Selection method for constructing multi-element indexes in green watershed

Abstract

The invention provides a selection method for constructing multi-element indexes of a green watershed, which is used for systematically analyzing the multi-element indexes of water resources, water environment, water ecology and water economy of the green watershed based on a principal component analysis technology by fully utilizing big data analysis methods such as machine learning and the like, and belongs to the technical field of environment planning and management. The method comprises the steps of data collection and preprocessing; the principal component analysis process comprises the steps of constructing a data matrix, performing data standardization processing, calculating a covariance matrix, calculating a characteristic value and a characteristic vector, calculating an accumulated contribution rate and writing a principal component expression; screening indexes; and verifying and optimizing. According to the method, dimension reduction processing is carried out on the indexes through the principal component analysis method, the core indexes in the multi-element index system constructed for the green drainage basin are screened out, redundancy is effectively reduced, and the index screening precision and efficiency are improved. According to the method, the screening process of the green drainage basin indexes is optimized, and professional technical reference can be provided for environment planning and management.

Description

Technical Field

[0001] The invention provides a method for selecting multi-factor indicators for constructing a green watershed, and belongs to the technical field of environmental planning and management. Background Art

[0002] In the field of watershed management and environmental management, existing technologies mainly use traditional indicator selection methods such as expert judgment and analytic hierarchy process (AHP). These methods have been widely used in many fields, but big data statistical methods such as principal component analysis have not been specifically applied to the selection of indicators for green watersheds, and optimization and application for specific needs of green watersheds are also relatively lacking.

[0003] Green river basin construction refers to a river basin with sustainable development in river basin planning and environmental management, which mainly includes basic characteristics such as good water quality, rich biodiversity, healthy ecosystem, and green and low-carbon development.

[0004] Although there are some evaluation methods similar to those in the green watershed category, they still face challenges in practical application. For example, the index system and evaluation methods for constructing green watersheds are still in the preliminary research stage, and a lot of scientific research is needed to determine the relevant indicators. In addition, the existing index system research is mostly focused on a single water quality or ecological evaluation, lacking a multi-dimensional and comprehensive evaluation system, which limits the comprehensive application of these evaluation methods in green watersheds. The Yangtze River Scientific Research Institute has established an evaluation index system for public ecological rights and interests at the subjective level based on the core connotation of "Healthy Yangtze River" and following the four principles of scalability, scientific systematicity, policy orientation and comparability, reflecting the people's demands and satisfaction with ecological rights and interests in the Yangtze River Basin. At the objective level, an ecological environment system comprehensive status evaluation index system was constructed from the four dimensions of harmonious Yangtze River, clean Yangtze River, safe Yangtze River and beautiful Yangtze River. The system uses the traditional PSR (Pressure-State-Response) model to set specific indicators to ensure the representativeness, scientificity and operability of the indicators. First, the indicator status values ​​of different indicator elements are dimensionlessly normalized to ensure the comparability of the indicators. The hierarchical structure diagram of indicators was constructed, and the weight of indicators C1-C4 and factor layer B2 was allocated by the weighting method, each accounting for 1 / 5 of the weight. The indicator layer D was classified according to P, S, and R, and the weight was equalized from top to bottom. Quantitative indicators were referred to the existing national, local or industry standards or specifications as much as possible, and the relevant historical data, reference point conditions and related literature research results were also used to determine the values ​​through regional management goals. However, the indicator did not apply the method of selecting indicators to green watersheds. The evaluation system focuses more on traditional environmental quality and ecological health indicators, but pays less attention to indicators of green development, sustainability and ecological service value, and fails to fully cover indicators reflecting the development of green economy, such as the proportion of clean energy and green GDP. In addition, the existing methods also rely on expert judgment and have a certain degree of subjectivity, resulting in insufficient scientific basis and systematicness in the selection of indicators. The data related to green watersheds are of various types and highly correlated, making it difficult to extract key and effective indicators from them. Therefore, the construction of a multi-factor indicator system for green watersheds aims to overcome the defects of subjective indicator selection, high data complexity and insufficient evaluation tools in existing watershed management. Summary of the invention

[0005] The purpose of the present invention is to make full use of advanced big data analysis technologies such as machine learning, and propose a screening method for multi-factor indicators of green watershed construction based on principal component analysis (PCA, Principal Components Analysis). Through systematic data analysis and dimensionality reduction processing of multi-factor indicators such as environment, ecology, economy, and society in green watershed construction, the core indicators that are critical and representative for green watershed construction are screened and integrated. Through this method, the present invention aims to establish a foundation for establishing a comprehensive indicator system to achieve a systematic evaluation of green watershed construction. This indicator system will provide a scientific basis for accurately identifying the key factors affecting the construction of green watersheds, thereby providing strong support for sustainable management and decision-making of watersheds.

[0006] The technical solution of the present invention:

[0007] A method for selecting multi-factor indicators for green watershed construction includes the following steps:

[0008] S1. Data collection and acquisition;

[0009] Clarify the definition and scope of green river basin construction, and determine the environmental factors that need to be monitored and analyzed. Collect data on ecological, environmental, economic, and social indicators related to green river basin construction. On the basis of existing relevant indicators, summarize the principles and indicators that can be used as reference, and select multi-factor indicators covering water resources, water environment, water ecology, and water economy, while covering multiple dimensions such as green development, healthy water ecosystems, harmony between people and water, and regional coordination as much as possible.

[0010] Data can be collected by setting up monitoring points in the target watershed to collect samples and conduct on-site testing; using satellite or UAV remote sensing technology to obtain environmental data for large areas; using relevant data in existing databases and literature; and consulting experts to provide data.

[0011] S2, data preprocessing;

[0012] The collected raw data are preprocessed, including data cleaning and missing value processing, to improve the integrity and consistency of the data.

[0013] S3. Principal component analysis, including:

[0014] (1) Construct a data matrix;

[0015] For p indicators, there are n samples, and an n×p sample matrix can be constructed:

[0016]

[0017] (2) Data standardization;

[0018] Calculate the mean and standard deviation respectively:

[0019] Mean:

[0020] Standard Deviation: Normalize data within a matrix: The initial sample matrix is ​​normalized:

[0021] (3) Calculate the covariance matrix:

[0022]

[0023] in, So we can get:

[0024]

[0025] (4) Calculate eigenvalues ​​and eigenvectors:

[0026] Eigenvalue: λ1≥…≥λ p ≥0

[0027] Eigenvectors:

[0028] (5) Calculate the cumulative contribution rate:

[0029]

[0030] Where i=1,…,p.

[0031] (6) Write the principal component expression:

[0032] The i-th principal component: F i =a 1i X1+…+a pi X p

[0033] S4, indicator screening;

[0034] The evaluation of green watershed construction is a complex process with multiple dimensions and indicators, and the evaluation method involves a large number of indicators. According to the results of principal component analysis, the eigenvalues ​​are sorted and the core indicators that have significant contributions to green watershed construction are selected, requiring the cumulative variance contribution rate to reach 75% or 90%, which shows that these indicators play an important role in the evaluation of green watershed construction to reduce data redundancy and improve the accuracy of evaluation.

[0035] S5, verification and optimization;

[0036] Explain the selected key indicators and explain the specific role and significance of each indicator in the green watershed construction assessment. Combined with actual needs and application background, further comprehensive evaluation of the selected key indicators is carried out, considering the measurability, operability and practical application value of the indicators, verifying the rationality of the screening results, and ensuring that the selected indicators can effectively reflect the characteristics of green watershed construction.

[0037] Continuously optimize based on the new indicator system data, and regularly evaluate and adjust the screened indicators to maintain the accuracy and effectiveness of the green watershed construction assessment.

[0038] The key technology of the present invention is:

[0039] ①Use principal component analysis to select indicators;

[0040] Principal component analysis is an unsupervised machine learning method that uses the idea of ​​dimensionality reduction to transform the original variables with high correlation into a smaller number of linearly independent variables, namely principal components, through orthogonal transformation.

[0041] The number of principal components is less than the number of original variables, which can achieve data dimensionality reduction processing while minimizing information loss. By screening indicators through principal component analysis, the most representative principal components are extracted from a large number of environmental and ecological indicators, the data dimension is optimized, and data redundancy is effectively reduced, while ensuring that the selected indicators can accurately reflect the key characteristics of green watershed construction. The principal component analysis method compresses complex indicator sets into a few principal components, which have a high variance explanation ability and can significantly improve the effectiveness and scientificity of indicator screening. Indicators that are of reference significance for green watershed construction can be screened out from typical indicator systems.

[0042] ②Apply within the scope of green watershed construction;

[0043] By applying the key indicators selected by principal component analysis to the environmental assessment and management of green watershed construction, the ecological data in the green watershed can be analyzed to identify the core indicators that have an important impact on the ecological health and sustainable development of the watershed. These indicators will be used to evaluate and monitor the environmental quality of the watershed, guide the formulation of watershed management policies, and optimize resource allocation to achieve the sustainable development and ecological protection goals of green watershed construction.

[0044] Compared with the prior art, the present invention has the following technical effects:

[0045] Compared with the traditional multi-index selection method, the present invention screens the key indicators in the green watershed construction index system through principal component analysis, which significantly improves the accuracy and efficiency of index screening. Experimental data show that compared with the traditional multi-index screening method, the present invention reduces the number of indicators by about 30%, while retaining more than 75% to 90% of the variance contribution rate indicators, thereby reducing data redundancy and improving the accuracy of data analysis.

[0046] During implementation, the present invention can reduce the original complex data processing time by 70%, greatly improving the processing efficiency. In addition, due to the reduction of data dimension, the computing resources and storage requirements are significantly reduced, thereby reducing energy consumption and system load, and achieving environmental protection goals. The processed data is more concise, easy to operate and control, which simplifies the environmental assessment and management process of green watershed construction and reduces the labor intensity of operators.

[0047] By analyzing the experimental data in the actual application of green watershed construction, the present invention shows good application effects. For example, in the following case showing green watershed construction, after applying the index screening method of the present invention, the accuracy of environmental quality assessment and the rate of screening indicators are improved. These results prove that the present invention not only optimizes the index screening process, but also enhances the scientificity and operability of green watershed management, thereby improving the overall effect of environmental management and ecological protection of green watershed construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a gravel map of Example 1 of the present invention;

[0049] Figure 2 It is the factor loading matrix heat map of the present invention;

[0050] Figure 3 This is a gravel diagram of Example 2 of the present invention.

[0051] Specific implementation method

[0052] The specific technical solution of the present invention is explained in conjunction with two embodiments.

[0053] 1. This embodiment 1 takes the happy rivers and lakes as the research object.

[0054] (1) Variance explanation table

[0055] By analyzing the variance explanation table and scree plot, the number of principal components was determined according to their contribution to the explanation of the variables.

[0056] Table 1 Variance explanation table

[0057]

[0058]

[0059] The variance explanation table shows that at the sixth principal component, the characteristic root of the total variance explained is less than 1, and the contribution rate of the variable explanation reaches 93.962. Therefore, the first six principal components are selected as the main factors to explain the data variation.

[0060] (2) Gravel map

[0061] The scree plot helps determine the number of principal components to choose by showing how much each principal component contributes to the degree of explanation of the data variation. Figure 1 The scree plot shows that the eigenvalues ​​after the sixth principal component decrease rapidly, and the change trend is gentle. The characteristic root of the total variance explained is less than 1, and there is a significant change between principal components 6 and 7. This verifies the rationality of selecting 6 principal components.

[0062] (3) Factor loading coefficient table

[0063] The factor loading coefficients are shown in Table 2, which are used to show the strength of the relationship between each original variable and the principal component. By analyzing the factor loading coefficients, we can identify which variables contribute more to the principal component and decide whether to retain these variables for further analysis and evaluation.

[0064] Table 2 Factor loading coefficient table

[0065]

[0066] (4) Factor loading matrix heat map

[0067] The factor loading matrix heat map is used in principal component analysis to visualize the loading coefficient between each original variable and the principal component. Through the change of color, the heat map intuitively shows the contribution of the variable to the principal component, thereby helping to identify the relationship between important variables and principal components. This graphical display helps to more clearly understand the principal component structure and the role of the variables, thereby simplifying the process of identifying key indicators.

[0068] like Figure 2 As shown in the figure, the indicators that contribute more to the principal component 1 include post-flood recovery capacity, water security rate, residents' life happiness index, urban and rural residents' water affinity index, and historical water culture protection and inheritance index; the indicators that contribute more to the principal component 2 include per capita water resources, groundwater resource protection index, ecological flow compliance rate of important rivers and lakes, and aquatic biological integrity index; the indicators that contribute more to the principal component 3 include the water landscape influence index; the indicators that contribute more to the principal component 4 include the mortality rate of people in flood disasters; the indicators that contribute more to the principal component 5 include the compliance rate of flood control projects; and the contributions to the principal component 6 are relatively low.

[0069] (5) Component matrix

[0070] The component matrix table shows the loading coefficient of each original variable on each principal component in the principal component analysis. The component matrix table is shown in Table 3, which provides the contribution of each variable to the principal component and helps analyze the degree of explanation of each principal component on the data variation.

[0071] Table 3 Ingredient matrix

[0072]

[0073]

[0074] (6) Factor weight analysis

[0075] As shown in Table 4, the weight calculation results of the principal component analysis show that the weight of principal component 1 is 34.73%, the weight of principal component 2 is 27.254%, the weight of principal component 3 is 17.158%, the weight of principal component 4 is 9.497%, the weight of principal component 5 is 6.204%, and the weight of principal component 6 is 5.156%. The maximum index weight is principal component 1 (34.73%), and the minimum is principal component 6 (5.156%).

[0076] Table 4 Factor weight analysis table

[0077] name Variance explained (%) Cumulative variance explained (%) Weight (%) Principal component 1 32.633 32.633 34.730 Principal component 2 25.608 58.241 27.254 Principal component 3 16.123 74.364 17.158 Principal component 4 8.923 83.287 9.497 Principal component 5 5.830 89.117 6.204 Principal component 6 4.845 93.962 5.156

[0078] The standardized weights are used to perform weighted summation of the principal components to form the final comprehensive factor:

[0079]

[0080] The weights of principal components 1, 2, and 3 are high, so the indicators that contribute more to principal components 1, 2, and 3 are selected as the main indicators of the happy rivers and lakes index system: they are post-flood recovery capacity, water security rate, residents' life happiness index, urban and rural residents' water affinity index, and historical water culture protection and inheritance index; per capita water resources possession, groundwater resources protection index, ecological flow compliance rate of important rivers and lakes, and aquatic biological integrity index; water landscape influence index.

[0081] (7) Verification and optimization

[0082] According to the needs of actual applications, the selected key indicators are interpreted and comprehensively evaluated to ensure that they are of practical significance for the evaluation of green watersheds. Combined with application feedback, the indicator system is continuously adjusted and optimized to ensure its long-term effectiveness and applicability.

[0083] 2. This second embodiment takes the green development index as its object.

[0084] In the second embodiment, the green development index is taken as the research object, and the principal component analysis method is used to analyze and optimize the index system to screen out the core indicators that have a significant impact on the construction of green watersheds.

[0085] (1) Variance explanation table

[0086] The number of principal components was determined by analyzing the variance explained table and scree plot shown in Table 5 .

[0087] Table 5 Variance explanation table

[0088]

[0089]

[0090]

[0091] Table 5 shows that when the number of principal components reaches 15, the characteristic root of the total variance explained is less than 1, and the cumulative variance explanation rate contribution rate reaches 91.477%. Therefore, the first 15 principal components are selected to ensure sufficient explanatory power.

[0092] (2) Gravel map

[0093] Figure 3 A scree plot is shown for determining the appropriate number of principal components. Figure 3 From the changing trend of the eigenvalues ​​of the principal components, it can be seen that when the number of principal components reaches 15, the characteristic root of the total variance explained is lower than 1, and the eigenvalue begins to flatten out, which verifies the rationality of selecting the first 15 principal components.

[0094] (3) Factor loading coefficient table

[0095] The factor loading coefficient table shown in Table 6 shows the contribution of each original variable in each principal component. The indicators that contribute more to principal component 1 are per capita GDP, energy consumption per unit GDP, sulfur dioxide emissions per unit GDP, nitrogen oxide emissions per unit GDP, per capita urban life electricity consumption, comprehensive utilization rate of industrial solid waste, per capita sulfur dioxide emissions, per capita newly added afforestation area in the year, and chemical oxygen demand removal rate of industrial wastewater; the indicators that contribute more to principal component 3 are per capita ammonia nitrogen emissions; the contribution of each indicator to other principal components is relatively low.

[0096] Table 6 Factor loading coefficient table

[0097]

[0098]

[0099] (4) Component matrix

[0100] The component matrix table shown in Table 7 shows the load coefficient of each original variable on each principal component, which helps analyze how each principal component explains the variation of the data. By analyzing the component matrix, we can identify which variables have a large load on the main principal components and then select them as the core indicators of the green development index.

[0101] Table 7 Ingredient matrix

[0102]

[0103]

[0104] (5) Factor weight analysis

[0105] The factor weight analysis table shown in Table 8 shows the explanation rate of each principal component for the total variance and the cumulative variance explanation rate. By calculating the weight of each principal component, the indicators that have an important impact on the green development index can be further screened. The weight of principal component 1 is 27.853%, the weight of principal component 2 is 15.803%, and the weight of principal component 3 is 9.995%. The maximum indicator weight is principal component 1 (27.853%), and the minimum is principal component 15 (1.912%).

[0106] Table 8 Factor weight analysis table

[0107]

[0108]

[0109] The standardized weights are used to perform weighted summation of the principal components to form the final comprehensive factor:

[0110] F=(0.255 / 0.915)×F1+(0.145 / 0.915)×F2+(0.091 / 0.915)×F3+(0.08 / 0.915 )×F4+(0.056 / 0.915)×F5+(0.049 / 0.915)×F6+(0.041 / 0.915)×F7+(0.036 / 0. 915)×F8+(0.029 / 0.915)×F9+(0.027 / 0.915)×F10+(0.025 / 0.915)×F11+(0.023 / 0.915)×F12+(0.02 / 0.915)×F13+(0.02 / 0.915)×F14+(0.017 / 0.915)×F15

[0111] The weights of principal components 1, 2, and 3 are high, so the indicators that contribute more to principal components 1, 2, and 3 are selected as the main indicators of the green development index system. They are per capita GDP, energy consumption per unit of GDP, sulfur dioxide emissions per unit of GDP, nitrogen oxide emissions per unit of GDP, per capita urban life electricity consumption, comprehensive utilization rate of industrial solid waste, per capita sulfur dioxide emissions, per capita newly added afforestation area in the year, chemical oxygen demand removal rate of industrial wastewater, and per capita ammonia nitrogen emissions.

[0112] (6) Optimization and verification of the indicator system

[0113] Combined with the calculation results of the green development index, actual application verification is carried out to ensure the scientificity and practicality of the index system. Through the analysis of test data, the calculation model is continuously optimized, and the weights and screening criteria are adjusted so that the green development index can more accurately reflect the green development status and changing trends of the region.

Claims

1. A method for selecting multi-factor indicators for green watershed construction, characterized in that: The main steps include: S1. Data collection and acquisition; Clarify the definition and scope of green watershed construction and identify environmental factors that need to be monitored and analyzed; Collect data on ecological, environmental, economic and social indicators related to green watershed construction; S2, data preprocessing; Preprocess the collected raw data, including data cleaning and missing value processing, to ensure the integrity and reliability of the data; S3, principal component analysis, including: constructing data matrix; data standardization; calculating covariance matrix; calculating eigenvalues ​​and eigenvectors; calculating cumulative contribution rate; writing principal component expression; S4, indicator screening; According to the results obtained from principal component analysis, the eigenvalues ​​were sorted and key indicators that made significant contributions to the evaluation of green watershed construction were selected; S5, verification and optimization; Explain the selected key indicators and explain the specific role and significance of each indicator in the green watershed construction assessment; at the same time, combine the actual needs and application scenarios to conduct a comprehensive assessment of the selected key indicators, consider the measurability, operability and practical application value of the indicators, verify the rationality of the screening results, and ensure that the selected indicators can effectively reflect the characteristics of the green watershed; Continuously optimize based on the new indicator system data, and regularly evaluate and adjust the selected indicators.

2. The method for selecting multi-factor indicators for green watershed construction according to claim 1 is characterized in that: S1 selects multi-factor indicators covering water resources, water environment, water ecology and water economy, and covers multiple dimensions such as green development, healthy water ecosystem, harmony between man and water, and regional coordination.

3. The method for screening multi-factor indicators for green watershed construction according to claim 1 is characterized in that: The principal component analysis in S3 specifically includes: (1) Construct a data matrix; For p indicators, there are n samples, and an n×p sample matrix can be constructed: (2) Data standardization; Calculate the mean and standard deviation respectively: Mean: Standard Deviation: Normalize data within a matrix: The initial sample matrix is ​​normalized: (3) Calculate the covariance matrix: in, get: (4) Calculate eigenvalues ​​and eigenvectors: Eigenvalue:λ1≥…≥λ p ≥0 Eigenvectors: (5) Calculate the cumulative contribution rate: where i=1,…,p (6) Write the principal component expression: The i-th principal component: F i =a 1i X1+…+a pi X p .

4. The method for selecting multi-factor indicators for green watershed construction according to claim 1 is characterized in that: The key indicator for S4 screening requires the cumulative variance contribution rate to reach 75% or 90%.

Images