Intelligent evaluation and zoning optimization method for coastal zone ecological resilience

Abstract

The application provides a kind of coastal zone ecological resilience intelligent evaluation and zoned optimization method, it is related to geographic information technology field, including: based on social-ecological system theory, from ecological resistance, ecological resilience and ecological stability three dimensions, the index system of coastal zone ecological resilience evaluation is constructed, the pre-processing and standardization of the obtained multi-source data are carried out;Based on the index system and the data after standardization, the ecological resilience index of each evaluation unit in the research area is calculated by using comprehensive weighting and approximation ideal solution sorting method, and the spatial distribution pattern of ecological resilience is analyzed;With the ecological resilience index as the target variable, the social, economic and environmental factors as the characteristic variables, a machine learning regression model is constructed, and the non-linear influence and interaction of each characteristic variable on the ecological resilience index are analyzed using the interpretable artificial intelligence framework to identify the key driving factors and thresholds;Based on the analysis results, the zoned ecological resilience optimization control strategy is developed.

Description

Technical Field

[0001] This application belongs to the field of geographic information technology, specifically relating to an intelligent evaluation and zoning optimization method for coastal ecological resilience. Background Technology

[0002] As a crucial zone for land-sea interaction, the coastal zone is both a biodiversity-rich area and a region of high concentration of human economic activity. Under the dual pressures of global climate change and human development, its ecosystems exhibit extremely high vulnerability and sensitivity. Traditional ecological protection and assessment methods often focus on the static stability of ecosystems, failing to effectively characterize the dynamic capacity of systems to maintain function, absorb shocks, and achieve restructuring and recovery when disturbed—that is, "ecological resilience." Therefore, scientifically understanding and quantitatively assessing the ecological resilience of coastal areas has become an urgent need for achieving sustainable development and precise management.

[0003] Currently, related technological research and practice suffer from the following limitations: First, in terms of evaluation system construction, existing indicator systems are often simplistic or insufficiently adapted to the complex characteristics of the coastal zone's "natural-social" composite system. They fail to systematically integrate and reflect the multidimensional attributes of its resistance (resistance to disturbance), resilience (repair after damage), and stability (sustainable structural function), leading to biased assessment results. Second, regarding assessment models, traditional methods often employ linear models or subjective weighting, making it difficult to capture the complex nonlinear relationship between ecological resilience and socio-economic driving factors. Furthermore, the models themselves are often considered "black boxes," lacking interpretability of their internal decision-making mechanisms, resulting in assessment results that are "what but not why." Finally, in terms of application, most studies stop at the static measurement and spatial display of regional ecological resilience levels, lacking in-depth analysis of key driving mechanisms and the ability to generate mechanism-based, spatially differentiated optimization strategies, thus failing to provide direct and actionable decision support for the refined governance of coastal zones. Summary of the Invention

[0004] This application provides an intelligent evaluation and zoning optimization method for coastal ecological resilience to solve one of the aforementioned technical problems.

[0005] The technical solution adopted in this application is as follows: This application provides an intelligent assessment and zoning optimization method for coastal zone ecological resilience, including: S1. Based on the socio-ecological system theory, an evaluation index system for the ecological resilience of coastal areas is constructed from three dimensions: ecological resistance, ecological resilience, and ecological stability. The acquired multi-source data are preprocessed and standardized. S2. Based on the aforementioned indicator system and standardized data, the ecological resilience index of each evaluation unit within the study area is calculated using a comprehensive weighting and approximation of ideal solution ranking method, and the spatial distribution pattern of ecological resilience is analyzed. S3. Using the ecological resilience index as the target variable and socio-economic environmental factors as feature variables, construct a machine learning regression model, and use an interpretable artificial intelligence framework to analyze the nonlinear effects and interactions of each feature variable on the ecological resilience index, and identify key driving factors and thresholds; based on the analysis results, formulate regional ecological resilience optimization and regulation strategies.

[0006] According to one embodiment of this application, in step S1, the indicators of the ecological resilience dimension include: population density, built-up area ratio, per capita water resource consumption, carbon emissions per unit of GDP, industrial wastewater discharge per unit of GDP, and carbon dioxide emissions per unit of GDP. The indicators for the ecological resilience dimension include: comprehensive utilization rate of water resources, land resource optimization index, comprehensive utilization rate of solid waste, harmless treatment rate of garbage, centralized sewage treatment rate, and air quality excellence rate. The indicators for the ecological stability dimension include: green coverage rate of built-up areas, per capita green space area, biodiversity index, environmental investment ratio, vegetation coverage, and the proportion of water conservancy environment and public facilities management.

[0007] According to one embodiment of this application, in step S2, the comprehensive weighting and approximation of the ideal solution ranking method specifically involves: determining the objective weights of each indicator using the entropy weight method, constructing a weighted standardized decision matrix; calculating the Euclidean distance from each evaluation unit to the positive and negative ideal solutions, and then applying the formula:

[0008] Calculate the ecological resilience index for each evaluation unit. ,in To determine the distance to the ideal solution, This is the distance to the negative ideal solution.

[0009] According to an embodiment of this application, step S2, analyzing the spatial distribution pattern of ecological resilience, includes: using kernel density analysis to characterize the spatial clustering density of ecological resilience, and using global and local spatial autocorrelation analysis to identify the spatial correlation pattern of ecological resilience and high-value and low-value clustering areas.

[0010] According to one embodiment of this application, in step S3, the socio-economic environmental factors include at least one of the following: economic development level, financial investment intensity, industrial structure, education level, urbanization level, technological innovation capability, social consumption level, and infrastructure level.

[0011] According to one embodiment of this application, the machine learning regression model is an XGBoost model based on gradient boosting decision trees; and the interpretable artificial intelligence framework is a SHAP framework based on game theory.

[0012] According to one embodiment of this application, the analysis using the SHAP framework includes: calculating the SHAP value of each feature variable to quantify its average marginal contribution, and drawing a SHAP dependency graph to visualize the nonlinear impact and threshold effect of a single feature variable on the ecological resilience index.

[0013] According to one embodiment of this application, the analysis using the SHAP framework further includes: performing interaction analysis between feature variables to reveal the synergistic or antagonistic effects of different socioeconomic environmental factors on the ecological resilience index.

[0014] According to one embodiment of this application, step S3, formulating a zoned ecological resilience optimization and regulation strategy specifically includes: for the low-value clustering areas identified by the spatial analysis, and in combination with the negative impact threshold of the key driving factors, formulating a differentiated regulation strategy with at least one of the following as the core: green transformation of industrial structure, increased environmental investment, improved resource utilization efficiency, and strengthened ecological space protection.

[0015] A second aspect of this application provides an intelligent assessment and zoning optimization system for coastal ecological resilience, comprising: The indicator construction and preprocessing module is used to construct an evaluation indicator system for the ecological resilience of coastal areas from three dimensions: ecological resistance, ecological resilience, and ecological stability, based on the socio-ecological system theory, and to preprocess and standardize the acquired multi-source data. The assessment and spatial analysis module is used to calculate the ecological resilience index of each evaluation unit in the study area based on the aforementioned indicator system and standardized data, using a comprehensive weighting and approximation of ideal solution ranking method, and to analyze the spatial distribution pattern of ecological resilience. The analysis and optimization strategy generation module is used to construct a machine learning regression model with the ecological resilience index as the target variable and socio-economic environmental factors as feature variables, and to analyze the nonlinear effects and interactions of each feature variable on the ecological resilience index using an interpretable artificial intelligence framework, identify key driving factors and thresholds; and to formulate regional ecological resilience optimization and regulation strategies based on the analysis results.

[0016] Due to the adoption of the above technical solution, the beneficial effects achieved by this application are as follows: This application, through step S1, systematically constructs an indicator system based on socio-ecological system theory, encompassing three dimensions: ecological resilience, ecological resilience, and ecological stability. This approach overcomes the limitations of single-dimensional evaluation, ensuring that the assessment comprehensively covers the capabilities of coastal ecosystems in responding to stress, recovering from damage, and maintaining long-term balance, thus laying a solid foundation for subsequent scientific quantification.

[0017] In step S2, a comprehensive weighting and ideal solution approximation ranking method is used to aggregate multi-dimensional indicator data into a unified Ecological Resilience Index (ERI), achieving scientific and objective quantification of complex attributes. Simultaneously, spatial distribution pattern analysis of the index is conducted. This method allows the abstract resilience level to be visually presented spatially, clearly identifying high-value clusters (hotspots) and low-value clusters (cold spots) of ecological resilience, accurately revealing the heterogeneity within regions, and providing spatial targeting for subsequent regional policy implementation.

[0018] Step S3 constructs a machine learning regression model and analyzes it using an interpretable artificial intelligence framework. This combined approach effectively overcomes the shortcomings of traditional linear models and "black box" machine learning models. It not only accurately fits the complex nonlinear relationship between ecological resilience and its driving factors, but also quantifies the contribution of each feature variable (socioeconomic and environmental factors) through an interpretable framework, visualizes its nonlinear influence path and threshold effect, and analyzes the interactions between factors. This transforms model predictions into mechanistic explanations of "why it is so," accurately identifying the key levers and critical conditions that constrain or enhance ecological resilience.

[0019] Based on the analysis results of the driving mechanism in step S3, combined with the spatial pattern analysis in step S2, a regional ecological resilience optimization and regulation strategy is formulated. This approach ensures that the optimization strategy is not arbitrary but rooted in insights into deep-seated influencing mechanisms, and is designed differently for the prominent problems of different spatial regions (such as high- and low-resilience clusters). For example, for regions where the identified key negative driving factors exceed the threshold, targeted regulation (such as industrial transformation) can be enforced; for low-value clusters, environmental investment or ecological restoration can be prioritized. This achieves a closed loop of assessment, analysis, and optimization, greatly enhancing the decision support value and application effectiveness of the technical solution. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating an intelligent assessment and zoning optimization method for coastal ecological resilience provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application; Figure 3 The overall flowchart provided for the embodiments of this application; Figure 4 This is a schematic diagram of XGBoost training results and feature evaluation provided in an embodiment of this application; Figure 5 A schematic diagram of the SHAP-based driver dependency curves constructed for embodiments of this application; Figure 6 This is a schematic diagram of the SHAP-based interactive analysis provided for an embodiment of this application.

[0021] Figure label: 810, Processor; 820, Communication interface; 830, Memory; 840, Communication bus. Detailed Implementation

[0022] To more clearly illustrate the overall concept of this application, a detailed explanation is provided below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of this application is not limited to the specific embodiments disclosed below. It should be noted that, unless otherwise specified, the embodiments of this application and the features thereof can be combined with each other.

[0024] In this application, unless otherwise expressly specified and limited, the "above" or "below" of the second feature can mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. In the description of this specification, references to terms such as "an embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in one or more embodiments or examples.

[0025] Example 1 like Figure 1 As shown, a smart evaluation and zoning optimization method for coastal ecological resilience includes: S1. Based on the socio-ecological system theory, an evaluation index system for the ecological resilience of coastal areas is constructed from three dimensions: ecological resistance, ecological resilience, and ecological stability. The acquired multi-source data are preprocessed and standardized.

[0026] Specifically, based on the socio-ecological system theory, which posits that coastal areas are complex systems formed by the close coupling and interaction of natural ecosystems and human social systems, ecological resilience assessments must consider both natural and socio-economic attributes.

[0027] The dimensional division transforms the abstract concept of ecological resilience into three measurable and analyzable functional dimensions: Ecological resilience refers to the ability of a system to maintain its original structure and function without drastic decline or collapse when subjected to external pressures or disturbances (such as pollution emissions or excessive resource consumption). This dimension mainly selects indicators that characterize the intensity of the pressure source or the system's exposure, and is usually a negative indicator.

[0028] Ecological resilience refers to the ability of a system to recover to its original or near-original state after being damaged, through self-repair or with the help of human intervention. This dimension mainly selects indicators that characterize the efficiency of resource recycling and the effectiveness of pollution control, and these are usually positive indicators.

[0029] Ecological stability refers to a system's ability to maintain its structural integrity, functional continuity, and long-term equilibrium, which depends on the system's underlying ecological conditions and continuous external safeguards. This dimension primarily selects indicators that characterize the underlying ecological conditions and long-term safeguard capabilities, typically positive indicators.

[0030] Based on the above three dimensions, a series of secondary indicators that can specifically represent the core characteristics of each dimension are selected to form a structured, multi-level evaluation indicator system.

[0031] For example: Ecological resilience indicators: For example, "industrial wastewater discharge per unit of GDP" can be selected as a secondary indicator. This indicator belongs to the ecological resilience dimension. Specifically, it represents the amount of industrial wastewater discharged for every unit of economic output (GDP) generated in a coastal area. Higher discharge volumes indicate greater immediate pollution pressure on the water environment from economic activities, higher intensity of disturbance to the system, and greater challenges to its resilience. This indicator is negative; lower values ​​are more conducive to ecological resilience.

[0032] Ecological resilience indicators: For example, "comprehensive utilization rate of solid waste" can be selected as a secondary indicator. This indicator belongs to the ecological resilience dimension. Specifically, it means the proportion of solid waste that is comprehensively utilized relative to the total amount of solid waste generated in coastal areas. A higher utilization rate indicates a stronger ability of the system to recycle waste into resources, and a stronger ability to achieve material balance and functional recovery within the system through recycling after being disturbed by resource consumption and waste emissions. This indicator is a positive one.

[0033] Ecological stability indicators: For example, "environmental investment ratio" can be selected as a secondary indicator. This indicator belongs to the ecological stability dimension. Specifically, it represents the proportion of GDP invested in environmental pollution control, ecological protection, and restoration in coastal areas. This proportion reflects the long-term capital commitment of the social system to maintain and enhance the stability of the natural ecosystem. The higher the investment ratio, the stronger the external support the system receives, and the more conducive it is to maintaining its long-term stability. This indicator is a positive one.

[0034] Preprocessing: The raw data obtained comes from diverse and heterogeneous sources such as statistical yearbooks, environmental bulletins, and remote sensing imagery. These data differ in statistical definitions, time scales, spatial resolutions, and units, making direct comprehensive calculations impossible. The purpose of preprocessing is to unify the data foundation and eliminate incomparable factors.

[0035] Standardization Purpose and Principle: Evaluation indicators have different physical meanings and dimensions, and their numerical values ​​and ranges of variation vary significantly. Directly using the original values ​​will allow indicators with larger absolute values ​​to dominate the comprehensive evaluation. The purpose of standardization (normalization) is to eliminate the influence of indicator dimensions, mapping the original values ​​of each indicator to a unified, dimensionless numerical range through mathematical transformation. This ensures that all indicators are on the same scale, are comparable, and can be used for comprehensive calculations such as weighted summation.

[0036] Standardization method: Min-Max Normalization is used. This method compresses the original data to the [0, 1] interval through linear transformation. Different transformation formulas are used for positive indicators (the larger the indicator value, the better) and negative indicators (the smaller the indicator value, the better) to ensure that the standardized data are consistent in direction.

[0037] For example: Example (Standardization of Positive Indicators): Take "Green Coverage Rate of Built-up Areas" as an example, which is a positive indicator. Assume that in all evaluation units of a certain study area, the maximum value of this indicator is 60%, and the minimum value is 10%. For one evaluation unit, its original value is 40%. Using range standardization: First, calculate the range (maximum value - minimum value) = 60% - 10% = 50%. Then, subtract the minimum value from the original value of this unit, and divide by the range, i.e., (40% - 10%) / 50% = 0.6. After standardization, the "Green Coverage Rate of Built-up Areas" indicator value for this unit becomes 0.6. This process transforms the value of this indicator for all evaluation units to between 0 and 1, and the larger the original value, the closer the standardized value is to 1.

[0038] Example (Negative Indicator Standardization): Take "population density" as an example, which is a negative indicator. Assume the maximum value is 2000 people / km² and the minimum value is 200 people / km². For one evaluation unit, its original value is 1000 people / km². Using the negative indicator formula of range standardization, we transform it as follows: First, calculate the range = 2000 - 200 = 1800. Then, subtract the original value of the unit from the maximum value, and divide by the range, i.e., (2000 - 1000) / 1800 ≈ 0.556. After standardization, the "population density" indicator value of this unit becomes 0.556. This process also compresses the value to the [0,1] range, but the logic is that the smaller the original value (population density), the closer the standardized value is to 1, conforming to the unified standard of "the larger the value, the better the evaluation," for subsequent calculations.

[0039] S2. Based on the aforementioned indicator system and standardized data, the ecological resilience index of each evaluation unit within the study area is calculated using a comprehensive weighting and approximation of ideal solution ranking method, and the spatial distribution pattern of ecological resilience is analyzed.

[0040] Specifically, based on the constructed indicator system and preprocessed standardized data, a comprehensive weighting and ideal solution approximation ranking method is used to calculate the ecological resilience index of each evaluation unit. Comprehensive weighting refers to determining weights by combining the objective information content of the indicators (such as through entropy weighting) to overcome subjective arbitrariness. The ideal solution approximation ranking method is a multi-objective decision analysis method. Its principle is to define a theoretical "positive ideal solution" (a vector composed of the optimal values ​​of each indicator) and a "negative ideal solution" (a vector composed of the worst values ​​of each indicator) for each indicator. By calculating the distance between each actual evaluation unit and these two ideal solutions in multi-dimensional space, the degree of closeness of the unit to the optimal state is comprehensively evaluated, and this closeness is quantified as the ecological resilience index. Subsequently, based on the calculated ecological resilience indices of each unit, spatial distribution pattern analysis is conducted. This includes using kernel density estimation methods to estimate the density of point or area index values ​​through a moving smooth window to visualize the continuous spatial distribution and clustering hotspots of ecological resilience; at the same time, spatial autocorrelation analysis (such as global and local Moran indices) is used to quantitatively examine and identify whether the ecological resilience index values ​​exhibit significant clustering patterns in space (such as high-value clusters or low-value clusters), thereby revealing their spatial dependence and heterogeneity characteristics.

[0041] For example, consider the coastal zone of a peninsula in a certain province, assuming it includes multiple evaluation units such as City A, City B, and City C. First, the entropy weight method is used to analyze all standardized indicators such as "population density," "green coverage rate," and "wastewater treatment rate." The calculation logic of the entropy weight method is: the greater the data difference of an indicator among all units, the more information it contains, and the higher its objective weight. For example, if the data on "carbon emissions per unit of GDP" differs greatly among cities, then the indicator has a higher weight; if the data on "per capita green space area" are very similar, then its weight is lower. After obtaining the weights, a weighted decision matrix is ​​constructed. Next, positive and negative ideal solutions are determined. A positive ideal solution is a virtual optimal unit, composed of the optimal values ​​of all indicators (maximum values ​​for positive indicators and minimum values ​​for negative indicators); a negative ideal solution is the opposite. For example, a positive ideal solution might consist of the lowest "population density," the highest "green coverage rate," and the highest "wastewater treatment rate," etc. Subsequently, the Euclidean distance between the weighted vector of Qingdao's indicators and the two ideal solution vectors is calculated, yielding a distance to the positive ideal solution and a distance to the negative ideal solution. Finally, the distance to the negative ideal solution is divided by the sum of these two distances, and the resulting ratio (between 0 and 1) is Qingdao's ecological resilience index. A higher ratio indicates that its ecological resilience is closer to the ideal optimal state. After calculating the indices for all cities, spatial analysis is performed. Kernel density analysis treats each city's index value as a spatial point, generating a continuous surface map through a smoothing function. Darker areas in the map represent a greater concentration of cities with high ecological resilience. Simultaneously, spatial autocorrelation analysis calculates a global Moran's index. If this index is significantly positive, it indicates that cities with high ecological resilience tend to be close to each other, and cities with low ecological resilience also tend to be close to each other. Furthermore, through local analysis, "high-high" clusters (those with high indices themselves and their neighbors also have high indices) and "low-low" clusters (such as several adjacent counties dominated by heavy industry) can be specifically identified on the map.

[0042] S3. Using the ecological resilience index as the target variable and socio-economic environmental factors as feature variables, construct a machine learning regression model, and use an interpretable artificial intelligence framework to analyze the nonlinear effects and interactions of each feature variable on the ecological resilience index, and identify key driving factors and thresholds; based on the analysis results, formulate regional ecological resilience optimization and regulation strategies.

[0043] Specifically, the ecological resilience index of each evaluation unit calculated in the previous step is used as the target variable for model prediction. Selected socio-economic and environmental factors such as economic development level, industrial structure, and environmental investment are used as feature variables to construct a machine learning regression model. This model aims to learn the complex mapping relationship between the ecological resilience index and numerous driving factors. Its advantage lies in its ability to automatically capture and fit non-linear relationships between variables; for example, ecological resilience may first increase and then decrease with the growth of a certain factor. Subsequently, an interpretable artificial intelligence framework is used to analyze this "black box" machine learning model. This framework, based on the contribution allocation idea in cooperative game theory, can quantitatively decompose the specific contribution of each feature variable (such as the proportion of industrial structure) to the final predicted value (ecological resilience index) for each model prediction. This contribution value can be positive or negative. By analyzing all samples, the average contribution of each feature can be obtained to identify key driving factors. More importantly, this framework can generate feature dependency graphs to visualize the relationship between changes in individual feature values ​​and their contribution, thereby revealing their nonlinear impact patterns and accurately identifying the thresholds for feature values ​​where the contribution changes directionally or significantly (for example, when the proportion of industrial added value exceeds 45%, its contribution to ecological resilience turns from positive to negative). Simultaneously, the framework can analyze the interaction effects between any two feature variables, determining whether they act independently on ecological resilience or have synergistic or antagonistic effects. Based on the key negative driving factors obtained from the above analysis, their impact thresholds, and the interaction relationships between factors, combined with spatial pattern information such as high and low value clusters of ecological resilience identified in the previous spatial analysis, differentiated control strategies for different regions can be formulated. For example, for identified "low-low" ecological resilience clusters, if the analysis finds that "industrial structure" is a key negative factor and its threshold has been exceeded, a strict "green transformation of industrial structure" strategy will be formulated for that region; if a strong synergistic effect is found between "environmental investment" and "technological innovation," a corresponding strategy of "increasing incentives for green technology research and development" will be formulated, thus achieving a closed loop from mechanism diagnosis to precise policy implementation.

[0044] For example, taking the coastal zone of a certain province as an example, the ecological resilience index calculated for each county / district is used as the target variable, and eight socio-economic and environmental indicators collected from each county / district, including "industrial structure (proportion of industrial added value)," "proportion of environmental investment," and "GDP per capita," are used as feature variables. These are input into an XGBoost model for training. This model can learn that ecological resilience does not simply increase linearly with "GDP per capita," but rather follows a complex pattern where the growth rate slows down after reaching a certain level. After training, the SHAP framework is used for analysis. The analysis reveals that among all features, "industrial structure" and "environmental investment" have the highest average absolute SHAP values ​​and are identified as key driving factors. Further plotting the SHAP dependency graph for "industrial structure" shows that when the proportion of industrial added value in a county / district is below 40%, its SHAP value is mostly positive or close to zero, indicating that it has a slight positive or neutral impact on ecological resilience; however, when this proportion exceeds the 40% threshold, its SHAP value sharply turns into a significant negative value, indicating that an excessively high proportion of industry becomes a key factor inhibiting ecological resilience. Interactive analysis also revealed that the SHAP values ​​of the two characteristics, "environmental investment" and "technological innovation," were simultaneously positive or simultaneously negative in most samples, indicating a synergistic promoting or inhibiting effect on ecological resilience. Based on these findings, and combined with the "low-low" agglomeration areas identified in the spatial analysis, optimization strategies were formulated for these specific zones: first, an "industrial structure adjustment strategy," setting a 40% industrial share as a red line, mandating industrial upgrading and green transformation for counties exceeding this threshold; second, a "synergistic enhancement strategy," increasing environmental protection fiscal expenditure in the region and establishing a green technology innovation fund to maximize resilience enhancement through the synergistic effect of both. For the identified "high-high" agglomeration areas, a strategy focused on "maintaining advantages and preventing risks" was formulated, such as strictly controlling the industrial share threshold and maintaining and moderately increasing environmental investment.

[0045] According to one embodiment of this application, in step S1, the indicators of the ecological resilience dimension include: population density, built-up area ratio, per capita water resource consumption, carbon emissions per unit of GDP, industrial wastewater discharge per unit of GDP, and carbon dioxide emissions per unit of GDP. The indicators for the ecological resilience dimension include: comprehensive utilization rate of water resources, land resource optimization index, comprehensive utilization rate of solid waste, harmless treatment rate of garbage, centralized sewage treatment rate, and air quality excellence rate. The indicators for the ecological stability dimension include: green coverage rate of built-up areas, per capita green space area, biodiversity index, environmental investment ratio, vegetation coverage, and the proportion of water conservancy environment and public facilities management.

[0046] Specifically, the indicators of the ecological resilience dimension are used to measure the strength of coastal ecosystems in withstanding external pressures or disturbances. Among them: Population density and the proportion of built-up areas reflect the degree of physical encroachment and development intensity of human activities on natural space.

[0047] Per capita water consumption reflects the pressure on freshwater resources needed to maintain the operation of the socio-economic system.

[0048] Carbon emissions per unit of GDP, industrial wastewater emissions per unit of GDP, and carbon dioxide emissions per unit of GDP measure the immediate pollution load and impact of economic activities on the atmosphere and water bodies from three levels: gaseous pollutants, liquid pollutants, and greenhouse gas emission intensity, respectively.

[0049] The indicators of ecological resilience are used to measure the system's ability to achieve functional restoration and resource regeneration through self-regulation and human intervention after damage. Among them: The comprehensive utilization rate of water resources and the optimization index of land resources reflect the inherent ability to achieve intensive, efficient and circular utilization of resources under the constraints of water and soil resources.

[0050] The comprehensive utilization rate of solid waste, the harmless treatment rate of garbage, and the centralized sewage treatment rate together reflect the governance efficiency of artificially assisted remediation through the interception, treatment, and transformation of pollutants by end-of-pipe treatment facilities.

[0051] The rate of good air quality directly reflects the self-cleaning and recovery status of the atmospheric environment after being disturbed and the overall environmental quality level.

[0052] Indicators in the ecological stability dimension are used to measure the ability of an ecosystem to maintain structural integrity, functional continuity, and long-term balance. Among them: The green coverage rate, per capita green space area, and vegetation coverage of built-up areas together constitute the green space foundation and ecological barrier for mitigating natural disasters and regulating microclimate.

[0053] The biodiversity index reflects the species diversity and the complexity and structural integrity of the food chain within a region.

[0054] The proportion of environmental investment reflects the long-term financial support provided by the social system for ecological protection and pollution control.

[0055] The proportion of water conservancy, environment and public facilities management reflects the proportion of public expenditure used to support the operation and maintenance of ecological and environmental infrastructure and the degree of policy emphasis.

[0056] According to one embodiment of this application, in step S2, the comprehensive weighting and approximation of the ideal solution ranking method specifically involves: determining the objective weights of each indicator using the entropy weight method, constructing a weighted standardized decision matrix; calculating the Euclidean distance from each evaluation unit to the positive and negative ideal solutions, and then applying the formula:

[0057] Calculate the ecological resilience index for each evaluation unit. ,in To determine the distance to the ideal solution, This is the distance to the negative ideal solution.

[0058] Specifically, the entropy weight method is used to determine the objective weights of each evaluation indicator. The principle of the entropy weight method is to measure the information content of each indicator based on the dispersion of its data across all evaluation units. The greater the data difference for a particular indicator, the smaller its information entropy, and the greater the amount of information it contains; therefore, the higher the objective weight assigned to that indicator in the comprehensive evaluation. Conversely, the smaller the data difference, the greater the information entropy, the smaller the amount of information, and the lower its weight. This method allows for the objective determination of the weights of each indicator based on the distribution of the data itself, reducing the influence of subjective judgment.

[0059] After obtaining the weights of each indicator, the standardized data matrix from the previous steps is multiplied by the corresponding indicator weights to construct a weighted standardized decision matrix. This matrix contains both information on the relative importance of the indicators and comparable data after removing dimensions.

[0060] Subsequently, the calculation of the approximation ideal solution ranking method is performed. This stage requires defining two virtual reference points: the positive ideal solution and the negative ideal solution. The positive ideal solution consists of the optimal values ​​of all evaluation indicators in the weighted standardized matrix (maximum values ​​for positive indicators and minimum values ​​for negative indicators); the negative ideal solution consists of the worst values ​​of all indicators (minimum values ​​for positive indicators and maximum values ​​for negative indicators). Next, the Euclidean distances from the indicator vector of each actual evaluation unit to the positive and negative ideal solution vectors are calculated, denoted as […]. and .

[0061] Finally, according to the formula Calculate the ecological resilience index for each evaluation unit. The geometric meaning of this formula is: the degree of relative distance between the evaluation unit and the worst state (negative ideal solution). The value ranges from 0 to 1. The closer the value is to 1, the further the overall state of the unit is from the worst state and closer to the best state, that is, the higher the level of ecological resilience. Conversely, the closer the value is to 0, the closer its state is to the worst state and the lower the level of ecological resilience. Through this process, multi-dimensional indicator information is integrated into a single, comparable quantitative index.

[0062] According to an embodiment of this application, step S2, analyzing the spatial distribution pattern of ecological resilience, includes: using kernel density analysis to characterize the spatial clustering density of ecological resilience, and using global and local spatial autocorrelation analysis to identify the spatial correlation pattern of ecological resilience and high-value and low-value clustering areas.

[0063] Specifically, the first intelligent assessment and zoning optimization method for coastal ecological resilience is kernel density analysis. This method is used to characterize the spatial clustering density distribution of ecological resilience indices. Its basic principle is to treat the geographic center of each assessment unit as a spatial point and assign a corresponding ecological resilience index value to each point. Using a set search radius (bandwidth) and a mathematical function (kernel function), the ecological resilience density value at any location within the entire study area is systematically calculated. This process is similar to generating a continuous, smooth surface on a spatial map; the darker the color or the higher the value on the surface, the more concentrated and dense the assessment units with high ecological resilience are in space; conversely, a darker color indicates lower ecological resilience or sparse distribution. The results can intuitively reveal the spatial clustering hotspots and cold spots of ecological resilience at the macroscopic level.

[0064] The second method is global and local spatial autocorrelation analysis. Global spatial autocorrelation analysis aims to determine whether there are significant clustering or dispersion patterns (i.e., spatial correlation) in the overall spatial distribution of the ecological resilience index of all evaluation units within the entire study area. This is achieved by calculating statistics such as the global Moran index: if the index is significantly positive, it indicates that units with high ecological resilience tend to be adjacent to other high-value units, and units with low ecological resilience also tend to be adjacent to other low-value units, showing a positive spatial correlation (clustering); if it is significantly negative, it shows a negative spatial correlation (dispersion); if it is not significant, it indicates that the spatial distribution is random. Local spatial autocorrelation analysis, based on the global analysis, further identifies spatial correlation patterns in local areas. It calculates the local spatial correlation index between each evaluation unit and its neighboring units, specifically identifying four types of local spatial relationships: "high-high" clustering areas (both the unit itself and its neighboring units have high ecological resilience indices), "low-low" clustering areas (both the unit itself and its neighboring units have low indices), "high-low" anomaly areas (the unit itself has high indices but its neighboring units have low indices), and "low-high" anomaly areas (the unit itself has low indices but its neighboring units have high indices). This method can accurately locate hotspots (high-value clusters) and cold spots (low-value clusters) of ecological resilience on spatial maps, providing clear spatial targets for subsequent zoning optimization.

[0065] According to one embodiment of this application, in step S3, the socio-economic environmental factors include at least one of the following: economic development level, financial investment intensity, industrial structure, education level, urbanization level, technological innovation capability, social consumption level, and infrastructure level.

[0066] Specifically, in step S3, the independent variables used to construct the machine learning regression model, namely socioeconomic environmental factors, are selected from a pre-defined set of multi-dimensional indicators closely related to regional development and management. These factors aim to reflect the potential impact mechanisms of human socioeconomic activities on coastal ecosystems from different perspectives.

[0067] This set mainly includes the following eight categories of factors: Economic development level: usually characterized by indicators such as per capita GDP, reflecting the overall economic scale and wealth creation capacity of a region.

[0068] Financial investment intensity: This is usually represented by indicators such as the proportion of local government expenditure to regional GDP, reflecting the level of regulation and investment in the operation of the social economy.

[0069] Industrial structure: It is usually characterized by indicators such as the proportion of industrial added value to regional GDP, reflecting the share of the secondary industry in the economic composition, especially the industrial sector that may bring higher environmental burden.

[0070] Education level: Usually represented by indicators such as the proportion of education expenditure to regional GDP, reflecting society's investment in human capital development and the overall quality of the population.

[0071] Urbanization level: usually characterized by indicators such as the proportion of urban population to total population, reflecting the degree of spatial concentration of population and social structure.

[0072] Technological innovation capability: This is usually represented by indicators such as the number of green patents granted per 10,000 people, reflecting the region's activity in research and development and transformation of environmentally friendly technologies.

[0073] Social consumption level: usually represented by indicators such as per capita retail sales of consumer goods, reflecting the activity level of final demand in the regional market and residents' consumption capacity.

[0074] Infrastructure level: usually characterized by indicators such as per capita road area, reflecting the degree of perfection of the physical carriers supporting socio-economic activities.

[0075] When constructing the driving model, at least one of the above eight categories of factors can be selected as a feature variable based on the research objectives and data availability.

[0076] According to one embodiment of this application, the machine learning regression model is an XGBoost model based on gradient boosting decision trees; and the interpretable artificial intelligence framework is a SHAP framework based on game theory.

[0077] Specifically, the machine learning regression model selected in step S3 is the XGBoost model. This model is an efficient implementation of the gradient boosting decision tree algorithm. Its working principle is to sequentially construct a series of decision trees to continuously correct the prediction errors of previous models. The learning objective of each new tree is to fit the residual gradient between the current prediction result and the true value. This iterative ensemble strategy gives the XGBoost model powerful nonlinear fitting capabilities, effectively capturing the complex, nonlinear mapping relationship between the ecological resilience index and various socio-economic environmental driving factors, and typically exhibits high prediction accuracy and generalization performance.

[0078] In step S3, the interpretability AI framework used to analyze the model is the SHAP framework. Based on the Shapley value concept in game theory, this framework provides a unified method for interpreting complex machine learning models (such as XGBoost). Its core principle is to treat each prediction by the model as the result of collaboration among all input features (i.e., driving factors) through the model as a "coalition." The SHAP framework fairly quantifies the marginal contribution of each feature to a specific prediction relative to the baseline prediction by calculating the Shapley value of each feature. By analyzing the SHAP values ​​of each feature across all samples, several interpretation goals can be achieved: first, assessing the overall importance of features (through the average absolute SHAP value); second, visualizing the non-linear dependency between individual feature values ​​and their SHAP values, thereby identifying key inflection points or thresholds; and third, revealing the interaction effects of any two features in influencing prediction, determining whether their effects are independent, synergistic, or offsetting. This transforms the model's "black box" predictions into understandable and traceable attribution analysis.

[0079] According to one embodiment of this application, the analysis using the SHAP framework includes: calculating the SHAP value of each feature variable to quantify its average marginal contribution, and drawing a SHAP dependency graph to visualize the nonlinear impact and threshold effect of a single feature variable on the ecological resilience index.

[0080] Specifically, the process of parsing using the SHAP framework includes two main parts: The first part involves calculating the SHAP value of each feature variable to quantify its average marginal contribution. For a trained machine learning model, the SHAP framework interprets the prediction results for each sample (i.e., each evaluation unit). It uses a cooperative game theory-based algorithm to calculate the contribution of each individual feature variable (such as "industrial structure proportion") to the model's final predicted ecological resilience index value, relative to a baseline prediction (usually the average of all sample predictions), given all feature values ​​for that sample. This contribution is the SHAP value of that feature for that sample. SHAP values ​​can be positive or negative; a positive value indicates that the feature drives the predicted value towards higher ecological resilience in that sample, while a negative value indicates that it drives the predicted value towards lower ecological resilience. By summing the SHAP values ​​of the same feature variable across all samples and calculating the average of their absolute values, the average marginal contribution of that feature variable can be obtained, thus objectively and quantitatively assessing which socioeconomic and environmental factors are the key drivers influencing regional ecological resilience differentiation.

[0081] The second part involves plotting a SHAP dependency graph to visualize the impact patterns. After calculating the SHAP values ​​for each feature of all samples, a scatter plot (SHAP dependency graph) can be created for each feature variable, with the actual value of the feature on the x-axis and its corresponding SHAP value on the y-axis. This graph visually shows how the contribution (SHAP value) of the feature value to the predicted ecological resilience changes as the feature value changes. By observing the trend of the scatter plots, it can be determined whether there is a linear or complex nonlinear relationship between the feature and ecological resilience (such as an inverted "U"-shaped relationship that first promotes and then inhibits). The locations of feature values ​​where the scatter plot distribution trend shows a significant turning point or where the sign of the SHAP value changes systematically mark the key thresholds at which the driving factor affects ecological resilience. For example, when the feature value is below a certain critical point, its SHAP value is generally positive or close to zero; after exceeding this critical point, the SHAP value generally turns negative, and this critical point is an inhibitory threshold.

[0082] According to one embodiment of this application, the analysis using the SHAP framework further includes: performing interaction analysis between feature variables to reveal the synergistic or antagonistic effects of different socioeconomic environmental factors on the ecological resilience index.

[0083] Specifically, the interaction analysis between characteristic variables using the SHAP framework refers to quantifying and revealing the joint effect between any two different socioeconomic environmental factors when they jointly affect the ecological resilience index. This joint effect is not a simple addition of the independent contributions of the two factors, but rather means that when the two factors simultaneously have specific values, their combined impact on the predicted ecological resilience value will be greater than (synergistic effect) or less than (antagonistic effect) the sum of their individual independent impacts.

[0084] In addition to calculating an independent SHAP value for each feature, the SHAP framework also calculates an interaction SHAP value for each pair of features. This value quantifies the additional joint contribution of the two features together to the prediction result in a specific sample. By analyzing the distribution pattern of the interaction values ​​of a pair of features across all samples, the relationship between them can be determined. Synergistic effect: When the interaction value of two features (such as "environmental investment ratio" and "green technology innovation capability") is positive in most samples, it indicates that in the model prediction, these two factors tend to jointly enhance the positive driving effect on ecological resilience or jointly reduce its negative impact. That is, when both exist at the same time or are at a high level, they can produce a "one plus one is greater than two" promoting effect.

[0085] Antagonistic effect: When the interaction value of two features (such as "industrial structure proportion" and "urbanization level") is negative in most samples, it indicates that the two factors tend to weaken each other's impact on ecological resilience in the model prediction. That is, one factor may inhibit or offset the effect of the other factor, resulting in a mutually restrictive effect.

[0086] No significant interaction: If the interaction value is randomly distributed around zero and has no systematic direction, it indicates that the two factors have statistically independent effects on ecological resilience.

[0087] By systematically analyzing the interaction SHAP values ​​between all important feature pairs, we can identify which combinations of factors maximize ecological resilience (key synergistic combinations) and which combinations may introduce additional risks or offset governance effects (key antagonistic combinations). This analysis provides a direct mechanistic basis for developing comprehensive, multi-policy synergistic optimization strategies.

[0088] According to one embodiment of this application, step S3, formulating a zoned ecological resilience optimization and regulation strategy specifically includes: for the low-value clustering areas identified by the spatial analysis, and in combination with the negative impact threshold of the key driving factors, formulating a differentiated regulation strategy with at least one of the following as the core: green transformation of industrial structure, increased environmental investment, improved resource utilization efficiency, and strengthened ecological space protection.

[0089] Specifically, the step S3 involves formulating a regional ecological resilience optimization and regulation strategy, which is based on the integration and decision-making of the analysis results from the preceding steps.

[0090] First, the spatial targeting of the strategy is based on the "low-value clusters" identified in the spatial pattern analysis in step S2. These areas refer to the "low-low" clusters determined through local spatial autocorrelation analysis, that is, spatially continuous areas with low ecological resilience indices and low indices in their surrounding neighboring units. These are key areas that require priority intervention.

[0091] Secondly, the core basis for strategy formulation is the "negative impact threshold of key driving factors" identified in step S3 of the SHAP framework analysis. Specifically, this refers to those characteristic variables (such as "industrial structure") that are determined to have a significant negative impact on ecological resilience, as well as the critical values ​​determined by the SHAP dependency graph where the nature or intensity of their impact changes significantly (e.g., when the industrial added value accounts for 40%, the contribution turns from positive to negative).

[0092] Developing differentiated regulation strategies involves combining the aforementioned spatial targeting with mechanistic basis to design and implement targeted intervention measures addressing the core shortcomings exposed in specific "low-value clusters." The core directions of these measures include, but are not limited to: Green transformation of industrial structure: If the key negative driving factor in the region is "industrial structure" and the current value has exceeded its negative threshold, then strategies should be formulated to reduce the proportion of industries with high environmental impact, such as setting red lines for industrial access, promoting the clean transformation of traditional industries, and cultivating green emerging industries.

[0093] Increase environmental investment: If the region’s “environmental investment ratio” is identified as a key positive factor and the current level is insufficient, or if investment is needed to offset the impact of other negative factors, then strategies should be developed to increase the intensity of fiscal investment in ecological and environmental protection and governance.

[0094] Improve resource utilization efficiency: If resilience indicators such as "comprehensive utilization rate of water resources" or "comprehensive utilization rate of solid waste" in the region are identified as key constraints, strategies should be developed to promote recycling technologies, improve resource recycling systems, and implement stricter process management.

[0095] Strengthen ecological space protection: If stability indicators such as "green coverage rate" or "per capita green area" of the region are identified as key weaknesses, strategies should be developed to delineate and strictly adhere to ecological protection red lines, implement ecological restoration projects, and increase blue and green spaces.

[0096] Ultimately, for a specific "low-value cluster area", based on the diagnostic results of its driving mechanism, the most suitable and urgent intervention measures will be selected from one or more of the above core directions to form a customized optimized control strategy.

[0097] A second aspect of this application provides a system comprising: The indicator construction and preprocessing module is used to construct an evaluation indicator system for the ecological resilience of coastal areas from three dimensions: ecological resistance, ecological resilience, and ecological stability, based on the socio-ecological system theory, and to preprocess and standardize the acquired multi-source data. The assessment and spatial analysis module is used to calculate the ecological resilience index of each evaluation unit in the study area based on the aforementioned indicator system and standardized data, using a comprehensive weighting and approximation of ideal solution ranking method, and to analyze the spatial distribution pattern of ecological resilience. The analysis and optimization strategy generation module is used to construct a machine learning regression model with the ecological resilience index as the target variable and socio-economic environmental factors as feature variables, and to analyze the nonlinear effects and interactions of each feature variable on the ecological resilience index using an interpretable artificial intelligence framework, identify key driving factors and thresholds; and to formulate regional ecological resilience optimization and regulation strategies based on the analysis results.

[0098] A second aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in any of the embodiments of the first aspect above.

[0099] Figure 2 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 2 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 may call logical instructions in the memory 830 to execute the method in any of the embodiments of the first aspect described above, the method including: S1. Based on the socio-ecological system theory, an evaluation index system for the ecological resilience of coastal areas is constructed from three dimensions: ecological resistance, ecological resilience, and ecological stability. The acquired multi-source data are preprocessed and standardized. S2. Based on the aforementioned indicator system and standardized data, the ecological resilience index of each evaluation unit within the study area is calculated using a comprehensive weighting and approximation of ideal solution ranking method, and the spatial distribution pattern of ecological resilience is analyzed. S3. Using the ecological resilience index as the target variable and socio-economic environmental factors as feature variables, construct a machine learning regression model, and use an interpretable artificial intelligence framework to analyze the nonlinear effects and interactions of each feature variable on the ecological resilience index, and identify key driving factors and thresholds; based on the analysis results, formulate regional ecological resilience optimization and regulation strategies.

[0100] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0101] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program, the computer program being able to be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer being able to perform the methods provided by the above methods, the method comprising: S1. Based on the socio-ecological system theory, an evaluation index system for the ecological resilience of coastal areas is constructed from three dimensions: ecological resistance, ecological resilience, and ecological stability. The acquired multi-source data are preprocessed and standardized. S2. Based on the aforementioned indicator system and standardized data, the ecological resilience index of each evaluation unit within the study area is calculated using a comprehensive weighting and approximation of ideal solution ranking method, and the spatial distribution pattern of ecological resilience is analyzed. S3. Using the ecological resilience index as the target variable and socio-economic environmental factors as feature variables, construct a machine learning regression model, and use an interpretable artificial intelligence framework to analyze the nonlinear effects and interactions of each feature variable on the ecological resilience index, and identify key driving factors and thresholds; based on the analysis results, formulate regional ecological resilience optimization and regulation strategies.

[0102] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the methods provided by the above methods, the method comprising: S1. Based on the socio-ecological system theory, an evaluation index system for the ecological resilience of coastal areas is constructed from three dimensions: ecological resistance, ecological resilience, and ecological stability. The acquired multi-source data are preprocessed and standardized. S2. Based on the aforementioned indicator system and standardized data, the ecological resilience index of each evaluation unit within the study area is calculated using a comprehensive weighting and approximation of ideal solution ranking method, and the spatial distribution pattern of ecological resilience is analyzed. S3. Using the ecological resilience index as the target variable and socio-economic environmental factors as feature variables, construct a machine learning regression model, and use an interpretable artificial intelligence framework to analyze the nonlinear effects and interactions of each feature variable on the ecological resilience index, and identify key driving factors and thresholds; based on the analysis results, formulate regional ecological resilience optimization and regulation strategies.

[0103] Example 2 As a complex zone with the most frequent interactions between land and sea and the most vulnerable ecosystems, the coastal zone plays an irreplaceable role in mitigating natural disasters, regulating climate change, and maintaining biodiversity. Scientifically constructing an evaluation index system that aligns with the "resistance-resilience-stability" characteristics of the coastal zone, developing an ecological resilience assessment model applicable to complex land-sea environments, and dynamically evaluating the evolution of coastal ecological resilience and its nonlinear driving mechanisms under the background of high-intensity human activities will help promote differentiated regional land space management, explore new pathways for constructing coastal ecological security barriers, and provide scientific decision-making support for high-quality development in coastal areas.

[0104] By comprehensively employing the entropy weight method and the TOPSIS model to construct an ecological resilience measurement framework, this approach effectively overcomes the subjectivity and information gaps inherent in single-weighting methods, significantly improving the scientific rigor and accuracy of the assessment results. Simultaneously, spatial autocorrelation methods are used to analyze the spatial clustering characteristics and hot / cold areas of ecological resilience distribution patterns, quantifying the degree of spatial correlation and spillover effects between regions. Furthermore, machine learning algorithms (XGBoost) and the game theory explanation framework (SHAP) are introduced to deeply reveal the nonlinear driving mechanism and threshold effect of key factors on ecological resilience. Based on attribution analysis results, zoning optimization strategies are formulated, achieving a leap from "status assessment" to "precise regulation," providing strong decision-making support for regional sustainable development.

[0105] This embodiment selects the Shandong Peninsula coastal zone as the study area. Located at the confluence of two seas, the Shandong Peninsula coastal zone boasts a long coastline encompassing key coastal cities such as Qingdao, Yantai, Weihai, Rizhao, Dongying, Binzhou, and Weifang. This region is not only the blue engine of Shandong Province's economic development but also a typical marine-terrestrial ecological transition zone. Unlike inland areas, the Shandong Peninsula coastal zone faces unique dual pressures: on the one hand, human interference from intensive port construction, land reclamation, and port-related industrial development; on the other hand, natural threats from marine disasters such as seawater intrusion, coastal erosion, and typhoon storm surges. Against this backdrop, scientifically constructing an ecological resilience evaluation system that conforms to the "marine-terrestrial complex" characteristics of the coastal zone and developing an assessment model suitable for highly dynamic coastal areas are particularly important. For example... Figure 3 As shown, this solution includes the following specific methods and steps: (I) Construction of an evaluation index system for the ecological resilience of coastal areas Sub-step 1-1: Construction of the indicator system: Based on the characteristics of the coastal zone's "nature-society" complex system, we systematically analyze the interaction between high-intensity human activities and the fragile marine and terrestrial environment. Around the three dimensions of ecosystem resistance, ecosystem resilience, and ecosystem stability, we construct a comprehensive indicator system for the ecological resilience of the coastal zone.

[0106] To assess the fundamental characteristics of coastal ecosystem resilience, six secondary indicators were selected for monitoring and analysis: population density, proportion of built-up areas, per capita water resource consumption, carbon emissions per unit of GDP, industrial wastewater discharge per unit of GDP, and sulfur dioxide emissions per unit of GDP. Population density and the proportion of built-up areas reflect the physical encroachment and spatial stress on coastal natural habitats caused by rapid urbanization; per capita water resource consumption reflects the resource consumption pressure required to maintain system operation; and carbon emissions per unit of GDP, industrial wastewater discharge, and sulfur dioxide emissions reflect the intensity and impact of environmental pollution emissions from economic production activities across multiple media (gas, liquid, and solid). To assess the fundamental characteristics of coastal ecosystem resilience, six secondary indicators were selected for monitoring and analysis: comprehensive water resource utilization rate, land resource optimization index, comprehensive solid waste utilization rate, harmless waste treatment rate, centralized sewage treatment rate, and air quality excellence rate. Among them, the comprehensive utilization rate of water resources and the land resource optimization index reflect the endogenous driving force for the coastal zone to achieve system function regeneration through intensive utilization under the background of resource scarcity. The comprehensive utilization rate of solid waste, the harmless treatment rate of garbage, and the centralized sewage treatment rate reflect the governance capacity to intercept pollutants entering the sea and achieve artificial assisted restoration through the construction of environmental infrastructure. The air quality excellence rate directly reflects the self-purification and recovery status of the environmental baseline after disturbance. Based on the basic characteristics of the stability of the coastal ecosystem, six secondary indicators were selected for monitoring and analysis: green coverage rate of built-up areas, per capita green space area, vegetation coverage, biodiversity index, environmental investment ratio, and the proportion of water conservancy, environmental, and public facilities management industries. Among these, the green coverage rate of built-up areas, per capita green space area, and vegetation coverage together constitute a green ecological barrier against natural disasters such as storm surges. The biodiversity index reflects the biodiversity of the coastal zone and the structural integrity of the ecological chain. The environmental investment ratio and the proportion of water conservancy, environmental, and public facilities management industries reflect the level of social capital investment and policy support for maintaining the long-term stable operation of the system.

[0107] Table 1. Coastal Zone Ecological Resilience Index System

[0108] (II) Construction of an assessment model for the ecological resilience of coastal zones Sub-steps 1-2: Based on the established indicator system, a coastal zone ecological resilience assessment model is established by comprehensively considering the internal dispersion of indicators and the correlation between indicators.

[0109] The amount of information contained in each evaluation indicator is measured by analyzing the dispersion of each indicator in the sample. The calculation process includes data standardization, calculating the weight distribution, information entropy, and difference coefficient of each indicator, and finally obtaining the weight of each indicator.

[0110] The specific method for constructing the model is as follows: Establish the original data matrix. , where m is the number of samples (year × number of cities). This refers to the number of indicators.

[0111] Process the raw data; positive indicators:

[0112] For negative indicators:

[0113] For standardized data, the range of values ​​is... .

[0114] Sub-step 2-2: Minimum Information Entropy Evaluation Model The Analytic Hierarchy Process (AHP) is a structured, systematic, and effective method for determining the relative importance of body weight. It allows for quantitative and qualitative analysis of multiple indicators. The AHP process can be represented by the following steps: Establish a standardized judgment matrix:

[0115] in , , , Standardized judgment matrix:

[0116] Calculation of eigenvalues ​​and eigenvectors:

[0117] in It is an eigenvector. These are the eigenvalues ​​of a given matrix.

[0118] Check consistency indicators ( ):

[0119] in It determines the largest eigenvalue of the matrix; Modified Entropy Method: The entropy method refers to a method for judging the degree of dispersion of an indicator, and is generally used to calculate objective... Weights. However, the previously standardized metric value would show as 0, to avoid... In this situation, The standardized indicators were transformed. Calculation ratio:

[0120] Calculate the entropy value:

[0121] Calculate the entropy weighting coefficient:

[0122] The minimum information entropy method is used, combined with the subjective and objective weights of each evaluation, to obtain the comprehensive weight. The calculation formula is as follows:

[0123] in Represents the overall weight. Refers to the year. The weights are obtained from the analytic hierarchy process. The weights are obtained using the entropy method.

[0124] TOPSIS Analysis Construct the weighted matrix:

[0125] Determine the ideal solution:

[0126] Determine the negative ideal solution:

[0127] Calculate distance:

[0128] Calculate the Ecological Resilience Index (ERI):

[0129] The value ranges from 0 to 1, with a higher value indicating better ecological resilience in the area.

[0130] Sub-steps 2-3: Spatial evolution trends of coastal ecological resilience Nuclear density analysis: set up For independent and identically distributed sample points, the kernel density estimator for:

[0131] in, This is the kernel function, typically using a Gaussian kernel; The bandwidth determines the smoothness. The distance from the valuation point to the event point (city center). Sub-steps 2-4: Spatial Relationship Pattern Analysis Using GeoDa software, calculate the Global Moran's I index:

[0132] like Furthermore, this significantly indicates spatial clustering of resilience in the Shandong Peninsula's ecological coastal zone. Further... LISA image recognition: HH Zone (High-High Agglomeration): Usually located in coastal cities with relatively good ecological foundations, such as Weihai and Yantai.

[0133] LL zone (low-low agglomeration): may be located in areas with high concentration of heavy industry.

[0134] Step (III) Analysis of the Driving Mechanisms and Optimization of Ecological Resilience in Coastal Zones Sub-step 3-1: Constructing the set of influencing factor indicators and regression model Eight driving factors were selected (Table 2): Table 2. Factors affecting ecological resilience

[0135] Construct a dataset with ERI values ​​as labels and the aforementioned eight factors as features. Train the dataset using the XGBoost algorithm. Compared to traditional linear regression, XGBoost effectively captures the complex nonlinear relationship between different socioeconomic factors and ecological resilience by boosting the decision tree through gradients. The specific steps are as follows: Samples from 34 counties and districts along the Shandong Peninsula coastline Let the first The true ecological resilience index of each sample is , No. The predicted value of the model after rounds of iterations is ,but The objective function of the round iteration is:

[0136] In the formula, is the loss function used to measure the deviation between the model's predicted values ​​and the actual values. In this study, because the ecological resilience index is a continuous variable, the mean squared error loss (MSE) is chosen. For classification problems, the logistic loss function can be used. : No. k The regularization term of the decision tree is used to control the complexity of the tree, and the formula is:

[0137] Where T is the number of leaf nodes in the tree. For the first The output weights of the leaf nodes (Leaf node penalty coefficient) and (Weight L2 regularization coefficient) is a hyperparameter that needs to be optimized through cross-validation. : Constant term, does not affect the optimization process, and can be ignored.

[0138] In this study: ① The target variable was set. y Ecological resilience of counties and districts along the Shandong Peninsula coastline; characteristic variables x This includes: core variables + control variables (9 influencing factors such as environmental regulations, economic development, and technological innovation). ② The sample is randomly divided into a training set (for model training) and a test set (for model evaluation) at an 8:2 ratio, using stratified sampling to ensure consistent ecological resilience distribution between the training and test sets. ③ Core hyperparameters are optimized using a grid search + 10-fold cross-validation approach to balance model accuracy and generalization ability. Sub-step 3-2: SHAP-based metric contribution and interaction analysis Interpreting the trained XGBoost model using the SHAP library: The SHAP additive explanatory model is a method for interpreting predictions from machine learning models based on Shapley values ​​in game theory. It quantifies the contribution of each feature to the model's prediction and reveals the relationship between features and predictions. An explanatory model for the influencing factors of coastal ecological resilience is constructed as follows:

[0139] In the formula, For the first The first sample One characteristic, , , The mean of all target variables for all samples. For the first The first sample The SHAP value of each feature.

[0140] Factor contribution analysis: Calculate the mean absolute SHAP value of each feature. In this embodiment, the analysis... The results showed that "economic development" and "industrial structure" had the highest SHAP values, indicating that these two factors were the most important for Shandong. Driving factors of ecological resilience differentiation in peninsula coastal zones (such as...) Figure 4 ). Nonlinear dependency analysis: Plot a scatter plot of the eigenvalues ​​of each influencing factor versus the SHAP value (e.g., Figure 5 ). Interaction analysis: Analyzing the interaction between environmental regulations and individual influencing factors (such as...) Figure 6 ). Sub-step 3-3: Simulation of wetland ecological environment quality optimization Based on the above quantitative analysis, a targeted optimization plan is proposed: Threshold control: For counties and districts where the proportion of industry exceeds the SHAP identification threshold, a green transformation of the industrial structure will be enforced and the approval of high-energy-consuming industries will be restricted. Addressing Weaknesses: Given the extremely high weight of water resource utilization rate in the indicator system, the entire province can promote reclaimed water recycling technology, making improving water resource utilization rate the primary project for enhancing regional resilience. Spatial Differentiation: For LL clusters identified by LISA, increase the proportion of environmental investment and per capita green space area, leveraging spatial spillover effects to drive resilience recovery in surrounding areas.

[0141] For any parts not mentioned in this application, existing technologies may be used or referenced.

[0142] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0143] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for intelligent evaluation and zoning optimization of coastal zone ecological resilience, characterized in that, Includes the following steps: S1. Based on the socio-ecological system theory, an evaluation index system for the ecological resilience of coastal areas is constructed from three dimensions: ecological resistance, ecological resilience, and ecological stability. The acquired multi-source data are preprocessed and standardized. S2. Based on the aforementioned indicator system and standardized data, the ecological resilience index of each evaluation unit within the study area is calculated using a comprehensive weighting and approximation of ideal solution ranking method, and the spatial distribution pattern of ecological resilience is analyzed. S3. Using the ecological resilience index as the target variable and socio-economic environmental factors as the feature variables, construct a machine learning regression model, and use an interpretable artificial intelligence framework to analyze the nonlinear effects and interactions of each feature variable on the ecological resilience index, and identify key driving factors and thresholds. Based on the analysis results, a regional ecological resilience optimization and regulation strategy was formulated.

2. The method according to claim 1, characterized in that, In step S1, the indicators of ecological resilience include: population density, proportion of built-up area, per capita water resource consumption, carbon emissions per unit of GDP, industrial wastewater discharge per unit of GDP, and carbon dioxide emissions per unit of GDP. The indicators for ecological resilience include: comprehensive utilization rate of water resources, land resource optimization index, comprehensive utilization rate of solid waste, harmless treatment rate of garbage, centralized sewage treatment rate, and air quality good rate. Indicators for the ecological stability dimension include: green coverage rate of built-up areas, per capita green space area, biodiversity index, proportion of environmental investment, vegetation coverage, and proportion of water conservancy environment and public facilities management.

3. The method according to claim 1 or 2, characterized in that, In step S2, the comprehensive weighting and approximation of the ideal solution ranking method specifically involves: determining the objective weights of each indicator using the entropy weight method, constructing a weighted standardized decision matrix; calculating the Euclidean distance from each evaluation unit to the positive and negative ideal solutions, and applying the formula: Calculate the ecological resilience index for each evaluation unit. ,in To determine the distance to the ideal solution, This is the distance to the negative ideal solution.

4. The method according to claim 3, characterized in that, In step S2, the analysis of the spatial distribution pattern of ecological resilience includes: using kernel density analysis to characterize the spatial clustering density of ecological resilience, and using global and local spatial autocorrelation analysis to identify the spatial correlation pattern of ecological resilience and high-value and low-value clustering areas.

5. The method according to claim 1, characterized in that, In step S3, the socio-economic environmental factors include at least one of the following: economic development level, financial investment intensity, industrial structure, education level, urbanization level, technological innovation capability, social consumption level, and infrastructure level.

6. The method according to claim 5, characterized in that, The machine learning regression model is the XGBoost model based on gradient boosting decision trees; the interpretable artificial intelligence framework is the SHAP framework based on game theory.

7. The method according to claim 6, characterized in that, The analysis using the SHAP framework includes: calculating the SHAP value of each feature variable to quantify its average marginal contribution, and drawing a SHAP dependency graph to visualize the nonlinear impact and threshold effect of a single feature variable on the ecological resilience index.

8. The method according to claim 7, characterized in that, The analysis using the SHAP framework also includes: conducting interaction analysis among feature variables to reveal the synergistic or antagonistic effects of different socioeconomic and environmental factors on the ecological resilience index.

9. The method according to claim 1, characterized in that, In step S3, the formulation of a zoned ecological resilience optimization and regulation strategy specifically includes: for low-value clustering areas identified by spatial analysis, and in combination with the negative impact threshold of the key driving factors, formulating a differentiated regulation strategy with at least one of the following as the core: green transformation of industrial structure, increase in environmental investment, improvement of resource utilization efficiency, and strengthening of ecological space protection.

10. A smart evaluation and zoning optimization system for coastal ecological resilience, characterized in that, include: The indicator construction and preprocessing module is used to construct an evaluation indicator system for the ecological resilience of coastal areas from three dimensions: ecological resistance, ecological resilience, and ecological stability, based on the socio-ecological system theory, and to preprocess and standardize the acquired multi-source data. The assessment and spatial analysis module is used to calculate the ecological resilience index of each evaluation unit in the study area based on the aforementioned indicator system and standardized data, using a comprehensive weighting and approximation of ideal solution ranking method, and to analyze the spatial distribution pattern of ecological resilience. The analysis and optimization strategy generation module is used to construct a machine learning regression model with the ecological resilience index as the target variable and socio-economic environmental factors as feature variables, and to use an interpretable artificial intelligence framework to analyze the nonlinear effects and interactions of each feature variable on the ecological resilience index, and to identify key driving factors and thresholds. And based on the analysis results, formulate regional ecological resilience optimization and regulation strategies.

Images