Multifunctional calculation and analysis system and method for community toughness influence factors

Abstract

The invention provides a multifunctional calculation and analysis system and method for community toughness influence factors, and belongs to the field of data analysis. According to the method, a Bayesian network structure model is established by adopting a decision-making laboratory analysis method, and the failure probability of each non-root node is calculated by introducing default nodes and using a Bayesian conditional probability formula. An expert questionnaire and a fuzzy comprehensive evaluation method are utilized to obtain a prior probability value of a root node of the community toughness BN network, and finally the most approximate cause chain enabling the community toughness to be invalid is found out through a reverse deduction function. In cooperation with the method, a corresponding analysis system is also generated. The system has relatively high practicability and convenience, not only can quantitatively analyze the community toughness, but also can intuitively and intuitively understand the influence of each influence factor on the community toughness, and has important guiding significance for community disaster-resistant planning and decision-making.

Description

Technical Field

[0001] The present invention belongs to the field of data analysis, and more specifically, relates to a multifunctional computing and analyzing system and method for factors affecting community resilience. Background Art

[0002] Community resilience refers to the ability of a community to minimize disaster risks, quickly restore its original functions, or even achieve a better state through self-organization and self-adjustment after encountering natural disasters or other emergencies. Understanding the factors affecting community resilience and calculating and analyzing these factors are of great significance for improving the community's ability to resist disasters and optimizing community resource allocation.

[0003] Community resilience is a complex, multi-dimensional theory that involves factors in the community's economy, environment, society, and other aspects. These factors include not only the community's hardware facilities, such as housing, transportation, and communications, but also the community's software facilities, such as community organizations, projects, and activities. The numerous influencing factors, complex influencing relationships, and difficult-to-quantify issues make community resilience analysis very difficult.

[0004] Traditional community resilience assessment methods mostly use qualitative analysis, questionnaire surveys, etc. Although these methods can reflect the situation of community resilience to a certain extent, they are inevitably affected by subjective factors. In addition, these methods often cannot accurately quantify the weight of each influencing factor, which is not conducive to scientific decision-making. At the same time, community resilience analysis involves many factors and complex relationships, requiring a large amount of data and analysis methods. Traditional analysis methods are often inefficient and cannot meet the fast-paced and high-efficiency needs of modern society. Summary of the invention

[0005] The present invention proposes a new multifunctional computational analysis method and system for community resilience influencing factors. It uses methods such as Bayesian networks and Leaky Noisy-OR models to quantify influencing factors, which can more accurately reflect the weights of influencing factors and improve the efficiency of analysis. In addition, it also provides an intuitive graphical interface, allowing users to understand the analysis results more intuitively and conveniently.

[0006] In order to achieve the above object, the present invention is implemented by adopting the following technical scheme: the method comprises:

[0007] The decision laboratory analysis method was used to establish the Bayesian network structure model;

[0008] By default node P L The introduction of and calculation of the Bayesian conditional probability formula are used to obtain the conditional probability of failure of each non-root node;

[0009] Obtain the prior probability value of the root node of the community resilience BN network, that is, the failure probability of the corresponding influencing factor variable;

[0010] Based on the Bayesian network model, the maximum causal chain that causes community resilience to fail is derived through reverse deduction.

[0011] In one embodiment, the establishment of the Bayesian network structure model includes:

[0012] A Bayesian network was initially established based on the influencing factors of each category of community resilience given by experts;

[0013] The experts’ scores on the degree of influence of various influencing factors are read, and the final community resilience Bayesian network structure module is calculated using the decision laboratory analysis method DEMATEL.

[0014] In one solution, the conditional probability of non-root node failure adopts the Leaky Noisy-OR model: adding the default node Z L , used to represent the remaining potential variables that are not considered but will cause the child node to fail;

[0015] Even if all parent nodes Z i None of them will fail, and the child node Y will also be in the default node Z L Failure in case of failure;

[0016] Use P L Indicates that only the default node P L The probability that child node Y fails when its parent node fails and other parent nodes do not fail.

[0017] In one solution, the prior probability value of the root node of the community resilience BN network is obtained by combining an expert questionnaire with a fuzzy comprehensive evaluation method to obtain the prior probability value of the root node of the community resilience BN network;

[0018] The seven-level judgment theory is used to divide the fuzzy language of variable factor failure probability into seven levels;

[0019] Different weight indicators are assigned to each expert, and the weighted average method is used to process the fuzzy rating number. By comparing the evaluation results of each expert with the average value, the closer the expert is, the greater the weight is. Suppose the fuzzy evaluation number of an expert is F k =(F k1 , F k2 , F k3 , F k4 ), where K = 1, 2, ..., n;

[0020] (1) Solve the arithmetic mean F of n expert trapezoidal fuzzy numbers a ;

[0021]

[0022] (2) Solve F k and the arithmetic mean F a distance;

[0023]

[0024] (3) Solve F k and the arithmetic mean F a similarity;

[0025]

[0026] (4) Solving the expert evaluation weight ω k ;

[0027]

[0028] (5) Solve the weighted comprehensive fuzzy evaluation number F;

[0029]

[0030] (6) Use the mean area method to solve the fuzzy set and obtain the prior probability of the root node;

[0031]

[0032] On the other hand, a multifunctional computing and analysis system for community resilience influencing factors is provided, and the analysis system is applicable to the method described, and the analysis system includes: displaying a factor classification list, reading and modifying expert sub-node influence scores, displaying a Bayesian network structure, reading expert scores on the failure probability of a page root node, displaying a final Bayesian network model, performing reverse deduction to obtain the most probable cause chain of community resilience failure, spider web diagram analysis of the influence of each factor, and modifying parameter configuration.

[0033] In one solution, the display factor classification list is based on the classification of various factors affecting community resilience provided by experts, and a Bayesian network structure is established on the web page in the form of a tree diagram;

[0034] Each node can dynamically display and retract the nodes in the hierarchy below it by clicking the mouse. The font size increases step by step from the root node to the center node, and each layer of nodes is arranged in the same horizontal position.

[0035] In one embodiment, the reading and modifying expert subnode influence degree scoring adopts the group distance data processing method in statistics;

[0036] Read the expert scoring file in the project program folder, traverse each data in the table, use Python's built-in collections module and numpy library to find the mode, and create a Counter object to count the number of occurrences of each element;

[0037] Then the function returns the element with the most occurrences and its number of occurrences, and the value of the first element is the mode;

[0038] Directly modify the data in the table on the front-end web page for subsequent calculation and output.

[0039] In one embodiment, the Bayesian network structure is displayed, and the influence degree between the read influence factors is matrix-operated through expert scoring, and the influence factor with the greatest influence degree is found as the root node and displayed after the leaf node with the least influence degree to form a fifth layer structure, thereby obtaining the final Bayesian network structure;

[0040] The specific operations of the matrix calculation process are as follows:

[0041] (1) Form a direct impact matrix;

[0042] (2) normative direct impact matrix;

[0043] (3) Solve the comprehensive impact matrix;

[0044] (4) Calculate four metrics.

[0045] In one scheme, the expert scoring of the failure probability of the root node of the page includes: reading the root node obtained by the influence degree score calculation into a folder, and the expert scores the failure probability of each influencing factor in the range of [0.1,1] for the second time, and then derives the failure probability and effectiveness probability of each leaf node according to the Bayesian conditional probability formula.

[0046] In one scheme, the reverse deduction is performed to obtain the most probable causal chain of community resilience failure, and the Bayesian network model is read, starting from the central node, to find the causal chain that has the highest probability of failure of the overall community resilience among the four major categories of influencing factors.

[0047] Beneficial effects of the present invention:

[0048] The present invention realizes multifunctional computational analysis of factors affecting community resilience by introducing advanced technologies such as Bayesian networks, Leaky Noisy-OR models, and fuzzy comprehensive evaluation methods. It can more accurately and quickly evaluate the resilience of the community, accurately identify the key factors affecting community resilience, and make up for the shortcomings of existing technologies. In addition, the present invention also has a unique reverse deduction function, which can identify the largest causal chain that makes community resilience ineffective, providing an important reference for formulating targeted strategies to enhance community resilience. It has significant social benefits and broad application prospects for further improving the community's ability to resist disasters and protecting the lives and property of community residents. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flow chart of the method of the present invention;

[0050] Figure 2 The software structure and workflow diagram of the system of the present invention;

[0051] Figure 3 A classified list of system factors of the present invention;

[0052] Figure 4 Scoring the influence of sub-nodes of the system of the present invention;

[0053] Figure 5 Displaying the Bayesian network structure for the system of the present invention;

[0054] Figure 6 A scoring interface for the root node of the system of the present invention;

[0055] Figure 7 The final Bayesian network model diagram of the system of the present invention;

[0056] Figure 8 Reverse deduction of the system model of the present invention;

[0057] Fig. 9 This is a spider web analysis diagram of the system of the present invention;

[0058] Fig.10 It is a histogram analysis example of the system of the present invention;

[0059] Fig.11 Calculate and implement parameter configuration for the system of the present invention;

[0060] Fig.12 Modify the structure before the system of the present invention;

[0061] Fig.13 The modified structure of the system of the present invention;

[0062] Fig.14 Modify the previous model for the system of the present invention;

[0063] Fig.15 This is the modified model of the system of the present invention. DETAILED DESCRIPTION

[0064] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Typical embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in the present invention. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0065] Unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0066] Embodiment 1

[0067] like Figure 1 As shown, a multifunctional computational analysis method for community resilience influencing factors includes:

[0068] S1. Use decision laboratory analysis method to establish Bayesian network structure model;

[0069] The Bayesian network structure model is established as follows:

[0070] S101. Preliminary establishment of a Bayesian network based on the influencing factors of each category of community resilience given by experts;

[0071] The nodes in the directed acyclic graph of the Bayesian network represent random variables, which can be observable variables, hidden variables, unknown parameters, etc. If two nodes are connected by a single arrow, one of the nodes is the "cause (parents)" and the other is the "effect (children). The arrow connecting the two nodes represents that the two random variables have a causal relationship, or are unconditionally independent, so the two nodes will generate a conditional probability value.

[0072] For example, assuming that node E directly affects node H, that is, E→H, then a directed arc (E, H) from node E to node H is established with an arrow pointing from E to H, and the weight (that is, connection strength) is represented by the conditional probability P(H|E).

[0073] In short, the random variables involved in a system are plotted in a directed graph based on whether they are conditionally independent, forming a Bayesian network. It is mainly used to describe the conditional dependencies between random variables, with circles representing random variables and arrows representing conditional dependencies.

[0074] Let G = (L, E) represent a directed acyclic graph, where L represents the set of all nodes in the graph, and E represents the set of directed connecting segments, and let X = {x i}, i∈L, x i is the random variable represented by a node i in its directed acyclic graph. If X is a Bayesian network of a directed acyclic graph, the joint probability of X can be expressed as:

[0075]

[0076] Among them, pa(i) represents the previous associated node corresponding to a certain node, that is, "parents".

[0077] Each vertex in a Bayesian network corresponds to a random variable, and each variable is probabilistically independent of its non-inherited nodes in the graph. The graph captures the qualitative structure of probability distributions and is developed for efficient reasoning and decision making. While Bayesian networks can represent arbitrary probability distributions, they provide computable advantages for these distributions that can be represented with simple structures.

[0078] S102. Read the experts’ scores on the degree of influence of various influencing factors, and use the decision laboratory analysis method DEMATEL to calculate the final community resilience Bayesian network structure module.

[0079] The impact scores are as follows:

[0080] For continuous data, especially the calculation of the mode of group interval series, it is necessary to use the group interval data processing method in statistics, namely formulas 3-1 and 3-2.

[0081]

[0082] In order to determine the Bayesian network structure of public health and safety resilience, that is, the directed acyclic graph of influencing factor variables, it is necessary to first determine the causal relationship between each node, and the decision laboratory analysis method (DEMATEL for short) is a more suitable tool in this link.

[0083] This algorithm is a decision support method in the field of system engineering and management science. It uses graph theory and matrix operations to analyze the causal relationship and interactive influence between internal elements of the system in order to more clearly understand the structure and dynamic characteristics of the system. This method is particularly suitable for solving decision-making problems with multiple factors intertwined, interrelated and complex relationships. It is widely used in the field of disaster management and is also applicable to the study of factors affecting community resilience under public health emergencies. The degree of influence and causal properties between factors are presented by the transformation of mathematical matrices, and then the internal levels of each system factor are analyzed. Factors with high causality can be used as the root node of the BN network structure.

[0084] Experts in public health emergencies, urban communities, resilience and other related fields, as well as community management staff, were invited to fill out a scoring questionnaire based on their professional knowledge and past experience, using a five-level scale of 0-4, from low to high, representing none, low, medium, relatively high, and high, to quantify the degree of influence between two factors. Through the logical relationship and direct impact matrix between the elements in the system, the influence and influence of each element on other elements can be calculated, and the cause degree and centrality of each element can be calculated as the basis for constructing the model, thereby determining the causal relationship between elements and the status of each element in the system.

[0085] Influence, influence, centrality and cause are the four measurement criteria for the influence of four elements in the system. They are all calculated based on the comprehensive influence matrix.

[0086] Influence degree: refers to the sum of the values ​​of each row of the comprehensive influence matrix, which represents the comprehensive influence value of the corresponding elements in each row on all other elements.

[0087] Influence degree: refers to the sum of the values ​​of each column of the comprehensive influence matrix, which indicates the comprehensive influence value of all other factors on the corresponding factors of each column.

[0088] Centrality: The centrality of a factor is obtained by adding the influence and influence of the factor, which is recorded as the centrality, indicating the position of the factor in the evaluation index system and the size of its role.

[0089] Causality: The causality of a factor is obtained by subtracting its influence from its influence.

[0090] S2, through the default node P L The introduction of and calculation of the Bayesian conditional probability formula are used to obtain the conditional probability of failure of each non-root node;

[0091] The Noisy-OR model is a commonly used computational simplification tool in Bayesian networks, which is used to represent n independent parent nodes Z1, Z2, ..., Z n The relationship between the node and the child node Y. All nodes are binary variables with only two states: 0 and 1. 0 represents no failure; 1 represents failure. This model is a probabilistic logic model used in probabilistic graphical models to model the causal relationship between multivariate Boolean events. It is often used to describe the possibility of a series of conditional factors leading to the occurrence of an event, and takes into account the uncertainty of these conditional factors themselves.

[0092] In the Noisy-OR model, there is an event Y (usually representing a certain failure or the occurrence of a certain state), which has multiple conditional factors Z1 to Z k When no conditional factors occur (i.e., all Zi are 0), Y is very likely not to occur (Y = 0). However, each Zi The probability of occurrence is not completely deterministic, even if Z i It is also possible that due to the existence of "noise" (i.e. uncertainty), Y may not necessarily occur. In other words, as long as there is at least one Z i happens, Y is likely to happen, but even if all Z i If both occur, Y may not necessarily occur. The following formula is usually used to express its probability distribution.

[0093]

[0094] Among them, P i It is in Z i The conditional probability that Y will occur if Z i =1, due to the influence of noise, Y may not always be 1.

[0095] However, the Bayesian network of community resilience may not cover all the influencing factors of public health emergencies. The premise of the Noisy-OR model that "the failure of child nodes is independent of the remaining variables except the parent node" does not hold. Although these potential variables are not identified and screened as parent nodes, they may still cause deviations in the results. The Leaky Noisy-OR model can make up for this shortcoming because the default node Z is added. L , which is used to represent the remaining potential variables that are not considered but may cause the child nodes to fail. Even if all parent nodes Z i Neither is invalid, child node Y may also be in the default node Z L Failure in case of failure. L Indicates that only the default node P L The probability that child node Y fails when its parent node fails and other parent nodes do not fail.

[0096] S3, obtain the prior probability value of the root node of the community resilience BN network, that is, the failure probability of the corresponding influencing factor variable;

[0097] Fuzzy comprehensive evaluation method is a technology that uses fuzzy mathematical theory to comprehensively evaluate things or objects affected by multiple factors. The main steps of the algorithm are as follows:

[0098] 1. Define evaluation factors: First, determine the various evaluation indicators (U) in the evaluation system, such as cost, quality, service, etc.

[0099] 2. Determine the fuzzy set and membership function: set the corresponding evaluation value range for each evaluation indicator, and give the membership of each evaluation value to each evaluation level (such as excellent, good, medium, and poor).

[0100] 3. Construct a fuzzy evaluation matrix: Determine the fuzzy set membership corresponding to each factor based on the actual performance data of each factor.

[0101] 4. Determine the weight coefficient: Determine the relative importance weight (a1, a2, ..., an) of each evaluation index through expert scoring, analytic hierarchy process (AHP), entropy weight method, etc.

[0102] 5. Fuzzy synthesis operation: Use fuzzy logic operations (such as weighted average, centroid method, maximum membership principle, etc.) to perform comprehensive operations on the fuzzy evaluation results of each indicator to obtain the overall fuzzy evaluation vector.

[0103] 6. Defuzzification: Defuzzify the comprehensive fuzzy evaluation results to obtain the final quantitative evaluation score.

[0104] The main advantages of fuzzy comprehensive evaluation method are:

[0105] 1. Processing fuzzy information: Applicable to situations where the attributes and standards of the evaluation object are not clear enough and there is uncertainty and ambiguity.

[0106] 2. Comprehensive evaluation: It can integrate multiple independent or related factors to more comprehensively reflect the characteristics of the evaluation object.

[0107] 3. Flexibility in weight allocation: Different evaluation indicators can be flexibly assigned different weights according to actual conditions.

[0108] 4. Intuitive interpretation of results: The evaluation results can be expressed using language variables, which is easy to understand and accept.

[0109] Since the past data of public health emergencies are limited and difficult to obtain, it is difficult to accurately obtain the prior probability of the root node of the BN network based on historical sample data. Therefore, this paper uses an expert questionnaire combined with a fuzzy comprehensive evaluation method to obtain the prior probability value of the root node of the community resilience BN network, that is, the failure probability of the corresponding influencing factor variable. Referring to Wickens' seven-level judgment theory, the fuzzy language of the failure probability of the variable factor is divided into seven levels. The trapezoidal fuzzy number and evaluation level used are shown in Table 1.

[0110] Table 1

[0111]

[0112] Since experts have different understandings and experiences in community resilience and emergency management, the arithmetic average method is not convenient to eliminate the differences and obtain accurate results. Different weight indicators are assigned to each expert, and the weighted average method is used to process the fuzzy rating number. By comparing the evaluation results of each expert with the average value, the closer the expert is, the greater the weight. Assuming that the fuzzy evaluation number of an expert is F k =(Fk1 , F k2 , F k3 , F k4 ), where K = 1, 2, ..., n;

[0113] (1) Solve the arithmetic mean F of n expert trapezoidal fuzzy numbers a ;

[0114]

[0115] (2) Solve F k and the arithmetic mean F a distance;

[0116]

[0117] (3) Solve F k and the arithmetic mean F a similarity;

[0118]

[0119] (4) Solving the expert evaluation weight ω k ;

[0120]

[0121] (5) Solve the weighted comprehensive fuzzy evaluation number F;

[0122]

[0123] (6) Use the mean area method to solve the fuzzy set and obtain the prior probability of the root node;

[0124]

[0125] S4. Based on the Bayesian network model, the maximum causal chain that causes community resilience to fail is obtained through reverse deduction. Starting from the central node, find the causal chain that has the highest probability of failure of total community resilience among the four major influencing factors, which is convenient for intuitive analysis of the factors affecting community resilience.

[0126] Embodiment 2:

[0127] like Figure 2 As shown, a multifunctional computing and analysis system for factors affecting community resilience is applicable to the method described, and the analysis system includes: displaying a factor classification list, reading and modifying expert sub-node influence scores, displaying a Bayesian network structure, reading expert scores of page root node failure probabilities, displaying a final Bayesian network model, performing reverse deduction to obtain the most probable cause chain of community resilience failure, spider web diagram analysis of the influence of each factor, and modifying parameter configuration.

[0128] The main function of the system is to initially establish a Bayesian network through the influencing factors of each category of community resilience given by experts, and then read the expert scores on the degree of influence between each influencing factor, and use the DEMATEL method to calculate the final community resilience Bayesian network structure. Further, based on the expert scores on the failure probability of the root node influencing factor, the Bayesian conditional probability formula is used to calculate the failure probability of the influencing factor of each leaf node, and finally a complete Bayesian network model that can display the effectiveness and failure probability of the node is generated. At the same time, the degree of failure of each influencing factor can be intuitively analyzed through the spider web diagram of the failure probability of each parent node of the leaf node. According to the final Bayesian network model, the maximum causal chain that causes the failure of community resilience is obtained through the reverse deduction function. In addition, some parameters can be modified to change the calculation results in order to make accurate calculations and analysis for specific studies.

[0129] like Figure 3 As shown in the figure, based on the classification of various factors affecting community resilience provided by experts, a Bayesian network structure is initially established on the web page in the form of a tree diagram through HTML, CS and JavaScript code for intuitive analysis. Each node can dynamically display and retract the nodes of the hierarchy below it by clicking the mouse. The font size increases from the root node to the center node step by step, and each layer of nodes is arranged in the same horizontal position, which is convenient for clearly analyzing the relationship between the factors at each level. The subsequent Bayesian network models all maintain the above properties. Figure 3 This is a diagram showing the effects achieved on the page.

[0130] The software framework is built through the Django model, which is an advanced Python Web framework that follows the MTV (Model-Template-View) design pattern. It is mainly used to build backend services, and also has certain front-end capabilities. The backend mainly handles tasks such as data storage, business logic processing, permission control, and API interface provision. It comes with a lightweight template system for generating HTML responses and supports integration with modern front-end frameworks and libraries.

[0131] like Figure 4As shown, the web page program design adopts the BS architecture front-end and back-end interaction, that is, the expert scoring file in the project program folder is read through the back-end code, and each data in the table is traversed. The mode of the data is calculated through Python and JavaScript specific functions, and the database or other data processing capabilities are used to cooperate. If there are multiple modes, the average is calculated and then rounded. If there is only one mode, it is directly taken as the output result, and then read into the page table for subsequent calculations, that is, processed in the Web back-end environment, and this part of the logic is processed in the view function. The results are returned to the front-end for display using front-end and back-end interactions (such as AJAX requests, API interface design, etc.). Seeking the mode of n Excel table data through Python is essentially to find the element with the most occurrences in the data set. Use Python's built-in collections module and numpy library to find the mode, create a Counter object to count the number of occurrences of each element, and then return the element with the most occurrences and its number of occurrences through the function. The value of the first element is the mode. At the same time, you can also use the front-end web page to directly modify the data in the table for subsequent calculations and output.

[0132] For continuous data, especially the calculation of the mode of group interval series, it is necessary to use the group interval data processing method in statistics.

[0133]

[0134] like Figure 5 As shown, the Bayesian network structure is displayed. Through the expert scoring function in the previous step, the influence degree between the read influencing factors is matrix-calculated, and the influencing factor with the greatest influence degree is found as the root node to be displayed after the leaf node with the least influence degree to form the fifth layer structure, and the final Bayesian network structure is obtained.

[0135] The specific operations of the matrix calculation process are as follows:

[0136] (1) Forming a direct impact matrix

[0137] The questionnaire scores are summed up and averaged and then rounded to the nearest integer. The matrix of the table data is the direct impact matrix A = [a ij ]n×n, where a ij Represents the row factor X i Parallel Factor X j The degree of influence of ij =0.

[0138] (2) Standard direct impact matrix

[0139] In order to unify the dimensions and normalize the matrix directly, the maximum value normalization method is adopted, that is, each row of the matrix A is summed and the maximum value is taken, and then the standard matrix B is obtained according to the following formula = [b ij ]n×n, where b ij In the interval [0,1].

[0140]

[0141] (3) Solving the comprehensive impact matrix

[0142] A community is a micro-city unit with multiple entities and multiple elements. It contains complex and diverse network relationships, and the factors of community resilience are also multiple. Matrix A represents the direct impact relationship between factors, and matrix B is multiplied by itself to represent the increasing indirect impact between factors. The final formula for solving the comprehensive impact matrix T is as follows:

[0143] T=B(EB) -1

[0144] (4) Calculate four metrics

[0145] Four measurement indicators of factors affecting community public health resilience are calculated based on the matrix T. Among them, the influence degree f is the sum of the values ​​of each row of the matrix T, which refers to the comprehensive influence of the corresponding factor of each row on other factors, denoted as f i ;

[0146]

[0147] The influence degree e is the sum of the values ​​of each column of the matrix T, which refers to the comprehensive influence degree of each column corresponding factor by other factors, denoted as e i ;

[0148]

[0149] Centrality m i is the influence f i and the degree of influence i The sum of the factors refers to the role played by the factor in the evaluation system;

[0150] m i =f i +e i

[0151] Cause degree n i is the influence f i and the degree of influence i If the difference is greater than zero, it means that this factor has a greater impact on other factors and is a causal factor of community public health resilience. If it is less than zero, it means that this factor is more affected by other factors and is a result factor of community public health resilience.

[0152] n i =f i -e i

[0153] like Figure 6 As shown, the expert score of the failure probability of the root node of the page is read, and the root node calculated by the previous function is read into the folder. The expert scores the failure probability of each influencing factor in the range of [0.1,1] for the second time, and then the failure probability and effective probability of each leaf node are obtained according to the Bayesian conditional probability formula.

[0154] The Bayesian conditional probability formula is used to calculate the failure probability of the leaf node under the condition of the known failure probability of the root node. This formula embodies "reverse thinking", that is, the probability distribution of the cause is reversely inferred from the known result.

[0155]

[0156] like Figure 7 As shown, the final Bayesian network model is displayed. The failure and validity probabilities of each node calculated by the backend in the previous step are filled into the previously established Bayesian network structure to form the final complete Bayesian network model. The nodes of the model include a name and a bracket below. The left side of the " / " in the bracket is the failure probability of the node, and the right side is the validity probability. Figure 7 This is the final Bayesian network model diagram.

[0157] like Figure 8 As shown, the reverse deduction is performed to obtain the most probable causal chain of community resilience failure. The above-mentioned total Bayesian network model is read, and starting from the central node, the causal chain with the highest probability of failure of the total community resilience among the four major categories of influencing factors is found. Each branch of the causal chain is blue, and the total causal chain is displayed in red, which is convenient for intuitive analysis of the causes affecting community resilience. Figure 8 This is the model reverse deduction interface.

[0158] Spider web diagram and tree diagram analysis of the influence degree of each factor. Read the failure probability of the four sub-nodes under "Community resilience of major public health events" to generate a spider web diagram; read all the root node influencing factors to generate a tree diagram. The color of each type of tree diagram is different for easy distinction. Select the corresponding node by clicking the ○ after the name, and then click the OK button to refresh the required spider web analysis graph, see Fig. 9 and 10 .

[0159] Modify parameter configuration. Experts can modify the PL parameters used in the above calculation process on this page. Fig.11 for this page.

[0160] Embodiment 3

[0161] Simulate the scoring of 10 experts and write them into the Excel table in the project folder and save them. At the same time, the scoring cannot be random, but needs to be based on the actual situation and be as close to the community's common sense and natural laws as possible. Each Excel table contains 20 matrix tables that need to be scored. The following table is one of the tables in one of the expert scoring files.

[0162] Table 2 Expert scoring table

[0163]

[0164]

[0165] At the same time, during the scoring process, you need to pay attention that the name and format of the table header cannot be changed. You can only modify the text data in the table. After scoring and saving the file, you need to close the Excel table and then restart the web page to update the read table.

[0166] After that, the webpage updates the Bayesian network structure and uploads it to the front-end interface, updates the root node scoring table interface, and updates the root node failure scoring Excel table in the project folder. In the table, the failure probability of each factor is entered according to the actual situation. After saving and closing, it is read to the front-end interface for subsequent calculations, and finally a Bayesian network model with the failure probability and effective probability of each node is formed. Table 3 is the root node failure probability scoring table.

[0167] Table 3 Root node failure probability score

[0168]

[0169]

[0170] Experimental results and discussion on different scoring data

[0171] Results and discussion of the first scoring experiment

[0172] First, we analyze the data table after a certain scoring and compare it with the corresponding generated Bayesian network structure. Table 4 shows the scoring situation before modification. Fig.12 For the corresponding structure.

[0173] Table 4 Scoring table before modification

[0174]

[0175] Then modify the table data to increase the influence of "heating system flow satisfaction degree" on other factors (increase the number in this column), make it a root node, modify "average annual water outage time for households" to increase the influence of other factors (increase the number in this column), make it no longer a leaf node, and modify "average annual power outage time for households" to reduce the influence of other factors (decrease the number in this column), make it a new leaf node. The modified table data is shown in Table 5 below, and the expected structure is as follows Fig.13 It can be seen that the scoring and structure display functions of the table meet the requirements.

[0176] Table 5 Modified scoring table

[0177]

[0178] Results and discussion of the second scoring experiment

[0179] First, we analyze the data table after a certain scoring and compare it with the corresponding generated Bayesian network structure. We set the scoring data to a low failure probability value table (the failure probability is distributed between 0.1 and 0.3). Table 6 shows the scoring situation before the modification. Fig.14 is the corresponding Bayesian model. Then, the data that need to be scored are changed from low probability values ​​to high probability values ​​(failure probabilities are distributed between 0.7 and 1.0), as shown in Table 7, and a new Bayesian network model is generated, such as Fig.15 Conduct comparative analysis.

[0180] Table 6 Scoring table before modification

[0181]

[0182] Table 7 Modified scoring table

[0183]

[0184] It can be seen that according to the changes before and after the scoring, the failure probability of the community resilience of the final central node for major public health events has changed as expected, proving that the model is feasible. It is found that if the second root node failure probability score data is too small, it will basically not affect the final central node failure probability. Therefore, the expert scoring does not need to be too conservative to prevent the analysis from being affected by the low sensitivity of the program.

[0185] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0186] It should be understood that the detailed description of the technical solutions of the present invention by means of the preferred embodiments is illustrative rather than restrictive. A person skilled in the art may modify the technical solutions described in the embodiments, or replace some of the technical features by equivalents, based on reading the specification of the present invention; and these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multifunctional computational analysis method for factors affecting community resilience, characterized by: The method comprises: using a decision laboratory analysis method to establish a Bayesian network structure model; By default node P L The introduction of and calculation of the Bayesian conditional probability formula are used to obtain the conditional probability of failure of each non-root node; Obtain the prior probability value of the root node of the community resilience BN network, that is, the failure probability of the corresponding influencing factor variable; Based on the Bayesian network model, the maximum causal chain that causes community resilience to fail is derived through reverse deduction.

2. The multifunctional calculation and analysis method for community resilience influencing factors according to claim 1 is characterized by: The Bayesian network structure model is established as follows: A Bayesian network was initially established based on the influencing factors of each category of community resilience given by experts; the experts' scores on the degree of influence of each influencing factor were read, and the final community resilience Bayesian network structure module was calculated using the decision laboratory analysis method DEMATEL.

3. The multifunctional calculation and analysis method of community resilience influencing factors according to claim 1 is characterized by: The conditional probability of non-root node failure adopts the Leaky Noisy-OR model: Based on the Noisy-OR model, the default node Z is added L , which is used to represent the remaining potential variables that are not considered but will cause the child node to fail.

4. The multifunctional calculation and analysis method for community resilience influencing factors according to claim 1 is characterized by: The prior probability value of the root node of the community resilience BN network is obtained by combining an expert questionnaire with a fuzzy comprehensive evaluation method to obtain the prior probability value of the root node of the community resilience BN network; The seven-level judgment theory is used to divide the fuzzy language of variable factor failure probability into seven levels; Different weight indicators are assigned to each expert, and the weighted average method is used to process the fuzzy rating number. By comparing the evaluation results of each expert with the average value, the closer the expert is, the greater the weight is. Suppose the fuzzy evaluation number of an expert is F k =(F k1 , F k2 , F k3 , F k4 ), where K = 1, 2, ..., n; (1) Solve the arithmetic mean F of n expert trapezoidal fuzzy numbers a ; (2) Solve F k and the arithmetic mean F a distance; (3) Solve F k and the arithmetic mean F a similarity; (4) Solving the expert evaluation weight ω k ; (5) Solve the weighted comprehensive fuzzy evaluation number F; (6) Use the mean area method to solve the fuzzy set and obtain the prior probability of the root node; 5. A multifunctional computing and analysis system for community resilience influencing factors, the analysis system being applicable to the method according to any one of claims 1 to 4, characterized in that: The analysis system includes: displaying a factor classification list, reading and modifying expert sub-node influence scores, displaying a Bayesian network structure, reading expert scores on page root node failure probabilities, displaying a final Bayesian network model, performing reverse deduction to obtain the most probable cause chain of community resilience failure, spider web diagram analysis of the influence of each factor, and modifying parameter configuration.

6. The multifunctional computing and analyzing system for community resilience influencing factors according to claim 5 is characterized by: The display factor classification list is based on the classification of various influencing factors of community resilience provided by experts, and a Bayesian network structure is established on the web page in the form of a tree diagram; each node dynamically displays and retracts the hierarchical nodes below the node by clicking the mouse, and the font size increases step by step from the root node to the central node, and each layer of nodes is arranged in the same horizontal position.

7. The multifunctional computing and analyzing system for community resilience influencing factors according to claim 5 is characterized by: The reading and modifying expert sub-node influence degree scoring adopts the group distance data processing method in statistics; Read the expert scoring file in the project program folder, traverse each data in the table, use Python's built-in collections module and numpy library to find the mode, and create a Counter object to count the number of occurrences of each element; Then the function returns the element with the most occurrences and its number of occurrences, and the value of the first element is the mode; Directly modify the data in the table on the front-end web page for subsequent calculation and output.

8. The multifunctional computing and analyzing system for community resilience influencing factors according to claim 5 is characterized by: The display Bayesian network structure is to perform matrix operations on the influence degree between the read influence factors through expert scoring, find out the influence factor with the greatest influence degree as the root node and display it after the leaf node with the least influence degree to form a fifth layer structure, and obtain the final Bayesian network structure; The specific operations of the matrix calculation process are as follows: (1) Form a direct impact matrix; (2) normative direct impact matrix; (3) Solve the comprehensive impact matrix; (4) Calculate four metrics.

9. The multifunctional computing and analyzing system for community resilience influencing factors according to claim 5, characterized in that: The expert scoring of the failure probability of the root node of the page includes: reading the root node obtained by the influence degree score calculation into a folder, and the expert scores the failure probability of each influencing factor in the range of [0.1,1] for the second time, and then derives the failure probability and effectiveness probability of each leaf node according to the Bayesian conditional probability formula.

10. The multifunctional computing and analyzing system for community resilience influencing factors according to claim 5, characterized in that: The reverse deduction is performed to obtain the most probable causal chain of community resilience failure, read the Bayesian network model, and starting from the central node, find the causal chain that has the highest probability of failure of the total community resilience among the four major categories of influencing factors.

Images