A method, device and equipment for identifying and evaluating social stability risks of a major construction project and a storage medium

Abstract

The application discloses a kind of major construction project social stability risk identification and evaluation method, device, equipment and storage medium.The method comprises: obtaining the characteristic attribute of project to be evaluated and multi-modal social data;Using "prior knowledge-posterior data" dual-source driving method, construct initial risk benchmark framework;Based on the framework design structured questionnaire survey, fusion multi-source data, construct correlation model.Model construction uses prior-posterior dynamic fusion to carry out bayesian parameter estimation, introduces adjustment effect analysis, multi-group analysis, path time-varying weighting to carry out model enhancement.Calculate risk transmission path, identify key risk dimension and key influence path;Key risk index is built, and model dynamic self-adapting update is realized to realize early warning and risk evaluation.The application realizes the accurate identification of risk factor, the quantitative revelation of transmission mechanism and the dynamic early warning and evaluation of risk state, significantly improves the scientificity and timeliness of major construction project social stability risk evaluation.

Description

Technical Field

[0001] This invention relates to the field of risk assessment and management technology, specifically to a method, apparatus, equipment, and storage medium for identifying and evaluating social stability risks of major construction projects, and in particular, a method for predicting risks by associating the project to be evaluated with social data. Background Technology

[0002] Major construction projects (such as airports, high-speed railways, nuclear power plants, chemical plants, and waste incineration plants) play a significant role in promoting regional economic and social development. However, they often trigger social conflicts and even mass incidents due to issues such as land acquisition and demolition, environmental impact, and the distribution of benefits, posing a risk to social stability. Therefore, conducting a scientific and systematic social stability risk assessment before project decision-making and implementation is crucial.

[0003] Traditional risk assessment methods often rely on expert scoring or simple indicator weighting, which are highly subjective and fail to reveal the inherent, non-linear interactions between complex risk factors. Firstly, existing technologies lack a mechanism for linking project data with social data. Traditional risk assessments often analyze the project in isolation, failing to effectively utilize a large amount of existing social feedback data from similar projects (such as public opinion, complaints, and records of mass incidents), resulting in a lack of cross-sectional comparison and historical experience to support risk identification. Secondly, data sources are limited, mainly relying on expert scoring or questionnaires, which are time-consuming, costly, and fail to reflect dynamically changing public sentiment and social concerns. Furthermore, assessment results are static, often "one-off," and cannot be linked to real-time social dynamics during project construction, making continuous risk monitoring and dynamic early warning difficult.

[0004] Therefore, how to effectively link the projects to be evaluated with massive amounts of social data, and identify and quantify the complex mechanisms of action among risk factors, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the problems of lack of historical reference, single data source, and static results in the social stability risk assessment of major construction projects in existing technologies, this invention aims to provide a method, apparatus, equipment, and storage medium for social stability risk identification and assessment that can link the project to be assessed with social data, establish a model for identifying and quantifying the complex interaction relationships between risk factors.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for identifying and assessing social stability risks in major construction projects, including:

[0008] Obtain the characteristic attributes of the major construction projects to be evaluated. The project characteristic attributes include at least the project type, construction scale, site location, land acquisition and demolition scope, and engineering and technical solutions, and then vectorize them.

[0009] Acquire multimodal social data associated with the projects to be evaluated;

[0010] The social data is associated and mapped with the characteristic attributes of the project to be evaluated to construct an initial risk benchmark framework for the project to be evaluated.

[0011] Based on the initial risk benchmark framework, a structured questionnaire survey was designed and implemented to obtain stakeholder perception data for the project to be evaluated.

[0012] The acquired multi-source data is preprocessed and fused to extract the observed variable data corresponding to multiple preset risk dimensions. A measurement model is constructed between the observed variable data and the latent variables of the corresponding risk dimensions, and a correlation model between different risk dimensions is constructed to jointly form a correlation analysis framework between risk factors.

[0013] The processed observed variable data are input into the correlation model for parameter estimation to obtain the standardized path coefficients between the latent variables of each risk dimension, so as to quantitatively characterize the mutual influence relationship and influence weight between different risk dimensions.

[0014] Based on the standardized path coefficients, key risk dimensions and key impact paths that affect the overall level of social stability risk are identified.

[0015] Based on the key risk dimensions and their impact weights, and combined with the real-time public opinion characteristics extracted from the social data, a social stability risk assessment report and dynamic risk warning information are generated for the project under evaluation.

[0016] Optionally, the acquisition of multimodal social data associated with the project to be evaluated includes data on similar historical projects that have been built or are under construction, social sentiment data, public safety incident data, socio-economic operation data, and group activity trend data.

[0017] Historical project data refers to existing or under-construction projects of the same type as the project to be evaluated, used as analogies to build an initial risk benchmark framework; and historical project data that are different from the project to be evaluated but similar in key characteristics (such as scale, location characteristics, sensitivity characteristics, etc.), which are obtained by screening from a cross-industry project database through a multi-dimensional similarity matching algorithm.

[0018] Relevant public opinion text data, complaint records, and reports of mass incidents were collected from news websites, social media (Weibo, WeChat, forums), and government complaint platforms. Natural language processing (NLP) technology was used to perform sentiment analysis (identifying positive / negative emotions) and topic clustering (identifying core issues of public concern, such as demolition compensation, environmental pollution, and employment impact) on the collected text data, extracting high-frequency risk keywords and patterns of negative sentiment evolution. Stakeholder complaint data and regional socio-economic statistics, including population density, resident income levels, and historical petition records, were obtained from statistical and urban planning departments. Geographic Information System (GIS) technology was used to spatially overlay the project site with surrounding environmentally sensitive areas (residential areas, schools, hospitals, and ecological protection zones) to identify potential environmental sensitivity risks.

[0019] Optionally, the construction of the initial risk benchmark framework for the project to be evaluated adopts a dual-source driven approach of "prior knowledge + posterior data," including:

[0020] Prior knowledge driven layer: Based on historical data of similar projects and historical data of dissimilar projects, the risk patterns and empirical rules contained therein are used to generate a prior knowledge risk factor framework applicable to the project to be evaluated through structured comparison and induction. The prior framework includes verified risk dimensions and their typical observation indicators.

[0021] The posterior data-driven layer integrates regional aggregated public opinion sentiment index, regional background index, and real-time monitoring data. Through cross-case knowledge transfer, regional sensitivity embedding, and real-time signal injection, it dynamically expands and adaptively adjusts the prior framework to generate a posterior enhanced risk factor framework that integrates multi-source information. This process maps risk keywords, focus areas, and sentiment tendencies extracted from social data to the specific characteristics of the project under evaluation, constructing a posterior enhanced risk factor framework for that project. (For example, if multiple comparable projects trigger public opinion crises due to "opaque demolition compensation standards," then the initial risk framework for the project under evaluation should emphasize and strengthen relevant observation indicators in the "land acquisition and resettlement" dimension.)

[0022] Optionally, the acquisition of stakeholder perception data for the project under evaluation is based on an initial risk benchmark framework, using a structured questionnaire designed to be customized to the specific characteristics of the project. The questionnaire items not only cover the traditional eight risk dimensions (project procedures, land acquisition and demolition, technological and economic factors, natural environment, social environment, economic interests, media sentiment, and specific risks), but are also customized based on specific risk points revealed by social data, enabling the questionnaire to more accurately capture the potential risks of the project under evaluation. Questionnaires are distributed and collected from stakeholders of the project (including residents in the project's impact area, surrounding businesses, local government departments, and experts); the collected questionnaire data is cleaned and screened, and reliability and validity analyses are performed to verify the validity and reliability of the data.

[0023] Optionally, before extracting the observed variable data based on the preset risk dimension, the method further includes: incorporating dynamic indicators such as public opinion sentiment index and complaint hotspot density extracted from social data as auxiliary observed variables into the dataset; conducting exploratory factor analysis on the fused data to verify the consistency between the risk dimension revealed by the social data and the preset risk dimension; and conducting confirmatory factor analysis on the verified risk dimension to confirm the measurement relationship between the observed variables and latent variables.

[0024] Optionally, the risk dimensions may include at least the following: project procedures, land acquisition and resettlement, project technical and economic management, impact on the local natural environment, impact on the local social environment, impact on local economic interests, media sentiment, and specific risks.

[0025] Optionally, the steps of constructing the correlation model include: proposing multiple initial hypothetical models about the influence relationship between latent variables of different risk dimensions, wherein the path relationship part of the hypothetical model is derived from the correlation pattern between public opinion focus revealed by social data; fitting the multiple initial hypothetical models with the fused observed variable data respectively; and selecting the optimal model as the final correlation model for risk analysis by comparing the goodness-of-fit indices of multiple models, wherein the goodness-of-fit indices include at least the chi-square value, the normalized fit index, the non-normalized fit index, and the comparative fit index.

[0026] Optionally, the step of inputting the processed observed variable data into the association model for parameter estimation includes: transforming the prior knowledge risk factor framework extracted from historical project data into a prior distribution of model parameters, using a Bayesian structural equation model for parameter estimation, and fusing prior information with stakeholder perception data obtained through questionnaires to obtain posterior path coefficients and their confidence intervals; wherein the form of the prior distribution and the hyperparameters are determined based on the mean and variance of the factor loadings of historical data.

[0027] Optionally, a prior-posterior dynamic fusion coefficient can be set in the Bayesian structural equation model, and adaptive fusion weights can be calculated based on the multi-dimensional similarity between the project to be evaluated and historical projects. ,in This is a vector of similarity between projects. The weights are learnable weight vectors; the posterior parameter estimates are weighted averages of the prior mean and the data estimates, and the fusion weights are dynamically adjusted based on similarity.

[0028] Optionally, the correlation model incorporates moderating effect analysis, using regionally aggregated public opinion sentiment index or regional background index as moderating variables to construct a moderating model with moderating effects. These moderating variables influence the path coefficients between specific risk dimensions. Its mathematical form is:

[0029]

[0030] in, Let m be the moderating variable. This is the moderating effect coefficient, used to quantify the moderating effect of regional characteristics on the intensity of risk transmission.

[0031] Optionally, the association model incorporates multi-group analysis, which divides the sample into multiple subgroups based on the regional background index, estimates the association model for each subgroup, obtains the differential path coefficients under different regional types, and identifies significant differences between groups through parameter invariance tests.

[0032] The correlation model dynamically weights the path coefficients, including: calculating the real-time volatility index of each observed variable based on real-time dynamic monitoring data; constructing a dynamic weighting function for the path coefficients; and applying time-varying weights to the path coefficients in the correlation model according to the real-time volatility index to obtain dynamic path coefficients for real-time assessment of changes in risk transmission intensity. The dynamic weighting function takes the following form:

[0033]

[0034] in, As the baseline path coefficient, This represents the real-time volatility index of the corresponding risk dimension at time t. It is a weighted function.

[0035] Optionally, after obtaining new questionnaire survey data or real-time monitoring data each time, the correlation model calculates the correction index and expected parameter changes of the correlation model; if the correction index exceeds a preset threshold, model optimization is automatically triggered. The model path settings are adjusted or parameter constraints are released based on the expected parameter changes, achieving adaptive updates of the correlation model.

[0036] Optionally, the identification of key risk dimensions and key impact paths includes: calculating the standardized total effect of each path, i.e., the sum of direct and indirect effects; ranking the risk dimensions based on the size of the total effect, and selecting the dimensions whose total effect exceeds a preset threshold as key risk dimensions; and identifying the indirect path that contributes the most to the total effect as the key impact path.

[0037] Optionally, identifying key risk dimensions and key impact paths includes constructing a key risk index ( )formula:

[0038]

[0039] in, For the first The combined weight of each key dimension is obtained by normalizing the total effect of that dimension. Real-time monitoring of the observed variable data for each key dimension is performed to calculate the real-time... The value is compared with a dynamic threshold to trigger an alert signal of the corresponding level.

[0040] Optionally, generating a social stability risk assessment report and dynamic risk warning information for the project under evaluation includes: comparing and analyzing the risk assessment results of the project under evaluation with the historical risk levels of comparable projects extracted from social data to assess the relative risk level of the project under evaluation; generating multi-level risk warning signals based on key risk dimensions identified by the correlation model and combined with public opinion fluctuations in real-time social data; and matching and recommending targeted risk mitigation measures and public communication plans from a pre-set strategy library for different key risk dimensions.

[0041] This invention also discloses a device for identifying and assessing social stability risks in major construction projects, the device comprising:

[0042] The project data acquisition module is used to acquire the characteristic attributes of the major construction projects to be evaluated;

[0043] The social data acquisition module is used to acquire multimodal social data associated with the project to be evaluated;

[0044] The association mapping module is used to associate and map social data with the characteristic data of the project to be evaluated, and to build an initial risk benchmark framework.

[0045] The questionnaire survey and processing module is used to design, implement, and process structured questionnaire survey data for the projects to be evaluated.

[0046] The data fusion and feature extraction module is used to preprocess and fuse multi-source data to extract observed variable data;

[0047] The model building and management module is used to build and manage the relationship model that includes measurement models and structural models;

[0048] The risk calculation and analysis module is used to input observed variable data into the parameter estimation of the correlation model and identify key risk dimensions and impact paths;

[0049] The report generation and early warning module is used to generate risk assessment reports and dynamic risk early warning information.

[0050] The present invention also discloses an electronic device, comprising: one or more processors; and a memory for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method for identifying and assessing social stability risks of major construction projects as described in the present invention.

[0051] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, such that when the program is executed by a processor, it implements the method for identifying and evaluating social stability risks of major construction projects as described in the present invention.

[0052] Compared with the prior art, the present invention has the following advantages and technical effects:

[0053] By achieving dual-source synergy between prior knowledge and posterior data, an initial risk factor framework integrating multi-source information is ultimately formed. The prior layer provides a stable risk dimension structure and core observation indicators to ensure the historical inheritance and reliability of the framework. The posterior layer achieves horizontal expansion through cross-case knowledge transfer, vertical deepening through regional sensitivity embedding, and dynamic evolution through real-time signal injection.

[0054] Multi-source data fusion improves the accuracy of assessments by combining objective project characteristics and attributes, large-scale social data, and targeted questionnaire survey data. This ensures both the breadth of the assessment (covering the entire online public opinion) and the depth of the assessment (reaching specific stakeholders), greatly improving the comprehensiveness and accuracy of risk identification.

[0055] Dynamic early warning capability: By combining static structured analysis results with dynamic social data monitoring, the system can continuously track and provide real-time early warnings of social stability risks. When negative fluctuations related to key risk dimensions appear in real-time public opinion data, the system can automatically generate early warning information, giving decision-makers valuable response time.

[0056] Scientific identification of key risk points: Through path analysis of correlation models, it is possible to quantitatively reveal "which risk factors affect overall social stability through which paths", thereby accurately identifying key links that require priority intervention and achieving precise allocation of limited resources.

[0057] Highly replicable technology: This invention provides a complete technical process, from data acquisition, association mapping, model building to result output, all of which can be automated or semi-automated through computer systems, making it easy to promote and apply in different types of major construction projects. Attached Figure Description

[0058] Figure 1 The overall flowchart of a method for identifying and evaluating social stability risks in major construction projects provided in an embodiment of the present invention is shown.

[0059] Figure 2 A schematic diagram of a social stability risk identification and evaluation model framework for major construction projects provided in an embodiment of the present invention.

[0060] Figure 3 A modular structure diagram of a device for identifying and evaluating social stability risks in major construction projects, provided as an embodiment of the present invention.

[0061] Figure 4 This is a schematic diagram of an electronic device for implementing the method according to an embodiment of the present invention. Detailed Implementation

[0062] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present application will now be described in detail with reference to the accompanying drawings and embodiments. The steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowcharts, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0063] Example 1:

[0064] like Figure 1 As shown, a method for identifying and evaluating social stability risks in major construction projects may include, but is not limited to, the following steps S1 to S8.

[0065] Step S1: Obtain the characteristic attributes of the project to be evaluated. The characteristic attributes include at least the project type, construction scale, site location, land acquisition and demolition scope, and engineering technical solution, and then vectorize them.

[0066] Step S2: Obtain multimodal social data associated with the project to be evaluated;

[0067] Step S3: Associate and map the social data with the characteristic attributes of the project to be evaluated to construct an initial risk benchmark framework for the project to be evaluated.

[0068] Step S4: Based on the initial risk benchmark framework, design and implement a structured questionnaire survey to obtain stakeholder perception data for the project to be evaluated;

[0069] Step S5: Preprocess and fuse the acquired multi-source data, extract the observed variable data corresponding to multiple preset risk dimensions, construct a measurement model between the observed variable data and the latent variables of the corresponding risk dimensions, and construct a correlation model between different risk dimensions to jointly form a risk factor correlation analysis framework.

[0070] Step S6: Input the processed observed variable data into the correlation model for parameter estimation to obtain the standardized path coefficients between the latent variables of each risk dimension, so as to quantitatively characterize the mutual influence relationship and influence weight between different risk dimensions.

[0071] Step S7: Based on the standardized path coefficients, identify the key risk dimensions and key impact paths that affect the overall social stability risk level.

[0072] Step S8: Based on the key risk dimensions and their impact weights, and combined with the real-time public opinion characteristics extracted from the social data, generate a social stability risk assessment report and dynamic risk warning information for the project to be evaluated.

[0073] As an example of the present invention, in step S1, during the process of obtaining the characteristic attributes of the project to be evaluated, a social stability risk assessment project for a waste-to-energy incineration project is provided. The project information is as follows: the total processing capacity of municipal solid waste is 1200 t / d, and the waste heat generated by waste incineration is used for power generation, including systems for waste receiving, storage, incineration, flue gas treatment, and power generation. After completion, the project can process 438,000 tons of waste annually, with an average annual power generation of 2.54 × 10⁸ kWh. The total investment is 840 million yuan. The current site of the project is wasteland and forest, with a total land area of ​​approximately 65,180 m², requiring the relocation of 82 households. The project is 4.8 km from the planned edge of the urban area and approximately 7.7 km from the airport. The project's process configuration includes two 600 t / d (tons / day) waste incineration lines, two 16 MW (megawatt) condensing steam turbines, and two 18 MW (megawatt) generator sets; equipped with a medium-temperature sub-high-pressure steam boiler (450℃, 6.4 MPa). The flue gas treatment process adopts a combined process of "SNCR in-furnace denitrification (ammonia water) + PNCR (dry high-efficiency denitrification) + semi-dry acid removal + circulating dry process (slaked lime) + activated carbon adsorption + bag filter dust collection + flue gas recirculation". ...

[0074] The vectorized result (simplified feature vector) is: [1, 1200, 470, 6.5, 1, 1, 1, 1]

[0075] 1: Project Type, "1" represents "Waste Incineration Power Generation", "0" represents other types. 1200: Daily Processing Capacity (tons / day), representing the construction scale numerically. 840: Total Investment (million RMB), converting 840 million RMB into a numerical value. 6.5: Land Area (ten thousand square meters), converting the land area numerically. 0: Site Location, "1" represents "Industrial Park", "0" represents "Other Areas". 1,1,1,1: Engineering Technology Scheme, representing whether "SNCR Denitrification", "Semi-dry Deacidification", "Activated Carbon Adsorption", and "Bag Filter" are adopted respectively, "1" indicates "Yes".

[0076] This allows for the calculation and comparison of these feature vectors, enabling quick identification of the project and precise comparison with historical projects in the database.

[0077] As an example of the present invention, in step S2, during the acquisition of multimodal social data associated with the project to be evaluated, n similar waste-to-energy incineration projects (e.g., with a total processing capacity of over 1200 t / d, involving land acquisition and demolition, and located in the urban-rural fringe) are selected from the historical project database as analogy objects. Their risk factor records, stability assessment reports, original questionnaire survey data, and risk event records are extracted.

[0078] Data collection on dissimilar projects: Based on a multi-dimensional similarity matching algorithm (project size, population density, location characteristics, etc.), m dissimilar projects with similar characteristics (such as large-scale chemical industrial park projects and large-scale logistics hub projects) are selected, and records of risk factors that occurred during their construction are extracted. For example, the "credibility of groundwater monitoring data" in a chemical industrial park project once raised public concerns, and this risk point has reference value for waste incineration projects that also involve environmental monitoring.

[0079] Data on online public opinion in the project area was collected, including: over 1,200 news media reports; over 8,500 social media discussion posts; over 300 related complaint records from government complaint platforms; and over 2,000 discussion posts from surrounding residents' forums. Natural language processing (NLP) technology was used to process this data. For example, sentiment analysis identified approximately 3,200 posts with negative sentiment, accounting for about 27%, with peak negative sentiment occurring during the announcement period of land acquisition and demolition for the project and the peak construction period. Thematic clustering extracted high-frequency risk keywords, including "low compensation standards," "poor quality resettlement housing," "noise pollution," "water pollution," "concerns about toxic emissions," "opposition to the site selection," "concerns about ecological damage," "concerns about health impacts," and "concerns about property devaluation." Analysis of the spatiotemporal distribution of negative public opinion revealed that residents within a 1km radius of the project site exhibited the strongest negative sentiment, which was highly correlated with construction progress.

[0080] Regional socio-economic background data collection: Obtain population data (population density, age structure), economic data (per capita income, employment rate), historical petition data (number and distribution of environmental complaints in the past five years), and environmentally sensitive point data (distribution of schools, hospitals, and residential areas) from the statistics department in Area A where the project is located.

[0081] Real-time dynamic monitoring data: During the project's progress, we continuously collect real-time public opinion data, complaint records, etc., for dynamic early warning.

[0082] In an optional embodiment of the present invention, step S3 associates and maps the social data with the characteristic attributes of the project to be evaluated, constructs an initial risk benchmark framework for the project to be evaluated, and adopts a dual-source driven method of "prior knowledge + posterior data", including but not limited to the following steps S301 to S302:

[0083] Step S301: Construct a prior knowledge-driven layer. Based on historical data of similar projects and historical data of dissimilar projects, and the inherent risk patterns and empirical rules, a prior knowledge risk factor framework applicable to the project under evaluation is generated through structured comparison and induction. This prior framework includes verified risk dimensions and their typical observation indicators. For example, based on n similar completed waste-to-energy incineration projects and m different types of projects with comparable characteristics, a prior knowledge risk factor framework applicable to the project under evaluation, containing verified risk dimensions and typical observation indicators, is generated through structured comparison and induction. The framework includes exogenous latent variables (risk dimensions, representing antecedents) and endogenous latent variables (representing outcomes), each of which can be measured through several observation variables.

[0084] As an example of this invention, exogenous latent variables (antecedent risk dimensions) may include: 1. Land acquisition and resettlement, which can be measured by three observed variables: X1 - reasonableness score of compensation standard, X2 - satisfaction score of resettlement plan, X3 - impact score of relocation transition period; 2. Environmental impact perception, which can be measured by three observed variables: X4 - degree of concern about flue gas emissions, X5 - degree of concern about leachate pollution, X6 - degree of concern about noise impact; 3. Health risk perception, which can be measured by three observed variables: X7 - degree of concern about carcinogenic risk, X8 - degree of concern about respiratory system impact, X9 - degree of concern about children's health impact; 4. Impact on local economy, which can be measured by three observed variables: X 10 -Expected impact on property values, X 11 -Expected impact on employment opportunities, X 12 - Expected impact on local businesses; ...

[0085] Endogenous latent variables (outcome risk dimension) may include: 1. Media sentiment risk, which can be measured by three observed variables: Y1 - perceived frequency of negative reports, Y2 - perceived social media discussion intensity, Y3 - public opinion guidance tendency; 2. Social stability risk, which can be measured by three observed variables: Y4 - probability of mass incidents, Y5 - probability of individual extreme behavior, Y6 - willingness to file petitions and complaints; ...

[0086] Step S302: Construct a posterior data-driven layer. Based on the prior knowledge risk factor framework, integrate regional aggregated public opinion sentiment index, regional background index and real-time monitoring data. Through cross-case knowledge transfer, regional sensitivity embedding and real-time signal injection, dynamically expand and adaptively adjust the prior framework to generate a posterior enhanced risk factor framework that integrates multi-source information.

[0087] This is a process of supplementation, correction, enhancement, and expansion. For example, regarding the theme of "low compensation standards," the focus on the "land acquisition and resettlement" dimension is strengthened, considering the scale of the 82 households involved in the project under evaluation; regarding the theme of "noise pollution," the dimension of "impact on the local social environment" is strengthened, considering the residential areas involved in the project; and regarding the theme of "water pollution," the dimension of "impact on the local natural environment" is strengthened, considering the characteristics of groundwater protection involved in the project. Based on the regional aggregated public opinion sentiment index, it was found that villagers in village M are highly sensitive to the topic of "dioxins" (the keyword "dioxins" accounted for 23% of historical complaints), therefore, "your level of concern about dioxin emissions" is added to the risk factor framework. Based on the regional background index (the region has a high degree of population aging), "your level of concern about the health impact on the elderly" is added.

[0088] As an example of the present invention, in step S4, based on the initial risk benchmark framework, a structured questionnaire survey is designed and implemented to obtain stakeholder perception data for the project to be evaluated. During this process, the questionnaire can be customized to the specific characteristics of the project. For example, for a waste-to-energy incineration project, the following customized observation indicators were specifically added: "Are you satisfied with the currently announced demolition compensation standards?" (addressing the widespread compensation disputes in public perception); "Are you worried that the project's construction and operation will affect the quality of surrounding water sources?" (addressing the specific risk of the project's landfill affecting groundwater); "How tolerant are you of nighttime construction noise?" (addressing the impact of construction on nearby residential areas).

[0089] Questionnaires were distributed and collected from stakeholders of the project to be evaluated (including residents in the project's impact area, surrounding businesses, local government departments, experts and scholars).

[0090] The collected questionnaire data was cleaned and screened, and reliability and validity analyses were performed to verify the validity and reliability of the data. For example, this project designed a Likert scale questionnaire with 35 items. 500 questionnaires were distributed to stakeholders within a 3-kilometer radius of the project, including residents, village committee officials, environmental organizations, and local government departments, with 468 valid questionnaires collected. The sample distribution included: 28% of residents within 1 kilometer of the project, 35% of residents within 1-2 kilometers, and 37% of residents within 2-3 kilometers; 82 households involved in land acquisition and demolition were also included.

[0091] In an optional embodiment of the present invention, during step S5, which constructs a model of the correlation between different risk dimensions to form a risk factor correlation analysis framework, a waste incineration power generation project integrates questionnaire survey data with dynamic indicators extracted from social data (such as the negative public opinion index of surrounding villages) to construct a structured risk feature dataset containing 468 samples and 35 observed variables (see example in Table 1). Based on risk theory and preliminary analysis, three competing hypothetical models regarding the influence relationships between eight risk dimensions are proposed. The above dataset is then imported into the three models for fitting. Based on the goodness-of-fit index (χ² / df=2.08, NFI=0.93, CFI=0.96, RMSEA=0.05), the optimal model is selected.

[0092] Table 1: Example of a structured risk feature dataset (partial)

[0093]

[0094] In an optional embodiment of the present invention, step S6 inputs the processed observed variable data into the parameter estimation of the correlation model to obtain the standardized path coefficients between the latent variables of each risk dimension. Here, latent variables refer to abstract concepts that cannot be directly measured and need to be indirectly reflected through multiple observable indicators (i.e., observed variables or manifest variables). In the social stability risk assessment of waste-to-energy incineration projects, "land acquisition and resettlement risk" is a latent variable. It cannot be directly measured by a single number, but its level can be comprehensively reflected through the scores of multiple specific questions in a questionnaire (such as "reasonableness of compensation standards" and "satisfaction with resettlement plans").

[0095] Suppose we have defined the following four risk dimensions (latent variables) for this project:

[0096]

[0097] A measurement model is a mathematical expression that describes how each latent variable is measured by its corresponding observed variable. For example, for the latent variable "land acquisition, demolition, and resettlement"... Its measurement model can be written as:

[0098]

[0099] in:

[0100] λ is the factor loading, representing the strength of the observed variable's response to the latent variable (usually estimated from the data); δ is the measurement error.

[0101] Furthermore, based on analysis or empirical assumptions, the risks associated with "land acquisition, demolition, and resettlement" ( This will exacerbate the "impact on the local social environment" ( "Impact on the local natural environment" This will also exacerbate the "impact on the local social environment". "Impact on the local social environment" This will further exacerbate the "media public opinion risk" ( The following system of linear equations represents the structural model:

[0102]

[0103] The coefficients (γ, β) are path coefficients, representing the influence between latent variables. Their magnitude and significance reveal the transmission mechanism between risk factors. For example, if =0.6 and significant, indicating that for every unit increase in land acquisition and demolition risk, the social and environmental risk will directly increase by 0.6 units.

[0104] In an optional embodiment of the present invention, step S6, in the process of estimating the parameters of the correlation model, includes one or a combination of the following methods:

[0105] S601. Prior risk patterns extracted from historical data of similar projects are transformed into prior distributions of model parameters. Bayesian structural equation modeling is used for parameter estimation, fusing prior information with current questionnaire data to obtain posterior path coefficients and their confidence intervals. The form and hyperparameters of the prior distribution are determined based on the mean and variance of factor loadings from historical data. For example, historical data shows that the mean factor loading of "concern about flue gas emissions" on "perception of environmental impact" is approximately 0.75, which is set as the prior mean, and the variance is set to 0.05. Bayesian structural equation modeling is used for parameter estimation, fusing prior information with data from 468 current questionnaires to obtain posterior path coefficients.

[0106] S602. Moderating effect analysis: Using regionally aggregated public opinion sentiment index or regional background index as moderating variables, a structural model with moderating effects is constructed. These moderating variables influence the path coefficients between specific risk dimensions. For example, the path "perceived environmental impact → social stability risk" is moderated by the regional public opinion index, mathematically expressed as:

[0107]

[0108] in This is a regional public opinion index. This is the moderating effect coefficient, used to quantify the strengthening or weakening effect of public opinion atmosphere on risk transmission.

[0109] S603. Introduce multi-group analysis, dividing the sample into multiple subgroups based on regional background indices, estimating the structural equation model for each subgroup, obtaining differentiated path coefficients under different regions, scenarios, and condition types, and identifying significant differences between groups through parameter invariance tests. For example, based on "distance from the project," the sample is divided into three subgroups (<1km, 1-2km, 2-3km), the structural equation model for each group is estimated, and the differences in risk structure at different distances are identified through parameter invariance tests.

[0110] S604. Dynamic weighting of path coefficients: Based on real-time dynamic monitoring data, a dynamic weighting function for path coefficients is constructed. The path coefficients in the structural equation model are time-varyingly weighted according to the real-time fluctuation index to obtain dynamic path coefficients.

[0111] S605. After obtaining new questionnaire survey data or real-time monitoring data, calculate the correction index and expected parameter changes of the structural equation model. If the correction index exceeds a preset threshold, model optimization is automatically triggered. Adjust the model path settings or release parameter constraints according to the expected parameter changes to achieve adaptive updates of the correlation model.

[0112] As an example: In subsequent stages (such as the construction or operation phase), after obtaining new questionnaire data, the correction index of the structural equation model is calculated. If the correction index exceeds a threshold (e.g., MI>10), model optimization is automatically triggered, adjusting the path settings according to expected parameter changes (e.g., adding a "construction impact" dimension) to achieve adaptive model updates.

[0113] In an optional embodiment of the present invention, the optimal model is run to obtain standardized path coefficients. As an example, key findings include: the path coefficient of "impact of land acquisition and resettlement" on "the local social environment" is 0.52 (P<0.001); the path coefficient of "impact of the local natural environment" on "media sentiment" is 0.48 (P<0.001); and the direct impact coefficient of "media sentiment" on "overall social stability risk" is 0.61 (P<0.001).

[0114] Furthermore, the standardized total effect of each path is calculated, which is the sum of the direct and indirect effects; the risk dimensions are ranked based on the size of the total effect, and the dimensions with a total effect exceeding a preset threshold are selected as key risk dimensions; at the same time, the indirect path that contributes the most to the total effect is identified as the key impact path.

[0115] As an example, the total effect (direct effect + indirect effect) of each risk dimension is calculated as follows:

[0116]

[0117] The key risk dimensions identified were: environmental impact perception (total effect 0.62), media sentiment (0.38), and health risk perception (0.35). Among them, the indirect effect of "environmental impact perception" through "health risk perception" (0.20) accounted for 32.3% of the total effect, forming a key impact path: environmental impact perception → health risk perception → social stability risk.

[0118] As an example, the formula for constructing a key risk index is as follows: =0.42 × Environmental Perception Score + 0.26 × Public Opinion Score + 0.32 × Health Perception Score. The weights are obtained by normalizing the total effect. Real-time monitoring of key dimension observation variable data is performed, real-time KRI values ​​are calculated, and compared with dynamic thresholds to trigger corresponding level of early warning signals.

[0119] Step S8: Based on the key risk dimensions and their impact weights, and combined with the real-time public opinion characteristics extracted from the social data, generate a social stability risk assessment report and dynamic risk warning information for the project to be evaluated.

[0120] As an example, a waste-to-energy incineration project, based on the above analysis, generated a risk assessment report, pointing out that the most critical risk dimensions of the project are "land acquisition and resettlement" and "impact on the local natural environment." Combined with real-time social data monitoring, when the system detects that the negative public opinion index of a village surrounding the project exceeds a preset threshold (e.g., 0.7), it automatically generates an orange warning signal and recommends: 1) immediately initiating a special communication meeting for the village; 2) strengthening the harmless treatment and environmental monitoring of leachate and publicizing the results to the public.

[0121] Example 2:

[0122] This invention also provides an apparatus for implementing the above method, referring to... Figure 3 ,include:

[0123] Project data acquisition module 301 is used to acquire feature data of the project to be evaluated;

[0124] The social data collection module 302 is used to collect historical data and social feedback data of the analog project;

[0125] The association mapping module 303 is used to construct the initial risk baseline framework;

[0126] Questionnaire Survey and Processing Module 304 is used to design, implement, and process questionnaire survey data;

[0127] The data fusion and feature extraction module 305 is used to construct a structured risk feature dataset;

[0128] The model building and management module 306 is used to build and filter relationship models;

[0129] Risk calculation and analysis module 307 is used to calculate path coefficients and identify key risk dimensions;

[0130] The report generation and early warning module 308 is used to generate evaluation reports and implement dynamic early warnings.

[0131] Example 3:

[0132] This invention also provides an electronic device, with reference to... Figure 4 It includes at least one processor 401, a memory 402, and a computer program stored in the memory and executable on the processor. When the processor 401 executes the program, it can implement the method described in Embodiment 1.

[0133] Example 4:

[0134] This invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in Embodiment 1.

[0135] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for identifying and assessing social stability risks in major construction projects, characterized in that, Includes the following steps: Obtain the characteristic attributes of the major construction projects to be evaluated. The project characteristic attributes include at least the project type, construction scale, site location, land acquisition and demolition scope, and engineering and technical solutions, and then vectorize them. Acquire multimodal social data associated with the projects to be evaluated; The social data is associated and mapped with the characteristic attributes of the project to be evaluated to construct an initial risk benchmark framework for the project to be evaluated. Based on the initial risk benchmark framework, a structured questionnaire survey was designed and implemented to obtain stakeholder perception data for the project to be evaluated. The acquired multi-source data is preprocessed and fused to extract the observed variable data corresponding to multiple preset risk dimensions. A measurement model is constructed between the observed variable data and the latent variables of the corresponding risk dimensions, and a correlation model between different risk dimensions is constructed to jointly form a correlation analysis framework between risk factors. The processed observed variable data are input into the correlation model for parameter estimation to obtain the standardized path coefficients between the latent variables of each risk dimension, so as to quantitatively characterize the mutual influence relationship and influence weight between different risk dimensions. Based on the standardized path coefficients, key risk dimensions and key impact paths that affect the overall level of social stability risk are identified. Based on the key risk dimensions and their impact weights, and combined with the real-time public opinion characteristics extracted from the social data, a social stability risk assessment report and dynamic risk warning information are generated for the project under evaluation.

2. The method according to claim 1, characterized in that, The acquisition of multimodal social data associated with the project to be evaluated includes data on similar historical projects that have been built or are under construction, social sentiment data, public safety incident data, socio-economic operation data, and data on the status of group activities. Historical project data refers to existing or under-construction projects of the same type as the project to be evaluated, used as analogies to build an initial risk benchmark framework; and historical project data that are different from the project to be evaluated but similar in key characteristics (such as scale, location characteristics, sensitivity characteristics, etc.), which are obtained by screening from a cross-industry project database through a multi-dimensional similarity matching algorithm. Related social sentiment text data, complaint records, and reports of mass incidents were collected from news portals, social media platforms (including Weibo, WeChat, and forums) and government complaint platforms. Natural language processing technology is used to perform sentiment analysis and topic clustering on the collected text data to identify the core risk issues of public concern. Obtain stakeholder complaint data and regional socio-economic statistics for the project area from statistical and urban planning departments, including population density, residents' income levels, and historical petition records; Using geographic information system (GIS) technology, the site selection location of the project under evaluation is spatially overlaid with surrounding environmentally sensitive points (including residential areas, schools, hospitals, and ecological protection areas) to identify potential environmental sensitivity risks.

3. The method according to claim 1, characterized in that, The construction of the initial risk benchmark framework for the project under evaluation adopts a dual-source driven approach of "prior knowledge + posterior data," including: Prior knowledge driven layer: Based on historical data of similar projects and historical data of dissimilar projects, the risk patterns and empirical rules contained therein are used to generate a prior knowledge risk factor framework applicable to the project to be evaluated through structured comparison and induction. The prior framework includes verified risk dimensions and their typical observation indicators. The posterior data-driven layer integrates regional aggregated public opinion sentiment index, regional background index, and real-time monitoring data. Through cross-case knowledge transfer, regional sensitivity embedding, and real-time signal injection, it dynamically expands and adaptively adjusts the prior framework to generate a posterior enhanced risk factor framework that integrates multi-source information.

4. The method according to claim 1, characterized in that, The acquisition of stakeholder perception data for the project to be evaluated includes: Based on the initial risk benchmark framework, a structured questionnaire was designed, which can be customized to the specific characteristics of the project to be evaluated. Questionnaires were distributed and collected from stakeholders of the project to be evaluated (including residents in the project's impact area, surrounding businesses, local government departments, experts and scholars); The collected questionnaire data were cleaned and screened, and reliability and validity analyses were performed to verify the validity and reliability of the data.

5. The method according to claim 1, characterized in that, Before extracting the observed variable data based on the preset risk dimension, the process also includes: Dynamic indicators such as public opinion sentiment index and complaint hotspot density extracted from social data are incorporated into the dataset as auxiliary observation variables; Exploratory factor analysis was conducted on the fused data to verify the consistency between the risk dimensions revealed by the social data and the pre-set risk dimensions; Confirmatory factor analysis was performed on the validated risk dimensions to confirm the measurement relationship between observed variables and latent variables.

6. The method according to claim 1, characterized in that, The step of inputting the processed observed variable data into the association model for parameter estimation includes: transforming the prior knowledge risk factor framework extracted from historical project data into a prior distribution of model parameters; using a Bayesian structural equation model for parameter estimation; and fusing prior information with stakeholder perception data obtained through questionnaires to obtain posterior path coefficients and their confidence intervals. The form of the prior distribution and the hyperparameters are determined based on the mean and variance of the factor loadings in the historical data. In the Bayesian structural equation model, prior-posterior dynamic fusion coefficients are set, and adaptive fusion weights are calculated based on the multi-dimensional similarity between the project to be evaluated and historical projects. ,in This is a vector of similarity between projects. The weights are learnable weight vectors; the posterior parameter estimates are weighted averages of the prior mean and the data estimates, and the fusion weights are dynamically adjusted based on similarity.

7. The method according to claim 1, characterized in that, The correlation model incorporates moderating effect analysis. Using regionally aggregated public opinion sentiment index or regional background index as moderating variables, a moderating model with moderating effects is constructed. These moderating variables influence the path coefficients between specific risk dimensions. Its mathematical form is: in, Let m be the moderating variable. This is the moderating effect coefficient, used to quantify the moderating effect of regional characteristics on the intensity of risk transmission.

8. The method according to claim 1, characterized in that, The association model incorporates multi-group analysis, which divides the sample into multiple subgroups based on the regional background index, estimates the association model for each subgroup, obtains the differential path coefficients under different regional types, and identifies significant differences between groups through parameter invariance tests.

9. The method according to claim 1, characterized in that, The correlation model dynamically weights the path coefficients, including: calculating the real-time volatility index of each observed variable based on real-time dynamic monitoring data; constructing a dynamic weighting function for the path coefficients; and applying time-varying weights to the path coefficients in the correlation model according to the real-time volatility index to obtain dynamic path coefficients for real-time assessment of changes in risk transmission intensity. The dynamic weighting function takes the form of: in, As the baseline path coefficient, This represents the real-time volatility index of the corresponding risk dimension at time t. It is a weighted function.

10. The method according to claim 1, characterized in that, The correlation model calculates a correction index and expected parameter changes each time new questionnaire data or real-time monitoring data is obtained. If the correction index exceeds a preset threshold, model optimization is automatically triggered. The model path settings are adjusted or parameter constraints are released based on the expected parameter changes, achieving adaptive updates to the correlation model.

11. The method according to claim 1, characterized in that, The identification of key risk dimensions and key impact paths includes: calculating the standardized total effect of each path, i.e., the sum of direct and indirect effects; ranking the risk dimensions based on the size of the total effect, and selecting the dimensions whose total effect exceeds a preset threshold as key risk dimensions; and identifying the indirect path that contributes the most to the total effect as the key impact path.

12. The method according to claim 1, characterized in that, The identification of key risk dimensions and key impact paths includes the construction of key risk indices ( )formula: Where, α k The comprehensive weight of the k-th key dimension is obtained by normalizing the total effect of that dimension; the observed variable data of the key dimensions are monitored in real time, and the real-time calculation is performed. The value is compared with a dynamic threshold to trigger an alert signal of the corresponding level.

13. The method according to claim 1, characterized in that, The process of generating a social stability risk assessment report and dynamic risk warning information for the project under evaluation includes: comparing and analyzing the risk assessment results of the project under evaluation with the historical risk levels of comparable projects extracted from social data to assess the relative risk level of the project under evaluation; generating multi-level risk warning signals based on key risk dimensions identified by the correlation model and combined with public opinion fluctuations in real-time social data; and matching and recommending targeted risk mitigation measures and public communication plans from a pre-set strategy library for different key risk dimensions.

14. A device for identifying and assessing social stability risks in major construction projects, characterized in that, include: The project data acquisition module is used to acquire the characteristic attributes of the major construction projects to be evaluated; The social data acquisition module is used to acquire multimodal social data associated with the project to be evaluated; The association mapping module is used to associate and map social data with the characteristic data of the project to be evaluated, and to build an initial risk benchmark framework. The questionnaire survey and processing module is used to design, implement, and process structured questionnaire survey data for the projects to be evaluated. The data fusion and feature extraction module is used to preprocess and fuse multi-source data to extract observed variable data; The model building and management module is used to build and manage the relationship model that includes measurement models and structural models; The risk calculation and analysis module is used to input observed variable data into the parameter estimation of the correlation model and identify key risk dimensions and impact paths; The report generation and early warning module is used to generate risk assessment reports and dynamic risk early warning information.

15. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method as described in any one of claims 1 to 13.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 13.

Images