjective weights, and adjust the deterministic judgment matrix based on the verification results. The historical module is electrically connected to the matrix module. The historical module is configured to acquire historical road risk data and obtain objective entropy weights based on the historical road risk data. Based on the subjective weights and objective entropy weights according to preset weight coefficients, the fusion weights are obtained. The fusion weights are used as fuzzy density, and the risk factor interaction coefficients of the fuzzy density are solved based on the fuzzy measure equation. The fuzzy measure function of the fuzzy density is constructed based on the interaction coefficients. The evaluation module, electrically connected to the historical module, is configured to acquire real-time risk factor data of the road to be evaluated, and obtain the risk factor vector of the road to be evaluated based on the real-time risk factor data. The evaluation module is also configured to obtain the comprehensive road risk value of the road to be evaluated based on the fuzzy measure function and the risk factor vector of the road to be evaluated. The evaluation module is further configured to obtain the statistical distribution between the comprehensive road risk value and the fusion weight, and based on the statistical distribution and Monte Carlo sampling, to perform multiple random perturbations on the real-time risk factor data and the fusion weight and repeatedly calculate the fuzzy integral to obtain the probability distribution of the comprehensive road risk value of the road to be evaluated and the confidence interval of the probability distribution of the comprehensive road risk value.

[0017] It is understood that the road safety comprehensive assessment method and system for overcoming the interaction effect of risk factors in the above embodiments of this application have the same beneficial effects, and will not be described again. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating a comprehensive road safety assessment method for overcoming the interaction effects of risk factors, provided in an embodiment of the present invention; Figure 2 This is a hierarchical structure diagram of the road risk assessment index system provided in the embodiments of the present invention; Figure 3 An improved fusion weight calculation flowchart provided for embodiments of the present invention; Figure 4 This is a schematic diagram of the fuzzy integral fusion model architecture provided in an embodiment of the present invention; Figure 5 This is a functional block diagram of a comprehensive road safety assessment system for overcoming the interaction effects of risk factors, provided as an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] Goal Level: This represents the overall goal that the entire evaluation system ultimately aims to achieve, and is the core evaluation result that all evaluation steps are geared towards.

[0021] Criteria Level: This level is used to structurally decompose the target level into the main evaluation dimensions or categories that affect the overall target. It is an intermediate level that connects the overall target with specific indicators.

[0022] Indicator Level: This layer is used to further refine the criteria layer, forming specific risk factor indicators that are quantifiable and observable. It is the most basic and directly collectable data layer in the assessment system.

[0023] Risk Factor: A quantifiable element or characteristic that can affect the level of road safety, used to reflect the intensity and changes of potential risk sources.

[0024] Judgment Matrix: A matrix structure used to compare the relative importance of each evaluation element in pairs, reflecting the weight relationship between different elements through expert scoring results.

[0025] Least Squares Optimization Model: This refers to a mathematical model that finds the optimal weight vector by minimizing the sum of squared deviations of the judgment matrix, and is used to obtain weight results that are most consistent with expert judgment.

[0026] Objective Entropy Weight: This refers to the objective weight calculated based on the difference and information content of each indicator data in the sample, and is used to reflect the magnitude of the indicator's contribution to the uncertainty of the system.

[0027] Fuzzy density: In fuzzy measure theory, it refers to the parameter used to characterize the importance and interaction strength of a single risk factor or combination of factors, and is the basis for constructing fuzzy measure functions.

[0028] Fuzzy Measure Equation: A mathematical equation used to solve for the fuzzy measure values ​​of each risk factor and its combination based on the fuzzy density, in order to characterize the nonlinear coupling relationship and overall importance between factors.

[0029] Risk Factor Data: refers to the raw or processed numerical information used to characterize the current state of various risk factors of a road, and is the basic input for calculating the comprehensive risk value of a road.

[0030] Confidence interval: refers to a numerical range obtained based on statistical inference, used to describe the range in which the comprehensive risk value may fall at a given confidence level, thus reflecting the degree of uncertainty of the result.

[0031] The mechanism of road safety influence refers to the principles and intrinsic relationships by which various road risk factors affect road safety levels through structural characteristics, environmental conditions, and traffic behavior.

[0032] Interval Midpoint: The average of the two endpoints of a numerical interval. It is used to convert the interval judgment result into a deterministic value for subsequent calculation and analysis.

[0033] Dynamic Consistency Index: This index measures the consistency of weight changes in a judgment matrix over multiple rounds of adjustments or dynamic updates, ensuring that the weight system remains stable and reasonable in a dynamic environment.

[0034] Relative Contribution: refers to the proportion of importance of a certain risk factor in the overall assessment results, and is used to measure the degree of influence of the factor on the overall risk level.

[0035] The Choquet Fuzzy Integral Formula is a formula for nonlinearly aggregating a risk factor vector f given a fuzzy measure μ. Its mathematical expression is as follows: (Calculation formula) f(1)≤f(2)≤⋯f(n) represents the ascending order of risk factors; A(i)={(i),(i+1),…,(n)} represents the set consisting of the i-th factor and all subsequent factors after the ordering; μ(A(i)) is the fuzzy measure of the set A(i); f(0)=0. This formula is used to comprehensively consider the interaction effects between risk factors, achieving a more realistic nonlinear risk fusion than linear weighting.

[0036] Sample pair: In data analysis or model training, a pair of related samples is used for comparison, correlation analysis, or to build judgment relationships, such as paired comparison samples used to build a judgment matrix.

[0037] Empirical distribution estimation is a nonparametric method that directly estimates the distribution of a random variable using sample data, without relying on any specific distribution assumptions.

[0038] Kernel density estimation is a nonparametric statistical method that estimates the true probability density of data by smoothing the samples using a kernel function without requiring a predefined distribution.

[0039] Gaussian Mixture Model (GMM): A statistical model that treats data as a mixture of multiple Gaussian distributions, used to characterize complex distribution structures and perform clustering, density estimation, or probability prediction.

[0040] Marginal distribution: refers to the probability distribution obtained when considering one (or some) dimensions of a multidimensional random variable in isolation. It is a univariate or low-dimensional probability distribution obtained by integrating or summing over the other dimensions.

[0041] Joint distribution: refers to the probability distribution when two or more random variables take values ​​together in the same probability space. It is used to describe the overall dependency relationship and joint probability of occurrence between variables.

[0042] A statistical model is a model that uses mathematical structures and probabilistic laws to describe and infer the data generation process, and is used to characterize the relationship between variables, predict results, or explain data behavior.

[0043] Alignment / Geometric Feature: refers to the geometric orientation attributes of a road in space, such as curve radius, longitudinal slope, and horizontal curve change rate, which are used to reflect the impact of road geometry on driving safety.

[0044] Fuzzy Integral Computation: refers to a method based on fuzzy measures to perform nonlinear weighted aggregation of multiple risk factors, used to characterize the interaction effects between factors and obtain comprehensive evaluation results.

[0045] Random perturbation refers to the introduction of random variations into system inputs or parameters to simulate uncertainty, test model robustness, or perform probability distribution analysis.

[0046] Monte Carlo sampling is a method that uses a large number of random samples to simulate and repeatedly compute to estimate the probability distribution, expected value, or uncertainty index of a complex system.

[0047] like Figures 1-2 As shown in some embodiments of this application, this embodiment provides a comprehensive road safety assessment method to overcome the interaction effects of risk factors, including: Step S100: Construct a road risk assessment index system based on the target layer, criterion layer and indicator layer to obtain a risk factor set. Construct an interval judgment matrix according to the relative importance of each risk factor in the risk factor set, and convert the interval judgment matrix into a deterministic judgment matrix using the midpoint of the interval.

[0048] Specifically, the road risk assessment indicator system is constructed based on the target layer, criterion layer, and indicator layer. When obtaining the risk factor set, the following steps are taken: hierarchical modeling of road risks is performed according to the hierarchical analysis method structure of the target layer, criterion layer, and indicator layer; the road structure, traffic facilities, alignment features, and road environment in the preset road configuration are obtained, and the assessment dimensions in the preset road configuration are extracted based on the road structure, traffic facilities, alignment features, and road environment; each risk indicator in each dimension of the assessment dimension is evaluated based on the road safety impact mechanism, and the key risk indicators in each dimension of the assessment dimension are identified; the key risk indicators of each dimension are standardized and aggregated, and the set is determined as the risk factor set.

[0049] Specifically, when constructing an interval judgment matrix based on the relative importance of each risk factor in the risk factor set, and converting the interval judgment matrix into a deterministic judgment matrix using the interval median, the process includes: performing pairwise comparisons of each risk factor in the risk factor set based on the interval number, determining the comparison results as interval elements, and constructing an interval judgment matrix based on each interval element; obtaining the interval median of each interval element based on the relationship between the upper and lower bounds of each interval element; replacing each interval element in the interval judgment matrix with the interval median of each interval element, and determining the interval judgment matrix after replacing each interval element as a deterministic judgment matrix.

[0050] The preset road can be based on industry standards such as the "Technical Standards for Highway Engineering" (JTG B01-2014) and the "Design Specifications for Highway Traffic Safety Facilities" (JTG D81-2017).

[0051] Understandably, a hierarchical analysis (AHP) structure, comprising target, criterion, and indicator layers, is used to model road risks hierarchically, thus structuring and stratifying the complex road risk assessment problem. This method, from a top-down perspective, decomposes the overall road safety objective into assessment criteria of different dimensions, and further refines them into quantifiable risk indicators. This results in a clear hierarchical structure for the overall risk assessment system, facilitating systematic analysis and quantitative processing. Secondly, for each assessment dimension, key risk indicators from pre-defined road configurations, such as road structure, traffic facilities, alignment characteristics, and road environment, are obtained through road safety impact mechanisms. These indicators are then standardized and aggregated to form a set of risk factors that can be directly used for calculation. This process unifies complex and multi-source road safety information into quantifiable risk factors, providing foundational data for subsequent weight calculation and risk fusion. Finally, in the weight acquisition stage, an interval judgment matrix is ​​used to compare risk factors pairwise to express the fuzziness or uncertainty of the relative importance of different factors. By calculating the median of each interval element, the interval judgment matrix is ​​transformed into a deterministic judgment matrix, achieving a smooth transition from fuzzy evaluation to deterministic evaluation, and providing an operable mathematical basis for further weight calculation and consistency verification.

[0052] Step S200: Establish a least squares optimization model based on the deterministic judgment matrix, obtain subjective weights, perform consistency verification on the deterministic judgment matrix based on the subjective weights, and adjust the deterministic judgment matrix based on the verification results.

[0053] Specifically, when establishing a least squares optimization model based on the deterministic judgment matrix, obtaining subjective weights, and performing consistency verification on the deterministic judgment matrix based on these subjective weights, the process includes: obtaining the deviation values ​​of each element in the deterministic judgment matrix between each element and its corresponding weight ratio, and establishing an objective function for the sum of squared errors of each element based on these deviation values; establishing a least squares optimization model based on the sum of squared errors of each element, the objective function, the non-negativity constraint of the weights, and the normalization constraint of the weights; obtaining the subjective weight vector that minimizes the deviation of the judgment matrix in the deterministic judgment matrix by fitting and solving the least squares optimization model; determining the dynamic consistency index of the deterministic judgment matrix based on the relationship between the subjective weight vector and the deterministic judgment matrix, and determining whether to adjust the deterministic judgment matrix based on the relationship between the dynamic consistency index and the configured preset consistency index; when the dynamic consistency index is greater than or equal to the preset consistency index, it is determined that the deterministic judgment matrix will not be adjusted; when the dynamic consistency index is less than the preset consistency index, it is determined that the deterministic judgment matrix will be adjusted.

[0054] Specifically, when the dynamic consistency index is less than the preset consistency index, the deterministic judgment matrix is ​​adjusted, including: obtaining the ideal ratio matrix of the deterministic judgment matrix based on subjective weights; obtaining the element difference between each element of the deterministic judgment matrix and the corresponding ideal ratio in the ideal ratio matrix, mapping the element difference to the corresponding adjustment coefficient, and adjusting the element according to the adjustment coefficient, wherein: the adjustment coefficient is configured such that when the element difference is greater than the configured preset element difference threshold, the adjustment coefficient is less than 1; when the element difference is less than the configured preset element difference threshold, the adjustment coefficient is greater than 1; when the element difference is equal to the configured preset element difference threshold, the adjustment coefficient is 1; updating the deterministic judgment matrix according to the adjusted elements, and synchronously updating the inverse elements of each adjusted element in the deterministic judgment matrix; constructing a least squares optimization model for the second time based on the adjusted deterministic judgment matrix, and obtaining subjective weights for the second time based on the constructed least squares optimization model; performing a second consistency index verification on the adjusted deterministic judgment matrix based on the second obtained subjective weights, until the dynamic consistency index is greater than or equal to the preset consistency index.

[0055] Understandably, the subjective weights of risk factors are obtained by processing the deterministic judgment matrix using a least squares optimization model. Specifically, the deviation between each element in the judgment matrix and its corresponding weight ratio is used as the optimization objective. An error sum of squares objective function is constructed, and the optimal weight vector is solved under the constraints of weight non-negativity and normalization. This method can scientifically quantify the relative importance of each risk factor while preserving expert subjective judgment information, making the weight solution mathematically optimal and objective. Secondly, to ensure the consistency between the weights and the judgment matrix, a dynamic consistency index is introduced to verify the deterministic judgment matrix. By comparing the dynamic consistency index with a preset consistency standard, it is determined whether the judgment matrix needs adjustment. When consistency is insufficient, an ideal ratio matrix is ​​constructed, and the difference between the matrix elements and the ideal values ​​is calculated. An adjustment coefficient is generated to quantitatively correct the matrix elements, and the inverse elements are updated simultaneously. This process achieves dynamic optimization of the judgment matrix, keeping the subjective weights and the judgment matrix coordinated and reducing inconsistencies caused by subjective bias. Finally, the adjusted judgment matrix is ​​input twice into the least squares optimization model to resolve the subjective weights and perform consistency verification again, forming a closed-loop iteration until the dynamic consistency index reaches the preset standard. This iterative optimization mechanism can adaptively correct the judgment matrix, ensuring that the final weights conform to both expert judgment and consistency requirements, thereby improving the reliability of weight allocation and the stability of the evaluation system.

[0056] Step S300: Obtain historical road risk data, and obtain objective entropy weights based on the historical road risk data. Obtain fusion weights based on subjective weights and objective entropy weights according to preset weight coefficients. Use the fusion weights as fuzzy density, and solve the risk factor interaction coefficients of fuzzy density based on the fuzzy measure equation. Construct the fuzzy measure function of fuzzy density based on the interaction coefficients.

[0057] Specifically, the process of acquiring historical road risk data and obtaining objective entropy weights based on this data, and then obtaining fusion weights by combining subjective weights and objective entropy weights with preset weighting coefficients, includes: acquiring historical road risk data for each time period and performing normalization processing on each historical data; constructing a standardized matrix based on the normalized historical data; obtaining the probability distribution of each risk factor based on the standardized matrix and obtaining the entropy value of each risk factor based on the probability distribution; obtaining the objective entropy weight of each risk factor based on the entropy value, and linearly weighting the subjective weights and objective entropy weights according to preset weighting fusion coefficients to obtain a weight vector, wherein the weighting fusion coefficients are configured to be set based on the relative contribution of subjective weights and objective entropy weights; and processing the weight vectors according to normalization to obtain the fusion weights.

[0058] Specifically, when constructing a fuzzy measure function for fuzzy density based on the interaction coefficients of risk factors using the fusion weights as fuzzy densities and the fuzzy measure equation, the process includes: using the fusion weights as the marginal fuzzy densities of each risk factor to construct an initial fuzzy density vector; constructing a fuzzy measure interaction equation based on the marginal fuzzy densities of each risk factor; solving for the fuzzy measure parameters based on the fuzzy measure interaction equation to obtain the interaction coefficients between each risk factor; and obtaining the fuzzy density values ​​of any subset of risk factors according to the fuzzy measure generation formula based on the marginal fuzzy densities and interaction coefficients to construct a fuzzy measure function.

[0059] Understandably, objective entropy weights for each risk factor are obtained through historical road risk data. Specifically, historical data for each time period is normalized to construct a standardized matrix, and the probability distribution and entropy value of each risk factor are calculated. The entropy value objectively reflects the information content and uncertainty of each risk factor in the historical data, thus obtaining the objective weight of each factor. This process introduces objective patterns from historical data into the evaluation system, compensating for potential biases arising from relying solely on subjective expert judgment. Secondly, subjective weights and objective entropy weights are fused using a preset weight fusion coefficient to achieve linear weighting, resulting in the final fused weight vector. The fusion coefficient can be set based on the relative contribution of subjective weights and objective entropy weights, ensuring that the final weights consider both expert experience and the objective information of historical data, thereby improving the scientific rigor and accuracy of risk assessment. The fused weights are then normalized to ensure their applicability to subsequent fuzzy measure calculations. Finally, a fuzzy density and risk factor interaction model is constructed based on the fused weights. The fused weights are considered as the marginal fuzzy density of each risk factor, and the interaction coefficients between risk factors are solved using fuzzy measure interaction equations. These interaction coefficients quantify the interactions between different risk factors. For example, the impact of certain combinations of risk factors on road safety is far greater than the linear superposition of their individual effects, thus reflecting the complex risk interaction effects in real road environments. Finally, using marginal fuzzy density and interaction coefficients, the fuzzy density value of any subset of risk factors is calculated through a fuzzy measure generation formula, constructing a complete fuzzy measure function. This function can nonlinearly integrate various risk factors and their interaction effects, providing a foundation for subsequent fuzzy integral calculations of road risk values, enabling the assessment results to more accurately reflect the overall risk level of road safety.

[0060] Step S400: Obtain real-time risk factor data of the road to be evaluated; obtain the risk factor vector of the road to be evaluated based on the real-time risk factor data; and obtain the comprehensive road risk value of the road to be evaluated based on the fuzzy measure function and the risk factor vector of the road to be evaluated.

[0061] Specifically, the process of acquiring real-time risk factor data for the road to be evaluated, obtaining risk factor vectors for the road to be evaluated based on the real-time risk factor data, and obtaining the comprehensive road risk value for the road to be evaluated based on the fuzzy measure function and the risk factor vectors of the road to be evaluated includes: imputing missing values, filtering noise, and removing outliers from the real-time risk factor data of the road to be evaluated to obtain a risk factor dataset; normalizing the risk factor dataset and combining the normalized risk factors according to a preset index order to form the real-time risk factor vector of the road to be evaluated; calculating the incremental contribution value of each item in the real-time risk factor vector based on the fuzzy measure function and the Choquet fuzzy integral calculation formula; and summing the incremental contribution values ​​to obtain the comprehensive road risk value of the road to be evaluated.

[0062] Understandably, for the real-time risk factor data of the road to be evaluated, data preprocessing methods are used to impute missing values, filter noise, and remove outliers from the original data, forming a reliable risk factor dataset. This step ensures the integrity and accuracy of the input data, providing a robust data foundation for subsequent risk calculations and effectively avoiding evaluation biases caused by data quality issues. Secondly, the processed risk factor data is normalized and vectorized, combining each risk factor according to a preset index order to form a real-time risk factor vector for the road to be evaluated. This step achieves a unified representation of multi-dimensional heterogeneous risk factors, enabling comprehensive calculation of each index under the same dimension and providing standardized input for fuzzy integral processing. Then, fuzzy measure functions and Choquet fuzzy integrals are used to calculate the risk factor vector item by item. Specifically, by calculating the incremental contribution value of each risk factor and considering the interaction relationships between risk factors, nonlinear weighted aggregation of risk factors is achieved. This method can effectively reflect the comprehensive impact of different risk factors and their combinations on road safety, avoiding the problem of traditional linear weighting methods ignoring interaction effects. Finally, all incremental contribution values ​​are summed to obtain a comprehensive road risk value, which serves as a quantitative indicator of the overall safety status of the road to be evaluated. This technology combines real-time data, fuzzy measures, and Choquet fuzzy integrals to achieve a dynamic and nonlinear comprehensive evaluation of multidimensional risk factors and their interactions, providing a scientific and reliable basis for decision-making in real-time road safety monitoring and risk management.

[0063] Step S500: Obtain the statistical distribution between the comprehensive road risk value and the fusion weight, and based on the statistical distribution and Monte Carlo sampling, perform multiple random perturbations on the real-time risk factor data and the fusion weight, and repeatedly calculate the fuzzy integral to obtain the probability distribution of the comprehensive road risk value of the road to be evaluated and the confidence interval of the probability distribution of the comprehensive road risk value.

[0064] Specifically, obtaining the statistical distribution between the comprehensive road risk value and the fusion weight includes: obtaining sample pairs of fusion weight and comprehensive road risk value within several historical assessment periods, cleaning and aligning the sample pairs according to time order, and constructing a joint sample matrix composed of fusion weight and comprehensive road risk value; based on empirical distribution estimation, kernel density estimation, or Gaussian mixture model, determining the marginal distribution and joint distribution between fusion weight and comprehensive road risk value in the joint sample matrix, and establishing a statistical model based on the marginal distribution and joint distribution between fusion weight and comprehensive road risk value; obtaining the mean, variance, covariance, and correlation structure of fusion weight and comprehensive road risk value based on the statistical model; and constructing the statistical distribution between comprehensive road risk value and fusion weight based on the mean, variance, covariance, and correlation structure of fusion weight and comprehensive road risk value.

[0065] Understandably, a joint sample matrix is ​​constructed using historical assessment data. This involves cleaning and aligning sample pairs of fusion weights and comprehensive road risk values ​​from several historical assessment periods in chronological order, forming a joint sample matrix containing multidimensional variable relationships. This step structures the historical observation data, enabling subsequent statistical analysis to fully utilize past assessment information and reflect the actual correlation characteristics between fusion weights and risk values. Secondly, based on the joint sample matrix, empirical distribution estimation, kernel density estimation, or Gaussian mixture models are used to model the marginal and joint distributions of fusion weights and comprehensive road risk values, constructing a statistical model. These non-parametric or mixture model methods can characterize the complex probability distribution characteristics and variable correlations between fusion weights and comprehensive road risk values, thus providing a reliable mathematical foundation for risk uncertainty analysis. Subsequently, the mean, variance, covariance, and correlation structure are calculated based on the statistical model, forming the statistical distribution of fusion weights and comprehensive road risk values. This statistical distribution not only reflects the central trend of the risk values ​​but also characterizes the uncertainty and volatility characteristics of the risk, providing a quantitative basis for subsequent probabilistic inference. Finally, using Monte Carlo sampling and random perturbation methods, the real-time risk factor data and fusion weights were subjected to multiple random perturbations, and fuzzy integral calculations were repeatedly performed to obtain the probability distribution and confidence interval of the comprehensive road risk value. This step, by simulating uncertainty and risk fluctuations, achieves a probabilistic description of road safety risks, ensuring that the assessment results not only reflect the magnitude of the risk but also include its uncertainty and confidence level, providing a more scientific and robust basis for road safety management and decision-making.

[0066] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principle of this invention is further explained below in conjunction with a specific application scenario.

[0067] When establishing a road risk assessment index system and determining the integration weights, a three-tiered assessment framework of "target layer - criterion layer - indicator layer" is adopted, and the hierarchical relationship between each layer is clarified based on the principles of the analytic hierarchy process (AHP). Specifically, the target layer should comprehensively measure the overall road safety risk level; the criterion layer connects the target layer and the indicator layer, and, in conjunction with industry standards such as the *Highway Engineering Technical Standards* (JTG B01-2014) and the *Design Specifications for Highway Traffic Safety Facilities* (JTG D81-2017), selects several core dimensions that play a decisive role in road risk; the indicator layer is responsible for quantifying the risk level of each dimension in the criterion layer and is the direct data source for risk assessment (e.g., ...). Figure 3 (As shown). After constructing the three-level indicator system, the contribution of each level of indicators is quantified using an improved analytic hierarchy process (AHP). Traditional AHP suffers from strong subjectivity and rigid consistency checks, with a structure as shown... Figure 4 As shown, this module has been innovated and improved in the following aspects.

[0068] In traditional analytic hierarchy process (AHP), experts need to provide accurate numerical judgments, which often fail to accurately reflect the uncertainty in their thought process. To address this issue, this invention uses interval numbers to construct a judgment matrix, better representing the ambiguity of expert judgments, improving the robustness of weight calculations, and making the results more robust and stable. The interval judgment matrix can be represented as:

[0069] in, To determine the order of a matrix, we indicate that the element in the (i)th row and (j)th column of the matrix is ​​a closed interval; This is the lower bound of the interval; This is the upper bound of the interval; This indicates that the matrix has m rows and n columns.

[0070] Secondly, this invention employs the least squares method to optimize the final weight allocation, transforming the weights into a planning problem seeking the "optimal fit." This solves the fundamental difficulty in traditional methods, where inconsistent judgment matrices prevent the determination of weights or result in unreasonable weights. The optimization model is as follows:

[0071] in, To determine the elements of a matrix; Let be the weight ratio of the i-th indicator to the j-th indicator; the constraint condition is: This is a normalization constraint for the weights; Positive constraints on the weights. By solving this optimization problem, a scientifically sound final weight vector is obtained.

[0072] This invention designs a dynamic consistency check mechanism to fundamentally solve the problem of rigid consistency check processes in traditional analytic hierarchy process (AHP). The formula is as follows:

[0073] in, It is a static consistency indicator; It is the weighted standard deviation; It is an adjustment factor; It is a dynamic consistency indicator, when If the value is less than 0.1, the test passes; otherwise, the judgment matrix is ​​automatically adjusted until the consistency requirement is met.

[0074] Furthermore, to overcome the limitations of subjective judgment and achieve a complementary advantage between expert experience and data objectivity, this invention employs the entropy weight method to correct subjective weights. The entropy weight method is based on the principle that "the greater the dispersion of indicator data, the stronger its ability to distinguish the evaluation object, and the higher its weight should be," extracting the objective importance of each indicator from the actually collected multidimensional risk factor data. The formula is as follows:

[0075] in, It is a weighting coefficient used to balance the influence of subjective and objective weights, and is usually taken as 0.5-0.7; Subjective weighting; These are the objective weights calculated using the entropy weight method; It is the final fusion weight.

[0076] Furthermore, through the above steps, a complete risk indicator system and its fusion weight vector can be obtained. This weight vector not only serves as the fuzzy density input for the stage-two fuzzy integral calculation, but can also be used independently for road safety factor sensitivity analysis, providing a basis for subsequent road safety risk management.

[0077] Furthermore, to achieve accurate calculation of the comprehensive road risk value, this invention proposes a nonlinear fusion mechanism based on an improved fuzzy integral fusion model to overcome the limitation of traditional linear weighted models that ignore the dynamic interaction of risk factors. This mechanism uses multidimensional road data and the fusion weights obtained in Stage 1 as inputs to achieve fuzzy integral fusion of risk factors and calculation of the comprehensive road risk value.

[0078] Furthermore, in traditional fuzzy analysis, the setting of fuzzy density largely relies on manual setting or experience, which is highly subjective. Therefore, this method uses the fusion weights of the risk factors determined in stage one as the fuzzy density of the fuzzy integral, thereby achieving a seamless transition from the analytic hierarchy process (AHP) to the fuzzy integral fusion model in terms of model structure.

[0079] First, establish a fuzzy measurement system:

[0080] in, For the first The fusion weights of the risk factors satisfy... .

[0081] This approach allows fuzzy density to have a clear physical meaning, avoiding the problem of setting fuzzy density in traditional methods, and ensuring that the fuzzy fusion process has theoretical consistency and comparability at the weight level.

[0082] Furthermore, complex correlations often exist between road risk factors. For example, when "sharp bend + water accumulation" coexist, the combined risk is far higher than the result of simply adding the two together. To accurately characterize this interactive property, this invention introduces a fuzzy measure equation:

[0083] Wherein, λ is the interaction coefficient, reflecting the direction and strength of the interaction between risk factors. When λ = 1, it indicates a positive correlation (risk superposition effect); when λ = 2, it indicates a negative correlation (risk offsetting effect); and when λ = 3, it indicates an independent relationship.

[0084] Furthermore, based on the fuzzy measurement system obtained in the second step, this invention uses fuzzy integrals to achieve nonlinear fusion calculation of the comprehensive risk value. First, the standardized values ​​of various risk factors are arranged in ascending order as follows: Its overall risk value Represented as:

[0085] In actual calculations, discrete fuzzy integrals are used:

[0086] in, .

[0087] This nonlinear integral calculation method incorporates the ranking and weight of each road risk factor into the calculation process through a fuzzy measure function, thereby achieving a nonlinear comprehensive assessment of multidimensional risk factors.

[0088] Furthermore, to ensure the reliability of the fusion calculation, this method employs Monte Carlo simulation to quantify uncertainty. By repeatedly sampling the input data and weights within their possible distribution ranges and repeatedly calculating the comprehensive risk value, the final output risk value and its confidence interval provide an intuitive reference regarding the reliability of the evaluation results. Specifically:

[0089]

[0090]

[0091] Through multiple simulations, the probability distribution and confidence interval of the risk value can be obtained, thereby quantifying the stability and reliability of the assessment results. The final output of Stage Two is a comprehensive road risk value in the interval [0,1].

[0092] In the above embodiments, subjective weights obtained by utilizing interval judgment matrices and least squares optimization models enable expert experience to participate in weight construction in a structured and verifiable manner, and improve the reliability of the judgment matrix through consistency verification and adjustment. Simultaneously, objective entropy weights extracted from historical risk data can accurately reflect the information content differences of each risk factor in historical samples, preventing excessive reliance on human experience in weight allocation, thus constructing a fusion weight system that balances subjective perception and objective data. Secondly, by introducing fuzzy density and fuzzy measure functions, the interaction coefficients between risk factors are solved, overcoming the limitation of traditional linear weighting in expressing risk coupling relationships. This effectively characterizes risk superposition effects such as "sharp bend + water accumulation," enabling the comprehensive risk value to accurately reflect the nonlinear gain or inhibition relationships between multiple factors, improving the authenticity and sensitivity of road risk assessment results. Finally, by constructing statistical distributions and performing perturbation analysis on real-time risk factors and fusion weights using Monte Carlo sampling, the probability distribution and confidence interval of the comprehensive road risk value can be obtained, thereby elevating risk assessment from a "single value" to a "probabilistic result." This mechanism not only enhances the robustness and credibility of the assessment results, but also enables the quantitative assessment of the sources of uncertainty, providing decision-makers with a more comprehensive and reliable basis for risk warning.

[0093] In another preferred embodiment based on the above embodiments, such as Figure 5 As shown, this embodiment provides a comprehensive road safety assessment system that overcomes the interactive effects of risk factors, including: a matrix establishment module, a history module, and an assessment module.

[0094] Specifically, the matrix building module is configured to construct a road risk assessment indicator system based on the target layer, criterion layer, and indicator layer, obtaining a set of risk factors; the matrix building module is also configured to construct an interval judgment matrix based on the relative importance of each risk factor in the risk factor set, and convert the interval judgment matrix into a deterministic judgment matrix using the interval median; the matrix building module is further configured to build a least squares optimization model based on the deterministic judgment matrix, obtain subjective weights, and perform consistency verification on the deterministic judgment matrix based on the subjective weights, adjusting the deterministic judgment matrix based on the verification results; the historical module is electrically connected to the matrix building module, and the historical module is configured to acquire historical road risk data, obtain objective entropy weights based on the historical road risk data, and obtain fusion weights based on the subjective weights and objective entropy weights according to preset weight coefficients; based on the fusion... The weights are used as fuzzy density, and the risk factor interaction coefficients of the fuzzy density are solved based on the fuzzy measure equation. The fuzzy measure function of the fuzzy density is constructed based on the interaction coefficients. The evaluation module is electrically connected to the history module. The evaluation module is configured to acquire real-time risk factor data of the road to be evaluated and obtain the risk factor vector of the road to be evaluated based on the real-time risk factor data. The evaluation module is also configured to obtain the comprehensive road risk value of the road to be evaluated based on the fuzzy measure function and the risk factor vector of the road to be evaluated. The evaluation module is also configured to obtain the statistical distribution between the comprehensive road risk value and the fusion weights, and to perform multiple random perturbations on the real-time risk factor data and the fusion weights based on the statistical distribution and Monte Carlo sampling, and to repeatedly calculate the fuzzy integral to obtain the probability distribution of the comprehensive road risk value of the road to be evaluated and the confidence interval of the probability distribution of the comprehensive road risk value.

[0095] It is understood that the road safety comprehensive assessment method and system for overcoming the interaction effect of risk factors in the above embodiments of this application have the same beneficial effects, and will not be described again.

[0096] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0097] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0098] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0099] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A comprehensive road safety assessment method that overcomes the interaction effects of risk factors, characterized in that, include: A road risk assessment index system is constructed based on the target layer, criterion layer, and indicator layer to obtain a set of risk factors. An interval judgment matrix is ​​constructed according to the relative importance of each risk factor in the risk factor set, and the interval judgment matrix is ​​converted into a deterministic judgment matrix using the midpoint of the interval. A least squares optimization model is established based on the deterministic judgment matrix, subjective weights are obtained, and the consistency of the deterministic judgment matrix is ​​checked based on the subjective weights. The deterministic judgment matrix is ​​then adjusted based on the check results. Historical road risk data is acquired, and objective entropy weights are obtained based on the historical road risk data. The fusion weights are obtained by combining subjective weights and objective entropy weights according to preset weight coefficients. The fusion weights are used as the fuzzy density, and the risk factor interaction coefficients of the fuzzy density are solved based on the fuzzy measure equation. The fuzzy measure function of the fuzzy density is then constructed based on the interaction coefficients. Obtain real-time risk factor data of the road to be evaluated, obtain the risk factor vector of the road to be evaluated based on the real-time risk factor data, and obtain the comprehensive road risk value of the road to be evaluated based on the fuzzy measure function and the risk factor vector of the road to be evaluated. The statistical distribution between the comprehensive road risk value and the fusion weight is obtained. Based on the statistical distribution and Monte Carlo sampling, the real-time risk factor data and the fusion weight are subjected to multiple random perturbations and repeated fuzzy integral calculations to obtain the probability distribution of the comprehensive road risk value of the road to be evaluated and the confidence interval of the probability distribution of the comprehensive road risk value.

2. The comprehensive road safety assessment method for overcoming the interaction effects of risk factors as described in claim 1, characterized in that, When constructing a road risk assessment indicator system based on the target layer, criterion layer, and indicator layer, the resulting risk factor set includes: Road risk is modeled hierarchically based on the hierarchical analysis method structure of target layer, criterion layer and indicator layer; Obtain the road structure, traffic facilities, alignment features, and road environment from the preset road configuration, and extract the evaluation dimensions from the preset road configuration based on the road structure, traffic facilities, alignment features, and road environment; Based on the mechanism of road safety impact, the risk indicators of each dimension in the assessment dimension are evaluated, and the key risk indicators of each dimension in the assessment dimension are identified. Key risk indicators from each dimension are standardized and aggregated, and the aggregate is determined as a risk factor set.

3. The comprehensive road safety assessment method for overcoming the interaction effects of risk factors as described in claim 2, characterized in that, When constructing an interval judgment matrix based on the relative importance of each risk factor in the risk factor set, and converting this interval judgment matrix into a deterministic judgment matrix using the midpoint of the interval, the following steps are taken: Based on the interval number, each risk factor in the risk factor set is compared pairwise, the comparison results are determined as interval elements, and an interval judgment matrix is ​​constructed based on each interval element. Based on the relationship between the upper and lower bounds of each interval element, obtain the interval midpoint of each interval element; Replace each interval element in the interval judgment matrix with the midpoint of each interval element, and then determine the interval judgment matrix after replacing each interval element as the deterministic judgment matrix.

4. The comprehensive road safety assessment method for overcoming the interaction effects of risk factors as described in claim 3, characterized in that, When establishing a least-squares optimization model based on the deterministic judgment matrix, obtaining subjective weights, and performing consistency verification on the deterministic judgment matrix based on the subjective weights, the process includes: Obtain the deviation values ​​of each element in the deterministic judgment matrix between each element and its corresponding weight ratio, and establish the objective function of the sum of squared errors of each element based on the deviation values ​​of each element; A least squares optimization model is established based on the squared error of each element, the objective function, the non-negativity constraint of the weights, and the normalization constraint of the weights. Based on the fitting solution of the least squares optimization model, obtain the subjective weight vector that minimizes the bias of the judgment matrix in the deterministic judgment matrix; Based on the relationship between the subjective weight vector and the deterministic judgment matrix, the dynamic consistency index of the deterministic judgment matrix is ​​determined, and based on the relationship between the dynamic consistency index and the configured preset consistency index, it is determined whether the deterministic judgment matrix should be adjusted. When the dynamic consistency index is greater than or equal to the preset consistency index, it is determined that the deterministic judgment matrix will not be adjusted. When the dynamic consistency index is less than the preset consistency index, the deterministic judgment matrix will be adjusted.

5. The comprehensive road safety assessment method for overcoming the interaction effects of risk factors as described in claim 4, characterized in that, When the dynamic consistency index is less than the preset consistency index, the deterministic judgment matrix is ​​adjusted, including: Obtain the ideal ratio matrix of the deterministic judgment matrix based on subjective weights; Obtain the element-wise difference between each element of the deterministic judgment matrix and the corresponding ideal ratio in the ideal ratio matrix, map the element-wise difference to the corresponding adjustment coefficient, and adjust the element according to the adjustment coefficient, where: The adjustment coefficient is configured such that when the element difference is greater than the preset element difference threshold, the adjustment coefficient is determined to be less than 1; when the element difference is less than the preset element difference threshold, the adjustment coefficient is determined to be greater than 1; and when the element difference is equal to the preset element difference threshold, the adjustment coefficient is determined to be 1. The deterministic judgment matrix is ​​updated based on the adjusted elements, and the inverse elements of each adjusted element in the deterministic judgment matrix are updated simultaneously. Based on the adjusted deterministic judgment matrix, a least squares optimization model is constructed for the second time, and subjective weights are obtained for the second time based on the constructed least squares optimization model; The adjusted deterministic judgment matrix is ​​subjected to a second consistency index verification based on the subjective weights obtained in the second step, until the dynamic consistency index is greater than or equal to the preset consistency index.

6. The comprehensive road safety assessment method for overcoming the interaction effects of risk factors as described in claim 5, characterized in that, When acquiring historical road risk data and obtaining objective entropy weights based on this data, and then deriving the fusion weights according to preset weighting coefficients for both subjective and objective entropy weights, the process includes: Historical road risk data for each time period is obtained, and the historical data is normalized. Based on the normalized historical data, a standardized matrix is ​​constructed. The probability distribution of each risk factor is obtained based on the standardized matrix, and the entropy value of each risk factor is obtained based on the probability distribution. The objective entropy weight of each risk factor is obtained based on the entropy value, and the subjective weight and objective entropy weight are linearly weighted according to the preset weight fusion coefficient to obtain the weight vector. The weight fusion coefficient is configured to be set based on the relative contribution of the subjective weight and the objective entropy weight. The weight vector is processed by normalization to obtain the fusion weights.

7. The comprehensive road safety assessment method for overcoming the interaction effects of risk factors as described in claim 6, characterized in that, When using the fusion weights as the fuzzy density, and solving for the risk factor interaction coefficients of the fuzzy density based on the fuzzy measure equation, the fuzzy measure function of the fuzzy density is constructed based on the interaction coefficients, including: The fusion weights are used as the marginal fuzzy densities of each risk factor to construct an initial fuzzy density vector; A fuzzy measure interaction equation is constructed based on the marginal fuzzy density of each risk factor. The fuzzy measure parameters are then solved based on the fuzzy measure interaction equation to obtain the interaction coefficients between each risk factor. Based on the marginal fuzzy density and interaction coefficient, and according to the fuzzy measure generation formula, the fuzzy density value of any subset of risk factors is obtained, and the fuzzy measure function is constructed.

8. The comprehensive road safety assessment method for overcoming the interaction effects of risk factors as described in claim 7, characterized in that, When acquiring real-time risk factor data for the road to be evaluated, obtaining the risk factor vector of the road to be evaluated based on the real-time risk factor data, and obtaining the comprehensive road risk value of the road to be evaluated based on the fuzzy measure function and the risk factor vector of the road to be evaluated, the process includes: Missing values ​​are imputed, noise is filtered out, and outliers are removed from the real-time risk factor data of the road to be evaluated to obtain a risk factor dataset; The risk factor dataset is normalized, and the normalized risk factors are combined according to a preset index order to form a real-time risk factor vector for the road to be evaluated. The incremental contribution value of each item in the real-time risk factor vector is obtained based on the fuzzy measure function and the Choquet fuzzy integral calculation formula. The incremental contribution values ​​are summed to obtain the comprehensive road risk value of the road to be evaluated.

9. The comprehensive road safety assessment method for overcoming the interaction effects of risk factors as described in claim 8, characterized in that, When obtaining the statistical distribution between the comprehensive road risk value and the fusion weight, the following is included: Obtain sample pairs of fusion weights and comprehensive road risk values ​​from several historical assessment periods, and clean and align the sample pairs in chronological order to construct a joint sample matrix composed of fusion weights and comprehensive road risk values; Based on empirical distribution estimation, kernel density estimation, or Gaussian mixture model, the marginal distribution and joint distribution between the fusion weights and the comprehensive road risk value in the joint sample matrix are analyzed. A statistical model is then established based on the marginal distribution and joint distribution between the fusion weights and the comprehensive road risk value. Based on the statistical model, the mean, variance, covariance, and correlation structure of the fusion weights and the comprehensive road risk value are obtained. Based on the mean, variance, covariance, and correlation structure of the fusion weights and the comprehensive road risk value, a statistical distribution between the comprehensive road risk value and the fusion weights is constructed.

10. A comprehensive road safety assessment system for overcoming the interaction effects of risk factors, employing a comprehensive road safety assessment method for overcoming the interaction effects of risk factors as described in any one of claims 1-9, characterized in that, include: The matrix building module is configured to construct a road risk assessment indicator system based on the target layer, criterion layer, and indicator layer, thereby obtaining a set of risk factors. The matrix building module is also configured to construct an interval judgment matrix based on the relative importance of each risk factor in the risk factor set, and convert the interval judgment matrix into a deterministic judgment matrix using the interval median; the matrix building module is also configured to build a least squares optimization model based on the deterministic judgment matrix, obtain subjective weights, perform consistency verification on the deterministic judgment matrix based on the subjective weights, and adjust the deterministic judgment matrix based on the verification results; The historical module is electrically connected to the matrix module. The historical module is configured to acquire historical road risk data, obtain objective entropy weights based on the historical road risk data, and obtain fusion weights based on subjective weights and objective entropy weights according to preset weight coefficients. The fusion weights are used as the fuzzy density, and the risk factor interaction coefficients of the fuzzy density are solved based on the fuzzy measure equation. The fuzzy measure function of the fuzzy density is then constructed based on the interaction coefficients. The evaluation module is electrically connected to the history module. The evaluation module is configured to acquire real-time risk factor data of the road to be evaluated, and to obtain the risk factor vector of the road to be evaluated based on the real-time risk factor data. The evaluation module is also configured to obtain the comprehensive road risk value of the road to be evaluated based on the fuzzy measure function and the risk factor vector of the road to be evaluated. The assessment module is also configured to obtain the statistical distribution between the comprehensive road risk value and the fusion weight, and to perform multiple random perturbations on the real-time risk factor data and the fusion weight based on the statistical distribution and Monte Carlo sampling, and to repeatedly calculate the fuzzy integral to obtain the probability distribution of the comprehensive road risk value of the road to be assessed and the confidence interval of the probability distribution of the comprehensive road risk value.

Images