Navigation accident risk identification and evaluation method based on enhanced rough BWM evaluation method

Abstract

The invention provides a shipping accident risk identification and evaluation method based on an enhanced rough BWM evaluation method. The method comprises the following steps: constructing an identification prediction model by using a Bayesian network based on a neural network; determining the weight of a shipping accident risk factor through decision making; carrying out weight improvement by utilizing ROWA to obtain a weight coefficient of the shipping accident risk factor; constructing a rough decision matrix based on an elastic approximation factor, and evaluating shipping accident risk factors; and determining the relative weight of each shipping accident risk factor by adopting an aggregation evaluation method, converting the risk factors into numerical values, and arranging and grading the risk factors according to the numerical values. According to the method, a plurality of shipping accident risk factors are selected, the priority ranking problem under different standards is researched, a risk identification and evaluation method is fused, common risk factors in the shipping industry are deeply discussed, and effective evasion strategies are provided for different industries, so that the scientificity and practicability of risk management are improved.

Description

Technical Field

[0001] The present invention relates to the field of transportation, and particularly to a method for identifying and evaluating shipping accident risks based on an enhanced rough BWM evaluation method. Background Art

[0002] In the context of globalization today, the shipping industry, as a link connecting the world's continents, plays an important role in promoting the prosperity and development of international trade. Currently, international trade is booming, profoundly affecting the world's economic situation and context. As the world's largest cargo transportation system, the shipping industry undertakes most economic activities. However, maritime transportation faces many uncertain factors. Shipping risks can not only trigger ship accidents but also cause delays in cargo transportation. Therefore, great importance must be attached to safety and security measures.

[0003] Therefore, how to prioritize shipping accident risk factors under different criteria and formulate effective accident avoidance strategies has become an urgent problem to be solved. Summary of the Invention

[0004] In view of the above-mentioned technical problems, a method for identifying and evaluating shipping accident risks based on an enhanced rough BWM evaluation method is provided. The present invention mainly starts from shipping accident risk factors, selects multiple shipping accident risk factors, studies the problem of priority ranking under different criteria, integrates risk identification and evaluation methods, deeply explores common risk factors in the shipping industry, and provides effective avoidance strategies for different industries to improve the scientificity and practicality of risk management.

[0005] The technical means adopted by the present invention are as follows:

[0006] A method for identifying and evaluating shipping accident risks based on an enhanced rough BWM evaluation method, comprising:

[0007] Constructing an identification and prediction model using a Bayesian network based on a neural network;

[0008] Determining the weights of shipping accident risk factors through decision-making;

[0009] Improving the weights using ROWA to obtain the weight coefficients of shipping accident risk factors;

[0010] Constructing a rough decision matrix based on an elastic approximation factor to evaluate shipping accident risk factors;

[0011] Adopting an aggregation evaluation method to determine the relative weights of each shipping accident risk factor, converting the risk factors into numerical values, and arranging and grading the risk factors based on the numerical values.

[0012] Furthermore, the construction of the recognition and prediction model using the Bayesian network based on neural network specifically includes:

[0013] Construct a static Bayesian network using a rule-based pruning strategy to infer the probability associated with a set of Boolean variables. If the variable V represents a node, then (j = 1, 2, …, m) represents the set of parent nodes of the variable V. If V is a vector containing multiple random variables, the joint probability distribution of V is:

[0014]

[0015] Refine the results trained by the Bayesian model based on neural network through the recursive method. Regarding the marginal probability, if it is a node without parent nodes, obtain the marginal probability; if it is a node with parent nodes, obtain the parent nodes through the recursive method.

[0016] The Boolean variables provided by the conditional probability table CPT are:

[0017] P(Y j = True|Pa(Y j ))

[0018] where Y j represents a random variable, and Pa(Y j ) represents the parent node of Y j .

[0019] Express the complement value of the Boolean variable as:

[0020] P(Y j = False|Pa(Y j )) = 1 - P(Y j = True|Pa(Y j ))).

[0021] Furthermore, the risk factors of the shipping accident include: the number of affected areas D, the number of large ship losses L, the number of ship accidents A, and the total number of ship losses T.

[0022] Furthermore, the determination of the weights of the risk factors of the shipping accident through decision-making specifically includes:

[0023] Represent the standard set as C = {D, T, L, A}, and determine the best standard S B and the worst standard S W :

[0024]

[0025]

[0026] where Denote the scoring vector of the \(i\)-th decision maker under the best criterion \(S\) B . Denote the scoring vector of the \(i\)-th decision maker under the best criterion \(S\) W . Denote the score of the \(i\)-th decision maker's judgment on criterion \(j\) under the best criterion \(S\) B , Denote the score of the \(i\)-th decision maker's judgment on criterion \(j\) under the worst criterion \(S\) W ;

[0027] Define the distances between the conditional judgment sets of the best criterion \(S\) B and the worst criterion \(S\) W as follows:

[0028]

[0029] where \(d\) Bq denotes the distance of the conditional judgment set of vector \(B\), and \(d\) Wq denotes the distance of the conditional judgment set of vector \(W\).

[0030] Introduce the elastic approximation factor \(k\) to represent the clarity of the decision, and transform the judgment of sensitivity into EARN. The calculation process is as follows:

[0031]

[0032] where, is the lower bound approximation of , is the upper bound approximation of ; \(U\) represents the universal set, denotes the scoring vector that satisfies the condition, and \(H\) j denotes the scoring vector of the \(j\)-th criterion.

[0033] When \(k = 0\), the decision is clear, When \(k\in(0,1)\), there is ambiguity in the decision, and the degree of ambiguity is determined by the elastic approximation factor and the distance \(kd\) of the vector conditional judgment set; when \(k = 1\), the decision is fuzzy, then

[0034] Adopt the ordered weighted average (OWA) method to transform the clear value judgment set into the form of rough numbers;

[0035]

[0036] where, denotes the lower boundary, denotes the upper boundary, and \(g\) and \(f\) are respectively and The quantity of matter of the object in the middle; {o(1), o(2), …, o(g)} is a permutation of {(1), (2), …, (g)}; {o(1), o(2), …, o(f)} is a permutation of {(1), (2), …, (f)}, w v represents the corresponding v-th OWA weight, where w v ≥ 0,

[0037] The form of EARN is:

[0038]

[0039] Convert the form of EARN of the best criterion S B and the worst criterion S W into:

[0040]

[0041] Furthermore, the weight of the risk factors of shipping accidents is obtained by improving the weights using ROWA, specifically including:

[0042] The ROWA operator of the best criterion S B is expressed as:

[0043]

[0044] where, w a represents the a-th ROWA weight, 0 ≤ w a ≤ 1, EARN k (P Bq ) is restricted by , representing the lower and upper bounds of EARN k (P Bq ), respectively;

[0045] Calculate the rough fuzzy boundary interval of RBI Bq which refers to P Bq :

[0046]

[0047] The ROWA operator of the worst criterion S W can be expressed as:

[0048]

[0049] where, w b represents the b-th ROWA weight, 0 ≤ w b ≤ 1, EARN k(P qW ) is restricted by and represents the lower and upper bounds of EARN k (P qW ) respectively;

[0050] Let ROWA:

[0051]

[0052] where ω = (ω1, ω2, …, ω l ) is the weighted vector associated with the function ROWA, ω s ∈ [0, 1], s ∈ {1, 2, …, l}, b c is the c-th largest element in the data (f1, f2, …, f v , …, f l ), and R is the set of real numbers;

[0053] Let ω = (ω1, ω2, …, ω l ) be the weight vector of the ROWA operator, and define ω v as:

[0054]

[0055] where μ l is the mathematical expectation weighted by l and σ l l is the standard deviation obtained by l in σ l and the weight . The definitions of μ l and σ l are as follows:

[0056]

[0057] For the evaluated decision data f1, f2, …, f v , …, f l , use the obtained mean μ l and variance σ l to perform standardization processing to obtain (α1, α2, …, α v , …, α l ):

[0058]

[0059] Solve the values of θ(α v ) at α1, α2, …, α v , …, α l respectively, which are β1, β2, …, β v , …, β l ,

[0060]

[0061] Since ω k ∈ [0, 1] and perform unit normalization:

[0062]

[0063] obtain the weight vectors ω1, ω2, …, ω v , …, ω l ;

[0064] For each shipping accident risk factor, satisfy and calculate the weight of each criterion:

[0065] min λ

[0066] w v ≥ 0, for all v.

[0067] Where represents the weight of the best criterion S B ; is the weight of the worst criterion S W ; represents the average value of EARN k (P B ); represents the average value of EARN k (P W ).

[0068] Furthermore, construct a rough decision matrix based on the elastic approximation factor to evaluate the shipping accident risk factors, specifically including:

[0069] For each independent shipping accident risk factor m, form a scoring set, denoted as Where represents the specific score for the risk factor m in the d-th decision;

[0070]

[0071] Where, C d represents the d-th rough decision matrix, m = 1, 2, …, x; n = 1, 2, …, y, is the score for the shipping accident risk factor m under the n-th criterion Sn in the d-th decision; convert the rough numerical value into a definite fixed value:

[0072]

[0073] Among them, and are respectively the maximum and minimum values, and the EARN formula of c mn is as follows:

[0074]

[0075] Among them, and are respectively the upper and lower bounds of the boundary, and the rough decision matrix EARN under the elastic approximation factor k is obtained k :

[0076]

[0077] For the cost and benefit criteria, the normalization method of rough boundary is adopted to convert the rough values into a standardized expression form:

[0078]

[0079] Among them, is the standardized expression form.

[0080] Calculate the sum of the weighted comparability sequences of risk factors and the sum of the weighted comparability sequences of weight power exponents:

[0081]

[0082] Among them, S m represents the weighted comparability sequence, and E m represents the weighted comparability sequence of weight power exponents; w n represents the optimal weight obtained by the BWM algorithm.

[0083] Furthermore, the relative weights of each shipping accident risk factor are determined by using the aggregation evaluation method, and the risk factors are converted into numerical values, specifically including:

[0084] Determine the relative weights of each shipping accident risk factor by using the aggregation evaluation method:

[0085]

[0086]

[0087] Among them, θ is the empirical coefficient;

[0088] Determine the importance level of the shipping accident risk factors:

[0089]

[0090] Convert the fuzzy risk factors of shipping accidents into numerical representations, and arrange or classify the risk factors in an orderly manner based on these numerical values.

[0091]

[0092] Among them, ρ i is the coefficient of is a m crisp set of

[0093] Compared with the prior art, the present invention has the following advantages:

[0094] The present invention constructs a closed-loop research model for risk identification, assessment, and management. This research breakthroughly constructs a comprehensive and closed-loop shipping risk identification, assessment, and management framework. This framework not only systematically integrates risk identification, assessment, and the formulation of subsequent management strategies, but also deeply analyzes various shipping risks from multiple dimensions and perspectives. Through this closed-loop model, researchers can more comprehensively understand the internal relationships and influence mechanisms of various risk factors, thereby providing important and in-depth insights for risk prevention in the shipping industry. This comprehensive research method helps shipping companies, ship companies, ports, and other related industries more effectively identify potential risks, formulate targeted countermeasures, and thus enhance the safety and stability of the entire shipping system.

[0095] The present invention uses the static Bayesian network algorithm for risk criterion identification. In the risk identification stage, this research introduces the static Bayesian network (SBN) algorithm. This algorithm describes the dependence relationships between risk variables by constructing a directed acyclic graph and uses a conditional probability table (CPT) to express the probability distributions between these variables. Compared with traditional risk identification methods, the SBN algorithm has significant advantages in dealing with uncertainty and probabilistic inference problems. It can perform efficient and accurate probabilistic inference based on limited observational data and expert knowledge, thereby helping researchers more precisely identify the key risk factors in the shipping industry.

[0096] The shipping accident risk identification and assessment method provided by the present invention based on the enhanced rough BWM evaluation method combines the genetic algorithm, the best-worst method (BWM), and the rough comprehensive compromise method (CoCoSo) to propose a new shipping risk assessment framework, aiming to address the evaluation and optimization problems of shipping risk factors, especially in the face of uncertainty and fuzzy information. The design of this framework fully integrates the advantages of Bayesian networks in dealing with uncertainty and combines the processing ability of rough set theory for fuzzy information. By combining the genetic algorithm and the BWM algorithm, stable calculation of weights is achieved, improving the accuracy and reliability of the evaluation results.

[0097] Based on the above reasons, the present invention can be widely promoted in the fields of transportation and the like. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0099] Figure 1 It is a flowchart of a shipping accident risk identification and assessment method based on the enhanced rough BWM evaluation method in the present invention.

[0100] Figure 2 It is an algorithm framework diagram of a shipping accident risk identification and assessment method based on the enhanced rough BWM evaluation method in the present invention.

[0101] Figure 3 It is a labeled recognition prediction model diagram.

[0102] Figure 4 It is a prediction result diagram of a Bayesian model based on a neural network.

[0103] Figure 5 It is a refined probability distribution diagram obtained from the training of the static Bayesian network model in the embodiment of the present invention.

[0104] Figure 6 It is a diagram of the priority ranking of various shipping accident risks under different elastic approximation accuracies in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0105] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to describe the present invention in detail.

[0106] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. The following description of at least one exemplary embodiment is actually only illustrative and in no way restricts the present invention and its application or use. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0107] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they specify the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0108] Unless otherwise specifically stated, the relative arrangements of the components and steps, numerical expressions, and numerical values set forth in these embodiments do not limit the scope of the present invention. At the same time, it should be clear that, for the sake of convenience of description, the sizes of the various parts shown in the drawings are not drawn in actual proportional relationships. Technologies, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the said technologies, methods, and devices should be regarded as part of the authorized specification. In all the examples shown and discussed herein, any specific values should be construed as merely exemplary and not as a limitation. Therefore, other examples of the exemplary embodiments may have different values. It should be noted that: like reference numerals and letters denote like items in the following drawings, and thus, once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0109] As Figure 1 shown, the present invention provides a method for identifying and evaluating the risk of shipping accidents based on an enhanced rough BWM evaluation method, including:

[0110] Construct an identification and prediction model using a Bayesian network based on a neural network; the Bayesian network mainly consists of two core components. One is the structure model, which is represented by an acyclic directed graph, where each node represents a variable, and the directed edges between the nodes represent the conditional dependence relationships between these variables. The other is the parameter, which is used to represent the conditional probability table (CPT) of each variable, describing the probability distribution of the variable given the parent nodes.

[0111] In specific implementation, as a preferred implementation manner of the present invention, the constructing an identification and prediction model using a Bayesian network based on a neural network specifically includes:

[0112] Construct a static Bayesian network using a rule-based pruning strategy for inferring the probabilities associated with a set of Boolean variables. If the variable V represents a node, then (j = 1, 2,..., m) represents the set of parent nodes of the variable V. If V is a vector containing multiple random variables, then the joint probability distribution of V is:

[0113]

[0114] The conditional dependence relationship between network results and various variables remains unchanged throughout the inference process. The results obtained from training a neural network-based Bayesian model are refined through recursion. Regarding the marginal probability, for a node without parent nodes, its marginal probability; for a node with parent nodes, the parent nodes are obtained through recursion.

[0115] The Boolean variables provided by the conditional probability table CPT are:

[0116] P(Y j = True | Pa(Y j ))

[0117] where Y j represents a random variable, and Pa(Y j ) represents the parent nodes of Y j .

[0118] The complementary value of the Boolean variable is expressed as:

[0119] P(Y j = False | Pa(Y j )) = 1 - P(Y j = True | Pa(Y j ))

[0120] Determine the weights of shipping accident risk factors through decision-making; based on the prediction and recognition results obtained by the static Bayesian network algorithm, invite a group of experts to evaluate the relative importance of each criterion according to their own experiences, knowledge, and expertise, using the numbers 1 to 9 to determine the priority of the best criterion relative to other criteria. The higher the score, the more important the criterion, and obtain the best-worst vectors (B vector and W vector).

[0121] In specific implementation, as a preferred implementation manner of the present invention, the shipping accident risk factors include: the number of affected areas D, the number of large ship losses L, the number of ship accidents A, and the total number of ship losses T.

[0122] In specific implementation, as a preferred implementation manner of the present invention, the determination of the weights of shipping accident risk factors through decision-making specifically includes:

[0123] Represent the set of criteria as C = {D, T, L, A}, and determine the best criterion S B and the worst criterion S W :

[0124]

[0125] where represents the scoring vector of the i-th decision-maker for the best criterion S B , Denote the scoring vector of the \(i\)-th decision maker under the best criterion \(S\) W , Denote the score of the judgment of the \(i\)-th decision maker on criterion \(j\) under the best criterion \(S\) B ; Denote the score of the judgment of the \(i\)-th decision maker on criterion \(j\) under the worst criterion \(S\) ; where \(1\leq i\leq m\), \(1\leq j\leq q\). For example W denotes that the evaluation score of the best criterion is 9 compared with that of the worst criterion. Here, \(m\) is the number of decision makers and \(q\) is the number of criteria.

[0126] Define the distances of the conditional judgment sets of the best criterion \(S\) B and the worst criterion \(S\) W respectively as follows:

[0127]

[0128] where \(d\) Bq denotes the distance of the conditional judgment set of vector \(B\), and \(d\) Wq denotes the distance of the conditional judgment set of vector \(W\).

[0129] According to the rough set theory, rough numbers can be represented by using upper and lower approximation values, which can show the subjectivity and uncertainty in the decision-making of decision makers. To express this fuzziness and uncertainty more accurately, an elastic approximation factor \(k\) is introduced to represent the clarity of decision-making, and the judgment of sensitivity is transformed into EARN. The calculation process is as follows:

[0130]

[0131] where is the lower approximation of , and is the upper approximation of ; \(U\) represents the universal set, denotes the scoring vector that satisfies the condition, and \(H\) j denotes the scoring vector of the \(j\)-th criterion.

[0132] When \(k = 0\), the decision is clear. When \(k\in(0,1)\), there is fuzziness in the decision, and the degree of fuzziness is determined by the elastic approximation factor and the distance \(kd\) of the vector conditional judgment set; when \(k = 1\), the decision is fuzzy, then

[0133] To quantify and analyze the subjectivity and fuzziness of crisp values more precisely, the ordered weighted average (OWA) method is adopted. OWA is an aggregation technique that can take into account the relative importance and weights among different crisp values, so as to obtain a comprehensive optimal value. Convert the crisp value judgment set into the form of rough numbers;​

[0134]

[0135] Among them, represents the lower boundary, represents the upper boundary, and g and f are respectively and the amounts of substances of the objects in; {o(1), o(2), …, o(g)} is a permutation of {(1), (2), …, (g)}; {o(1), o(2), …, o(f)} is a permutation of {(1), (2), …, (f)}, and w v represents the corresponding v-th OWA weight, where w v ≥0,

[0136] According to the basic operation aspect of OWA, the judgment of the crisp value will be re - sorted to reflect the relative importance and priority among different judgments. This re - sorting process will ensure that when aggregating the crisp value judgments, the ordered relationship and weight distribution among them can be considered, so as to obtain a more reasonable and accurate comprehensive evaluation result.

[0137] The form of EARN of

[0138]

[0139] is: B and the worst criterion S W The form of EARN is transformed into:

[0140]

[0141] Use ROWA to improve the weights to obtain the weight coefficients of the shipping accident risk factors; in order to statistically aggregate the individual rough importance judgments of six decision - makers on a certain attribute or characteristic, the ROWA operator is used as a tool and the weights are improved. This operator takes into account the roughness (i.e., uncertainty) of the experts' evaluations and the orderliness of their judgments, so as to obtain a comprehensive and accurate aggregation result.

[0142] Specifically in implementation, as a preferred implementation manner of the present invention, the use of ROWA to improve the weights to obtain the weight coefficients of the shipping accident risk factors specifically includes:

[0143] The ROWA operator of the best criterion S B is expressed as:

[0144]

[0145] Among them, wa denotes the a-th ROWA weight, 0 ≤ w a ≤ 1, EARN k (P Bq ) is restricted by and represents the lower and upper bounds of EARN k (P Bq ) respectively;

[0146] Calculating RBI Bq refers to the rough fuzzy boundary interval of P Bq :

[0147]

[0148] The ROWA operator of the worst standard S W can be expressed as:

[0149]

[0150] where, w b denotes the b-th ROWA weight, 0 ≤ w b ≤ 1, EARN k (P qW ) is restricted by and represents the lower and upper bounds of EARN k (P qW ) respectively;

[0151] Let ROWA:

[0152]

[0153] where, ω = (ω1, ω2, …, ω l ) is the weighted vector associated with the function ROWA, ω s ∈ [0, 1], s ∈ {1, 2, …, l}, b c is the c-th largest element in the data (f1, f2, …, f v , … f l ), … f

[0154] Let ω = (ω1, ω2, …, ω l ) be the weight vector of the ROWA operator, and the definition of ω v is:

[0155]

[0156] where, μ l is the mathematical expectation weighted by l obtained, σl from l in σ l and the weight the standard deviation, μ, obtained l and σ l are defined as follows:

[0157]

[0158]

[0159] For the evaluated decision data f1, f2, …, f v , …, f l Using the obtained mean μ l and variance σ l perform standardization to obtain (α1, α2, …, α v , …, α l ):

[0160]

[0161] Solve the values of θ(α v ) at α1, α2, …, α v , …, α l respectively as β1, β2, …, β v , …, β l ,

[0162]

[0163] Since ω k ∈[0, 1] and perform unit normalization:

[0164]

[0165] Obtain the weight vectors ω1, ω2, …, ω v , …, ω l ;

[0166] For each shipping accident risk factor satisfying and calculate the weight of each criterion:

[0167] minλ

[0168]

[0169] w v ≥0, for all v.

[0170] Where represents the weight of the best criterion S B of is the worst standard S W of the weight, denotes EARN k (P B ) average value, denotes EARN k (P W ) average value.

[0171] Construct a rough decision matrix based on the elastic approximation factor to evaluate the risk factors of shipping accidents; during the evaluation of risk factors, decision-makers express their subjective judgments on the importance of each risk factor based on the criteria according to a grading system from 1 to 9.

[0172] When specifically implemented, as a preferred implementation manner of the present invention, the constructing a rough decision matrix based on the elastic approximation factor to evaluate the risk factors of shipping accidents specifically includes:

[0173] For each independent risk factor m of shipping accidents, form a scoring set, denoted as where represents the specific score for the risk factor m in the d-th decision;

[0174]

[0175] where, C d represents the d-th rough decision matrix, m = 1, 2,..., x; n = 1, 2,..., y, is the score for the risk factor m of shipping accidents in the d-th decision under the n-th standard Sn; convert the rough value to a definite fixed value:

[0176]

[0177] where, and are respectively the maximum and minimum values, and the EARN formula of c mn is as follows:

[0178]

[0179] where, and are respectively the upper and lower bounds of the boundary, and obtain the rough decision matrix EARN under the elastic approximation factor k k :

[0180]

[0181] For the cost and benefit criteria, a normalization method with rough boundaries is adopted to convert rough values into a standardized expression form:

[0182]

[0183]

[0184] where is the standardized expression form.

[0185] Calculate the sum of weighted comparable sequences of risk factors and the sum of weighted comparable sequences of weight power exponents:

[0186]

[0187] where S m represents the weighted comparable sequence, and E m represents the weighted comparable sequence of weight power exponents; w n represents the optimal weight obtained by the BWM algorithm.

[0188] Adopt an aggregation evaluation method to determine the relative weight of each shipping accident risk factor, and convert the risk factors into numerical values, and arrange and classify the risk factors according to the numerical values.

[0189] Specifically, as a preferred implementation manner of the present invention, the adopting an aggregation evaluation method to determine the relative weight of each shipping accident risk factor and converting the risk factors into numerical values specifically includes:

[0190] Determine the relative weight of each shipping accident risk factor by using an aggregation evaluation method:

[0191]

[0192] where θ is an empirical coefficient;

[0193] Determine the importance level of shipping accident risk factors:

[0194]

[0195] Convert the fuzzy shipping accident risk factors into numerical representations, and arrange or classify the risk factors in an orderly manner according to these numerical values.

[0196]

[0197] where ρ i is the coefficient of is a m crisp set.

[0198] The present invention combines the Genetic Algorithm (GA), the Best-Worst Method (BWM), and the CoCoSo method, and proposes a new shipping risk assessment framework to address the assessment and optimization of shipping risk factors, especially in the face of uncertainty and fuzzy information. The design of this framework fully integrates the advantages of Bayesian networks in handling uncertainty and combines the fuzzy information processing ability of rough set theory. By combining the Genetic Algorithm and the BWM algorithm, stable calculation of weights is achieved, improving the accuracy and reliability of the assessment results.

[0199] Shipping risk management strategies adopt systematic, scientific, and technological means to comprehensively control all aspects from risk identification, assessment, control, monitoring to emergency response. By combining modern data analysis methods, intelligent technologies, and multi-party cooperation, various risks during the shipping process can be effectively reduced, enhancing the safety and economic benefits of shipping operations.

[0200] Embodiment

[0201] In this embodiment, the affected area (D), total number of ships (T), large ships (L), and accident volume (A) are introduced as criteria for evaluating risk factors, and the criterion set is C = {D, T, L, A}. Among them, the affected area (D) represents the number of affected areas; large ships (L) represent the number of large ship losses; accidents (A) represent the number of ship accidents; and the total number of ships (T) represents the total number of ship losses. Determine the best criterion (SB) and the worst criterion (SW). At this stage, no comparison is made. For example, for a certain decision-maker, the affected area (D) and large ships (L) may be the best and worst criteria. Details of the shipping accident risk criteria are shown in Table 1.

[0202] Table 1 Measurement Criteria for Shipping Risk Factors

[0203]

[0204]

[0205] Take the amount as the target variable. Through a Bayesian model based on neural networks, the occurrence probability of the target event under multiple criteria is predicted.

[0206] Table 2 Data of Risk Factors under Different Conditions

[0207]

[0208] In Table 2, the affected areas 1 and 2 represent data on five risk factors for different years, with the amount taken as the target variable for the analysis result. Compared with other indicators, the accident volume can best reflect the actual occurrence frequency of risk events and also affects the other three criteria. The change in the accident volume directly indicates the dynamic change of a certain risk factor over a period of time. Therefore, it is hoped that the prediction result can be biased towards this criterion.

[0209] The training process is as follows. In each iteration, the output probability of the model is obtained through forward and backward propagation, the loss is calculated, and the weights are adjusted through backpropagation. The loss and sample results are output once every 50 epochs, and the model is trained through 500 iterations. The binary cross-entropy loss function is used for optimization, with the aim of finding a set of parameters to minimize the error between the input data and the target data. The iterative output during the training process is shown in the following table, and the bar chart is drawn as attached Figure 4 as shown.

[0210] Table 3 Iteration Process

[0211] Epoch Loss Out sample 1 0.6205130815505981 0.4947044253349304 51 0.5123624205589294 0.18786972761154175 101 0.4714396893978119 0.13461028039455414 151 0.44212231040000916 0.10175592452287674 201 0.419248104095459 0.08435012400150299 251 0.3999759554862976 0.07572125643491745 301 0.3831673264503479 0.07217538356781006 351 0.3683832287788391 0.07153887301683426 401 0.35540464520454407 0.07247483730316162 451 0.3440625071525574 0.07414308935403824

[0212] The final predicted probability results after training using the Bayesian network model based on neural networks are listed in the following table, and the bar chart is drawn as attached Figure 5 as shown.

[0213] Table 4 Final Probability Distribution

[0214] Criteria True False LS 0.05 0.95 AQ 0.74 0.26 TS 0.69753 0.30247 DA1 0.41405097800000007 0.585949022 DA2 0.3314118220000001 0.668588178

[0215] The decision-makers evaluate the preference degree of SB relative to all other criteria, quantify these preference degrees into specific numerical values to form the B vector, denoted as PB. Similarly, the preference degree of SW relative to all other criteria is evaluated to form the W vector, denoted as PW. As shown in Table 5, it contains preference values from 1 to 9, where 1 represents no preference or equal preference, and 9 represents extremely high preference.

[0216] Table 5 Data of B Vector and W Vector

[0217]

[0218] Taking the f BD vector as an example, the specific preference vector of the decision-makers is f BD = {6, 4, 4, 3, 3, 5}. The elastic approximation accuracy k takes the values of 0, 0.5, and 1 respectively, and the ROWA weights are obtained based on the method of normal distribution:

[0219]

[0220] The EARN form results are calculated as follows:

[0221] When k = 0, the cognitive decision of the decision maker is very certain and there is no ambiguity. The ROWA operator form result of f BR is as follows:

[0222]

[0223] When k = 0.5, there is partial ambiguity in the cognitive decision of the decision maker. The ROWA operator form result of f BR is as follows:

[0224]

[0225] When k = 1, the cognition of the decision maker is completely ambiguous at this time. The ROWA operator form result of f BR is as follows:

[0226]

[0227] Other vectors are similar to the above steps. The only difference is that for the decision vectors with only two preference degrees in the vector, the weighted average value is taken when calculating ROWA. The preference degree rough numbers of vector B and vector W are shown in the following table.

[0228] Table 6 Preference degree rough numbers of vector B (k = 0.5)

[0229]

[0230] Table 7 Preference degree rough numbers of vector W (k = 0.5)

[0231]

[0232] By using the genetic algorithm to write code, through steps such as initializing the population, decoding the binary encoding, calculating the objective function value, calculating the fitness, selection operation, crossover operation, and mutation operation, we can gradually approach the optimal solution of the problem. The weight coefficient distribution under different elastic approximation accuracies is calculated as shown in the following table:

[0233] Table 8 Optimal weight distribution

[0234]

[0235] First, the decision makers score the preference degrees of risk factors under different criteria. The detailed data of the scores are shown in the following table.

[0236] Table 9 Collected decision evaluation data

[0237]

[0238]

[0239] Taking the elastic approximation factor k = 0.5 as an example, the rough decision matrix is calculated as shown in the following table.

[0240] Table 10 Rough Decision Matrix

[0241]

[0242] When evaluating these four criteria, they are distinguished from the two perspectives of cost and benefit. The rough values are calculated and normalized as shown in the following table.

[0243] Table 11 Normalized Standard Values

[0244]

[0245]

[0246] Calculate the sum of weighted comparability sequences and the sum of weighted comparability sequences with weight power exponents. The results when k = 0.5 are shown in Table 12.

[0247] Table 12 Weighted Comparability Sequences and Weighted Comparability Sequences with Weight Power Exponents

[0248] risk factor Si Ei Rf1 [0.3008,0.5272] [3.0215,3.3885] Rf2 [0.071,0.3732] [1.0255,3.1287] Rf3 [0.3305,0.6083] [1.6948,3.4109] Rf4 [0.0865,0.3465] [2.1871,2.8466] Rf5 [0.6299,0.8272] [3.4821,3.789]

[0249] Calculate a m1 , a m2 , a m3 The rough set is transformed into a crisp set, and the final evaluation values of shipping risks and the corresponding priority rankings are shown in the following table. The result graph is drawn as attached Figure 6 as shown.

[0250] Table 13 Relative Importance Values and Priority Rankings of Each Risk Factor

[0251]

[0252] The results show that under the condition that the variable precision is 0.5, the most important criterion is A, followed by T, and finally D and L. This study shows that "RF5 (piracy)", which belongs to the risks of safety and external threats, is the most important shipping risk factor. The next two important risks are "RF3 (fire)" and "RF1 (collision)" related to human operations. The next important risk is "RF4 (mechanical failure / damage)" related to equipment. The significance of the environmental risk of "RF2 (stranding)" is lower than that of other risks. The avoidance strategies provided for different industries tend to focus on the two most important risks. Spatially, the occurrence of shipping accidents is also related to regions. In terms of accident types, fires are very uncontrollable and have the highest occurrence probability. Secondly, the frequencies of collisions and groundings are high, and the number of ships lost is also relatively large. In terms of ship types, cargo ships account for a relatively large proportion of the ships lost, followed by fishing boats and passenger ships.

[0253] The results show that the risk rate of piracy is the highest, followed by fires, and the other risk factors are relatively low. Therefore, shipping companies, shipping lines, ports and related industries are in urgent need of formulating detailed countermeasures to improve safety in order to effectively address these risks.

[0254] The strategies for shipping companies are as follows:

[0255] To comprehensively enhance the safety of ships, the early warning and monitoring mechanisms should be optimized, and equipment such as satellite communication, AIS (Automatic Identification System), and high-precision radar should be fully utilized to achieve real-time and comprehensive monitoring of the ship's navigation route and the surrounding sea area, and promptly detect and avoid areas suspected of pirate activities. In addition, each ship should be equipped with an advanced fire protection system, ensuring the installation of automatic sprinkler systems, highly sensitive smoke detectors, and various types of fire extinguishers, and regular inspections and maintenance should be carried out to effectively respond to fire risks. Regular maintenance and inspections of the ship should be carried out, especially for key parts such as the propulsion system, navigation equipment, and hull structure, to ensure the stability and safety of the ship's performance. Investment should be made in advanced navigation equipment such as radar, GPS, AIS, and echo sounders to improve the ship's navigation accuracy, collision avoidance ability, and water depth monitoring ability during shallow water navigation. In terms of piracy prevention, a detailed emergency plan should be formulated, clarifying the procedures for sending signals, evacuating personnel, selecting safe havens, conducting night patrols, self-defense counterattacks, and communicating with rescue forces when encountering pirates. At the same time, each ship should be equipped with a fire emergency plan and regular drills should be carried out to improve the crew's response capabilities. Regular training of the crew is also essential, and the training content should cover skills such as identifying pirate ships, using anti-piracy equipment, psychological coping strategies, safe evacuation, communication with rescue forces, and fire prevention and emergency evacuation. In addition, comprehensive navigation training should be provided to improve the crew's handling and emergency response capabilities in complex situations. In terms of ship protection, non-lethal self-defense weapons such as high-pressure water guns, electric grids, barbed wire, and lasers can be equipped on the ship to effectively prevent pirates from boarding and promptly respond to potential threats. Strictly control the quantity of flammable items to ensure their storage meets safety standards. A safe compartment should also be set up on the ship as a shelter in case of emergencies. In terms of repair and replacement, it is recommended to use high-quality spare parts and materials to reduce the risk of mechanical failures. Finally, international cooperation should be strengthened, and close contacts should be established with navies, maritime organizations, and anti-piracy centers of various countries to promptly obtain intelligence on pirate activities and seek international support and cooperation to jointly enhance maritime safety.

[0256] To enhance the safety of ship operations, it is necessary to conduct regular shipping risk assessments, identify potential risk points, and take appropriate management measures. Diversifying and transferring some risks through means such as insurance is a necessary safeguard. Ensure that the ship is maintained on schedule, especially focusing on key inspections of fire hazard areas such as the electrical system and fuel storage area. When selecting cooperative ships, those with good maintenance records and low mechanical failure rates should be prioritized to reduce safety risks during operation. For different emergency situations (such as pirate activities, fires, collisions, etc.), detailed emergency response plans should be formulated, covering emergency contacts, evacuation routes, and rescue procedures. Based on pirate activity reports and risk assessments, reasonably plan safe shipping routes and try to avoid known high-risk areas. This requires close cooperation between shipping companies and professional institutions to obtain the latest risk information. At the same time, the shipping route should be reasonably selected according to data such as nautical charts, tides, and water depths to prevent the ship from entering shallow water areas and reduce the risk of grounding. To further enhance the ship's protection ability, it is possible to cooperate with professional maritime security companies to provide armed escort services for the fleet. In addition, research and development and application of ship safety technologies should be increased, using the Automatic Identification System (AIS) to track nearby ships or deploying drones for aerial reconnaissance to detect potential threats in a timely manner. During the design and construction of new ships, flame-retardant materials should be prioritized and design solutions that can reduce the fire risk should be adopted. Establishing a sound emergency response mechanism is crucial, including communication channels with local navies or security agencies, communication protocols and rescue procedures in emergency situations, to ensure a rapid response in case of pirate attacks and maximize the safety of the ship and crew. In addition, a comprehensive fire risk management system should be developed and maintained, covering risk assessment, preventive measures, monitoring, and emergency response plans. To further strengthen safety, a pirate activity information sharing mechanism should also be established to ensure that relevant intelligence can be promptly transmitted to ships and ports, guaranteeing the safety and smoothness of shipping activities.

[0257] Strategies for ports:

[0258] To enhance the safety management of ports, especially in areas with frequent pirate activities, ports should strengthen security measures, including installing additional surveillance cameras, increasing patrol frequencies, and enhancing the training of security personnel. Ports need to be equipped with sufficient and well-functioning fire-fighting facilities to ensure that these devices are always in normal working condition. At the same time, ports should implement a strict traffic management system, demarcate safe navigation channels, set speed limits, and provide professional pilotage services to ensure the safety of ships when entering and leaving the port. Monitor the dynamic of ships in real time through the VTS system, provide information and guidance in a timely manner, and prevent collisions and other accidents. When ships enter and leave the port, strict safety inspections are required to prevent pirates from approaching large ships using small boats or disguised vessels. For ships docked at the port, the port should provide comprehensive protection services, including armed escorts and the installation of monitoring equipment, to ensure the safety of ships during berthing. At the same time, the fire-fighting equipment of ships should be inspected regularly to eliminate fire hazards in a timely manner. In addition, the port should establish a professional fire-fighting team to ensure a rapid response and efficient fire extinguishing in case of a fire. Ports should have professional ship repair services, equipped with emergency repair facilities and technical support, so that ships can be repaired quickly when mechanical failures occur, thereby improving operational efficiency and ensuring safety.

[0259] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A shipping accident risk identification and assessment method based on an enhanced rough BWM evaluation method, characterized in that Including: Construct an identification and prediction model by using a Bayesian network based on a neural network; Determine the weights of shipping accident risk factors through decision-making; Use ROWA to improve the weights to obtain the weight coefficients of shipping accident risk factors; Construct a rough decision matrix based on the elastic approximation factor to evaluate shipping accident risk factors; Adopt an aggregation evaluation method to determine the relative weights of each shipping accident risk factor, convert the risk factors into numerical values, and arrange and classify the risk factors according to the numerical values.

2. The shipping accident risk identification and assessment method based on the enhanced rough BWM evaluation method according to claim 1, wherein, The construction of the identification and prediction model by using a Bayesian network based on a neural network specifically includes: A static Bayesian network is constructed using a rule-based pruning strategy for inferring probabilities associated with a set of Boolean variables. If variable V represents a node, then (j = 1, 2, …, m) represents the set of parent nodes of variable V. If V is a vector containing multiple random variables, then the joint probability distribution of V is: Refine the results trained by the Bayesian model based on the neural network through the recursive method. Regarding the marginal probability, for a node without parent nodes, obtain the marginal probability; for a node with parent nodes, obtain the parent nodes through the recursive method. The Boolean variables provided by the conditional probability table CPT are: P(Y j = True | Pa(Y j )) where, Y j represents a random variable, and Pa(Y j ) represents the parent nodes of Y j ; Represent the complement value of the Boolean variable as: P(Y j = False | Pa(Y j )) = 1 - P(Y j = True | Pa(Y j ))。 3. The shipping accident risk identification and assessment method based on the enhanced rough BWM evaluation method according to claim 1, wherein, The shipping accident risk factors include: the number of affected areas D, the number of large ship losses L, the number of ship accidents A, and the total number of ship losses T.

4. The shipping accident risk identification and assessment method based on the enhanced rough BWM evaluation method according to claim 1, characterized in that, The determination of the weights of shipping accident risk factors through decision-making specifically includes: Represent the standard set as C = {D, T, L, A}, and determine the best standard S and the worst standard S through decision-making B and the worst standard S W : Among them, represents the scoring vector of the i-th decision maker under the best criterion S B . represents the scoring vector of the i-th decision maker under the best criterion S W . represents the score of the judgment of the i-th decision maker on criterion j under the best criterion S B . represents the score of the judgment of the i-th decision maker on criterion j under the worst criterion S W . Define the distances of the condition judgment sets of the best standard S B and the worst standard S W as follows respectively: Among them, d Bq represents the distance of the B vector condition judgment set, and d Wq represents the distance of the W vector condition judgment set; Introduce the elastic approximation factor k to represent the clarity of the decision-making, convert the judgment of sensitivity into EARN, and the calculation process is as follows: Among them, is the lower bound approximation of is the upper bound approximation of; U represents the universal set, represents the scoring vector that satisfies the condition, H j represents the scoring vector of the j-th criterion, When k = 0, the decision is clear, When k ∈ (0, 1), there is ambiguity in the decision, and the degree of ambiguity is determined by the elastic approximation factor and the distance kd of the vector condition judgment set; when k = 1, the decision is fuzzy, then Adopt the ordered weighted average OWA method to convert the clear value judgment set into the form of rough numbers; Among them, represents the lower boundary, represents the upper boundary, and g and f are respectively and the amounts of substances of the objects in; {o(1), o(2), …, o(g)} is a permutation of {(1), (2), …, (g)}; {o(1), o(2), …, o(f)} is a permutation of {(1), (2), …, (f)}, w v represents the corresponding v-th OWA weight, where w v ≥ 0, The form of EARN is as follows: Convert the form of EARN for the best standard S B and the worst standard S W into:

5. The shipping accident risk identification and assessment method based on the enhanced rough BWM evaluation method according to claim 1, characterized in that, The use of ROWA to improve the weights to obtain the weight coefficients of shipping accident risk factors specifically includes: The optimal standard S B is represented by the ROWA operator as follows: Among them, w a represents the a-th ROWA weight, 0 ≤ w a ≤ 1, EARN k (P Bq ) is restricted by and respectively represents the lower bound and the upper bound of EARN k (P Bq ); Calculating RBI Bq Denote P Bq as the rough fuzzy boundary interval of: Worst standard S W The ROWA operator of can be expressed as: where, w b represents the b-th ROWA weight, 0 ≤ w b ≤ 1, EARN k (P qW ) is subject to , representing the lower and upper bounds of EARN k (P qW ) respectively; Let ROWA: Among them, ω = (ω1, ω2, …, ω l ) is the weighted vector associated with the function ROWA, ω s ∈[0, 1], s ∈ {1, 2, …, l}, b c is the c-th largest element in the data (f1, f2, …, f v , … f l ), and R is the set of real numbers; Let ω = (ω1, ω2, …, ω l ) be the weight vector of the ROWA operator, and define ω v as follows: Among them, μ l is the mathematical expectation weighted by l obtained, σ l is the standard deviation obtained by l in σ l and the weight obtained, the definitions of μ l and σ l are as follows: For the evaluated decision data f1, f2, …, f v , …, f l Using the obtained mean μ l and variance σ l Perform standardization to obtain (α1, α2, …, α v , …, α l ): Solve θ(α v ) separately for the values of α at α1, α2, …, α v , …, α l to obtain β1, β2, …, β v , …, β l , Since ω k ∈ [0, 1] and perform unitization processing: Obtain weight vectors ω1, ω2, …, ω v , …, ω l ; For each shipping accident risk factor satisfying and EARN k (P q W), calculate the weight of each criterion: w v ≥0, for all v. Among them, represents the weight of the best criterion S B . is the weight of the worst criterion S W . represents the average value of EARN k (P B ). represents the average value of EARN k (P W ).

6. The shipping accident risk identification and assessment method based on the enhanced rough BWM evaluation method according to claim 1, characterized in that, minλ For each independent shipping accident risk factor m, a scoring set is formed, denoted as where represents the specific score for the risk factor m in the d-th decision; Among them, C d represents the d-th rough decision matrix, where m = 1, 2, …, x; n = 1, 2, …, y, is the score of the d-th decision for the shipping accident risk factor m under the n-th criterion Sn; convert the rough value into a definite fixed value: Among them, and are respectively the maximum value and the minimum value, and the EARN formula of c mn is as follows: Among them, and are respectively the upper and lower bounds of the boundary of, and obtain the rough decision matrix EARN k : The construction of a rough decision matrix based on the elastic approximation factor to evaluate shipping accident risk factors specifically includes: Among them, is a standardized expression form; For the cost and benefit criteria, adopt the normalization method of rough boundaries to convert the rough values into a standardized expression form: Among them, S m represents the weighted comparability sequence, and E m represents the weight power exponent weighted comparability sequence; w n represents the optimal weight obtained by the BWM algorithm.

7. The shipping accident risk identification and assessment method based on the enhanced rough BWM evaluation method according to claim 1, characterized in that, Calculate the sum of the weighted comparability sequences of risk factors and the sum of the weighted comparability sequences of weight power exponents: The adoption of an aggregation evaluation method to determine the relative weights of each shipping accident risk factor and convert the risk factors into numerical values specifically includes: Determine the relative weights of each shipping accident risk factor by using the aggregation evaluation method: where θ is an empirical coefficient; Determine the importance level of shipping accident risk factors: Convert the fuzzy shipping accident risk factors into numerical representations, and arrange or classify the risk factors in an orderly manner according to these numerical values. where ρ i is 's coefficient, is the crisp set of a m .

Images