A coastal zone ecological disaster mitigation evaluation method and device based on random subspace, equipment and storage medium

Abstract

The application provides a random subspace-based coastal zone ecological disaster mitigation assessment method, device, equipment and storage medium, a sample matrix is constructed, and a feature transformation matrix, a discrete coding matrix and a clustering center matrix are initialized, in an iterative updating process, a similarity matrix of each sample belonging to a feature transformation matrix array vector set is calculated, different types of coastal zone samples can be associated to different feature subspaces, and therefore the limitation of a single subspace assumption is broken through; finally, key risk assessment indexes are screened out according to L2 norm ordering of each row vector of the feature transformation matrix, and targeted ecological disaster mitigation risk assessment on a coastal zone region with spatial heterogeneity is realized.

Description

Technical Field

[0001] This invention relates to the field of data mining, and in particular to a method, apparatus, equipment, and storage medium for coastal ecological disaster reduction assessment based on random subspace. Background Technology

[0002] The coastal zone, where land and sea meet, is ecologically fragile and subject to complex and variable disaster risks due to multiple factors including ocean dynamics, geology, climate change, and human activities. Scientific assessment of coastal zone ecological disaster risk is crucial for ecological disaster mitigation decision-making.

[0003] Coastal ecological disaster risk assessment requires comprehensive consideration of multiple dimensions, including annual average significant wave height, isobath distance, coastline type, coastal lowlands, coastline slope, built-up area, vegetation cover, population density, tourist arrivals, and coastline change rate. However, these indicators exhibit complex correlations and redundancies. Directly using all indicators for risk assessment is not only computationally expensive but may also lead to decreased accuracy due to noise and redundant information. Therefore, identifying the most representative subset of key features from high-dimensional coastal zone assessment indicators is crucial for improving the efficiency and accuracy of risk assessment.

[0004] Existing feature selection methods face a key technical bottleneck when processing coastal zone ecological risk assessment data: these methods are generally based on the single subspace assumption, which assumes that the linear combination of features of all coastal zone samples lies in the same low-dimensional subspace. However, the actual coastal zone environment exhibits significant spatial heterogeneity. Different coastal zones are influenced by different dominant factors; for example, some areas are dominated by ocean dynamics, while others are dominated by human activities. This results in the feature expressions of different types of coastal zones actually being distributed across multiple different subspaces. Under the single subspace assumption, existing methods struggle to accurately capture the differentiated correlations between different coastal zone samples and various assessment indicators, thus failing to identify truly distinctive key risk indicators for different types of coastal zone areas, ultimately affecting the accuracy of ecological risk assessment.

[0005] In view of the above, this application is hereby submitted. Summary of the Invention

[0006] This invention discloses a method, apparatus, equipment, and storage medium for coastal zone ecological disaster reduction assessment based on random subspace. It aims to solve the problem that existing feature selection methods, which are based on the assumption of a single subspace, cannot adapt to the spatial heterogeneity of the coastal environment, resulting in difficulty in accurately identifying key risk assessment indicators for different types of coastal areas.

[0007] The first embodiment of the present invention provides a coastal zone ecological disaster reduction assessment method based on random subspace, comprising: Obtain the raw indicator data for coastal zone ecological disaster reduction risk assessment, construct sample matrix X, and initialize feature transformation matrix W, discrete coding matrix Z, and cluster center matrix B; Based on the sample matrix X, iteratively update the cluster center matrix B, the discrete encoding matrix Z, and the feature transformation matrix W, and calculate the similarity matrix F of the column vector set to which each sample belongs to the feature transformation matrix W; When the termination condition is met, the row vectors in the feature transformation matrix W are sorted according to their L2 norm, and the indicators corresponding to the first k row vectors are selected as key risk assessment indicators. Based on the key risk assessment indicators, an ecological disaster reduction risk assessment is carried out on the coastal zone to be assessed.

[0008] Preferably, the raw index data includes annual average significant wave height, isobath distance from shore, coastline type, coastal lowlands, coastline slope, built-up area, vegetation cover, population density, number of tourists, and coastline change rate.

[0009] Preferably, the feature transformation matrix W is an m-row, r-column real matrix, initialized as an arbitrary matrix, where m is the feature dimension of the coastal zone risk assessment index and r is the projection dimension; The cluster center matrix B is an r-row, c-column real matrix, initialized as a column orthogonal matrix, where c is the number of clusters in the coastal zone region; The discrete coding matrix Z is a c-row, n-column real matrix, initialized as a column-orthogonal and non-negative matrix, where n is the number of coastal zone samples.

[0010] Preferably, updating the cluster center matrix B includes: According to the formula R=ZX T W calculates the auxiliary variable matrix R, where Z is the discrete coding matrix, and X... T Let X be the transpose of the sample matrix X, and W be the feature transformation matrix; Perform singular value decomposition on the auxiliary variable matrix R to obtain the left singular matrix U. R Singular value diagonal matrix Σ R and the right singular matrix V R , among which, U R Each column of V is a left singular vector of R. R Each column is a right singular vector of R, Σ R It is a diagonal matrix whose diagonal elements are singular values ​​of R; Based on the cyclic permutation invariance of the trace function, an equivalent expression for the objective function is calculated, specifically by summing σ over i from 1 to c. ii Multiply by ψii , where σ ii Σ is a singular valued diagonal matrix R The i-th diagonal element, ψ ii Let Ψ be the i-th diagonal element of the intermediate matrix Ψ, which is given by the formula Ψ=V R T BU R Calculations show that V R T V is a right singular matrix R The transpose of the matrix; The objective function reaches its maximum value when the intermediate matrix Ψ is a diagonal identity matrix, thus obtaining the cluster center matrix used to characterize the clustering characteristics of different types of coastal zone regions. Its update formula is B=V R *I r*c *U R T , where I r*c Let U be an r-row, c-column identity matrix. R T For the left singular matrix U R The transpose of .

[0011] Preferably, updating the discrete coding matrix Z includes: Construct an auxiliary variable matrix Q, transform the objective function into a least-squares optimization problem with square F-norm regularization, and separate the orthogonal constraints and non-negativity constraints of the discrete encoding matrix Z in a decoupling manner, where the F-norm is the square root of the sum of the squares of all elements of the matrix; The auxiliary variable matrix Q is updated based on the non-negative least squares problem, so that all elements of Q are non-negative. Perform singular value decomposition on the matrix containing the auxiliary variable matrix Q and the regularization term hyperparameter λ to obtain the left singular matrix U. R and the right singular matrix V R , where λ is a positive real number that controls the strength of regularization; Based on the left singular matrix U R and the right singular matrix V R Calculate the discrete encoding matrix Z that satisfies the Stifel manifold constraint, which requires that the column vectors of Z be mutually orthogonal and be unit vectors. The discrete encoding matrix Z is used to characterize the membership relationship between each coastal zone sample and the cluster center.

[0012] Preferably, updating the feature transformation matrix W includes: Construct an auxiliary variable matrix A, and use the augmented Lagrange method to decouple the L2,1 norm regularization term. The L2,1 norm is calculated by first calculating the square root of the sum of squares of the elements in each row of the matrix to obtain the L2 norm of that row, and then summing the L2 norms of all rows. The L2,1 norm is used to induce row sparsity of the feature transformation matrix W to achieve the screening of coastal zone risk assessment indicators. The optimization problem of L1 norm regularization is transformed into an equivalent objective function based on the iterative reweighting algorithm, where the L1 norm is the sum of the absolute values ​​of the elements of the vector; The scalar terms are transformed into matrix form using the trace function, and the feature transformation matrix W is updated based on the alternating minimization algorithm. The trace function is the sum of the diagonal elements of the matrix. The Lagrange multiplier matrix is ​​updated based on the difference between the auxiliary variable matrix A and the feature transformation matrix W. The feature transformation matrix W is used to map the high-dimensional coastal zone index data to a low-dimensional risk feature space.

[0013] Preferably, calculating the similarity matrix F includes: for the i-th coastal zone sample, according to the formula... Calculate the similarity vector f i , in Let x represent the projection of the coastal zone sample x onto the d-th column vector of the feature transformation matrix W. The L1 norm measure of the projection. The L2 norm measure of the projection is used to solve the similarity matrix F based on the closed simplex constraint. The similarity matrix F is used to measure the correlation strength between each coastal zone sample and different risk feature subspaces.

[0014] The second embodiment of the present invention provides a coastal zone ecological disaster reduction assessment device based on random subspace, comprising: The initialization unit is used to acquire the original indicator data for coastal zone ecological disaster reduction risk assessment, construct the sample matrix X, and initialize the feature transformation matrix W, discrete coding matrix Z, and cluster center matrix B. The update unit is used to iteratively update the cluster center matrix B, the discrete encoding matrix Z, and the feature transformation matrix W based on the sample matrix X, and to calculate the similarity matrix F of the set of column vectors to which each sample belongs; The risk assessment unit is used to sort the row vectors in the feature transformation matrix W according to the L2 norm when the termination condition is met, select the indicators corresponding to the first k row vectors as key risk assessment indicators, and conduct ecological disaster reduction risk assessment on the coastal zone to be assessed based on the key risk assessment indicators.

[0015] The third embodiment of the present invention provides a coastal zone ecological disaster reduction assessment device based on random subspace, including a memory and a processor. The memory stores a computer program, which can be executed by the processor to implement a coastal zone ecological disaster reduction assessment method based on random subspace as described in any of the above embodiments.

[0016] The fourth embodiment of the present invention provides a computer-readable storage medium, characterized in that it stores a computer program, which can be executed by the processor of the device where the computer-readable storage medium is located, to implement the coastal zone ecological disaster reduction assessment method based on random subspace as described in any of the above claims.

[0017] Based on the coastal zone ecological disaster reduction assessment method, apparatus, equipment, and storage medium provided by this invention, a sample matrix is ​​constructed and a feature transformation matrix, a discrete coding matrix, and a cluster center matrix are initialized. During the iterative update process, the similarity matrix of each sample belonging to the column vector set of the feature transformation matrix is ​​calculated, so that different types of coastal zone samples can be associated with different feature subspaces, thereby breaking through the limitation of the single subspace assumption. Finally, key risk assessment indicators are selected by sorting the L2 norm of each row vector of the feature transformation matrix, so as to realize targeted ecological disaster reduction risk assessment for coastal zone areas with spatial heterogeneity. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a downhill assist method for a pure electric vehicle provided in the first embodiment of the present invention. Figure 2 This is a schematic diagram of a downhill assist device for a pure electric vehicle provided in the second embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To better understand the technical solution of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] This invention discloses a method, apparatus, equipment, and storage medium for coastal zone ecological disaster reduction assessment based on random subspace. It aims to solve the problem that existing feature selection methods, which are based on the assumption of a single subspace, cannot adapt to the spatial heterogeneity of the coastal environment, resulting in difficulty in accurately identifying key risk assessment indicators for different types of coastal areas.

[0022] The first embodiment of the present invention provides a coastal zone ecological disaster reduction assessment method based on random subspace, which can be executed by a coastal zone ecological disaster reduction risk assessment device (hereinafter referred to as assessment device or system), specifically, by one or more processors within the assessment device, to at least achieve the following steps: S101, Obtain the original indicator data for coastal zone ecological disaster reduction risk assessment, construct sample matrix X, and initialize feature transformation matrix W, discrete coding matrix Z and cluster center matrix B; In this embodiment, the evaluation device can be a desktop computer, laptop computer, server, or other terminal with data processing capabilities. The evaluation device can be equipped with a corresponding operating system and application software, and the functions required in this embodiment can be realized through the combination of the operating system and application software.

[0023] In coastal zone ecological disaster risk assessment, the first step is to obtain raw indicator data for the coastal zone to be assessed. This raw indicator data is constructed based on the causal relationship between sources, pathways, receptors, and impacts, forming a complete coastal zone ecological disaster risk assessment system.

[0024] Specifically, the source index reflects the dynamic effects of the ocean, and the annual average significant wave height is selected as the core index to characterize the energy of ocean erosion. The pathway indicators reflect the impact of the natural conditions of the coastal zone on the transmission of ocean dynamics. They include four indicators: isobath distance from the shore, coast type, coastal lowlands, and coast slope. The isobath distance from the shore reflects the characteristics of nearshore water depth variation. Coast type distinguishes different landform types such as bedrock coast, sandy coast, and silty coast. Coastal lowlands characterize the distribution of low-lying areas that are easily submerged. Coast slope reflects the steepness of the coastal topography. Receptor indicators reflect the degree of human economic, social and ecological exposure in coastal areas, including four indicators: built-up land area, vegetation cover, population density and tourist visits, which respectively characterize the intensity of construction and development, ecological protection capacity, resident population size and floating population pressure; The influencing indicators reflect the final result of coastal erosion, and the rate of coastline change is selected as the indicator characterizing the dynamic changes of the coastline. The ten indicators corresponding to each coastal zone region are arranged into a sample vector, and the sample vectors of all coastal zone regions are arranged in rows to form a sample matrix X. Subsequently, the feature transformation matrix W, the discrete coding matrix Z, and the cluster center matrix B are initialized. The feature transformation matrix W is an m-row, r-column real matrix initialized to an arbitrary matrix, where m is the feature dimension of the coastal zone risk assessment indicators (i.e., the ten indicators), and r is the projection dimension. The cluster center matrix B is an r-row, c-column real matrix initialized to a column orthogonal matrix, where c is the number of clusters in the coastal zone region, and column orthogonality means that the column vectors of the matrix are mutually orthogonal. The discrete coding matrix Z is a c-row, n-column real matrix initialized to a column orthogonal and non-negative matrix, where n is the number of coastal zone samples, and non-negativity means that all elements of the matrix are greater than or equal to zero.

[0025] S102, based on the sample matrix X, iteratively update the cluster center matrix B, the discrete encoding matrix Z, and the feature transformation matrix W, and calculate the similarity matrix F of the column vector set to which each sample belongs to the feature transformation matrix W; In the process of iteratively updating the cluster center matrix B, firstly according to the formula R=ZX T W calculates the auxiliary variable matrix R, which is the transpose of the discrete coding matrix Z and the sample matrix X. T The feature transformation matrix W is then multiplied to obtain an auxiliary variable matrix R that integrates the clustering membership information and feature transformation information of the coastal zone samples. Subsequently, singular value decomposition is performed on the auxiliary variable matrix R, decomposing it into a product of three matrices to obtain the left singular matrix U. R Singular value diagonal matrix Σ R and the right singular matrix V R The left singular matrix U R Each column is a left singular vector of the auxiliary variable matrix R, and the right singular matrix V. R Each column is a right singular vector of the auxiliary variable matrix R, and the singular value diagonal matrix Σ R Let R be a diagonal matrix whose diagonal elements are the singular values ​​of the auxiliary variable matrix R. These singular values ​​are arranged in descending order and reflect the importance of the directions of each principal component. Based on the relationship between the squared F norm and the trace function of a matrix, the original objective function is transformed into an equivalent expression by utilizing the cyclic permutation invariance of the trace function. Specifically, the formula is to sum σ over i from 1 to c. ii Multiply by ψ ii , where σ ii Σ is a singular valued diagonal matrix R The i-th diagonal element is the i-th singular value, ψ ii Let Ψ be the i-th diagonal element of the intermediate matrix Ψ, which is given by the formula Ψ=V.R T BU R The calculation shows that this formula will transform the right singular matrix V R The transpose matrix, cluster center matrix B, and left singular matrix U R Perform matrix multiplication. According to the definition of the trace function and matrix inequalities, when the intermediate matrix Ψ is a diagonal identity matrix, the objective function reaches its maximum value because all singular values ​​are non-negative and the absolute value of the diagonal elements of the intermediate matrix Ψ is less than or equal to 1. In this case, the update formula for the cluster center matrix B is B = V. R *I r*c *U R T , where I r*c An r-row, c-column identity matrix is ​​used to maintain matrix dimension matching, U R T For the left singular matrix U R The transpose of the matrix. The cluster center matrix B obtained after the above update process satisfies the column orthogonality condition, which can characterize the clustering characteristics of different types of coastal zone regions and provide a vertical cluster center benchmark for subsequent identification of the category of each coastal zone sample.

[0026] In the iterative update of the discrete encoding matrix Z, since Z needs to simultaneously satisfy two conditions—column orthogonality and non-negativity—direct solution is difficult. Therefore, an auxiliary variable matrix Q is introduced to decouple these two constraints. Specifically, by setting the constraint Z=Q, the original objective function is transformed into a least-squares optimization problem with square F-norm regularization, where the F-norm is the square root of the sum of the squares of all elements of the matrix. This transformation allows the orthogonality and non-negativity constraints of the discrete encoding matrix Z to be handled separately in a decoupled manner, reducing the difficulty of solving the optimization problem. When updating the auxiliary variable matrix Q, the solution is based on the decoupled non-negativity least squares problem. The projection operation ensures that all elements of the auxiliary variable matrix Q are non-negative, thus guaranteeing that the final discrete encoding matrix Z satisfies the non-negativity requirement. This non-negativity constraint conforms to the physical meaning of the membership degree of coastal zone samples to each cluster center, i.e., the membership degree should not be negative. When updating the discrete encoding matrix Z, a matrix containing the auxiliary variable matrix Q and the regularization hyperparameter λ is first constructed. The regularization hyperparameter λ is a positive real number that controls the strength of regularization and is used to balance the weight relationship between the fitting terms and constraint terms of the objective function. Then, singular value decomposition is performed on this matrix to obtain the corresponding left singular matrix U. R and the right singular matrix V R The left singular matrix U obtained based on singular value decomposition. R and the right singular matrix V R Through the formula Z=U R V R TThe discrete coding matrix Z satisfying the Stifel manifold constraint is calculated. The Stifel manifold constraint requires that the column vectors of the discrete coding matrix Z be mutually orthogonal and unit vectors. This constraint ensures that the cluster membership of each coastal zone sample has clear class distinction. The discrete coding matrix Z obtained after the above update process satisfies both the column orthogonality and non-negativity constraints, which can accurately characterize the membership relationship between each coastal zone sample and the cluster center, laying the foundation for identifying key risk indicators of different types of coastal zone areas in the subsequent feature selection process.

[0027] In the iterative update of the feature transformation matrix W, the objective function contains an L2,1 norm regularization term, which is non-smooth and difficult to solve directly. Therefore, an auxiliary variable matrix A is introduced, and the augmented Lagrange method is used to decouple the L2,1 norm regularization term. The L2,1 norm is calculated by first taking the square root of the sum of the squares of the elements in each row of the matrix to obtain the L2 norm of that row, and then summing the L2 norms of all rows. This norm form can induce row sparsity in the feature transformation matrix W, that is, encourage the elements of the entire row to approach zero, thereby realizing the automatic screening of coastal zone risk assessment indicators. When the L2 norm of a row is small, it indicates that the coastal zone assessment indicator corresponding to that row has low importance and can be eliminated. By setting the constraint A=W, the augmented Lagrange method is used to transform the original optimization problem into a subproblem of alternately optimizing the auxiliary variable matrix A and the feature transformation matrix W. When updating the auxiliary variable matrix A, the optimization problem of L1 norm regularization is transformed into an equivalent objective function based on the iterative reweighting algorithm. The L1 norm is the sum of the absolute values ​​of the vector elements. The iterative reweighting algorithm transforms the non-smooth L1 norm problem into a series of weighted smooth optimization problems for iterative solution by introducing a diagonal weight matrix. When updating the eigenvalue transformation matrix W, the scalar terms in the objective function are uniformly transformed into a matrix-based trace function form using the definition of the trace function. The trace function is the sum of the diagonal elements of the matrix. This transformation allows the objective function to be expressed in a unified matrix operation form, facilitating subsequent solution. Subsequently, the eigenvalue transformation matrix W is updated based on the alternating minimization algorithm. By fixing other variables and taking the partial derivatives of W with respect to zero, a closed-form solution for W is obtained. After updating the auxiliary variable matrix A and the eigenvalue transformation matrix W, the Lagrange multiplier matrix is ​​updated based on the difference between A and W. The Lagrange multiplier matrix is ​​used to measure the degree to which the constraint condition A=W is satisfied and guides the iterative process to gradually converge to the optimal solution where the constraint condition holds. The feature transformation matrix W obtained through the above update process can map the high-dimensional index data of the coastal zone to a low-dimensional risk feature space. At the same time, its row sparsity structure provides a basis for subsequent screening of key risk assessment indicators based on the L2 norm of the row vectors.

[0028] In calculating the similarity matrix F, it is necessary to measure the correlation strength between each coastal zone sample and different column vectors of the feature transformation matrix W, thereby identifying the feature subspace to which each coastal zone sample belongs. For the i-th coastal zone sample, its index vector x is first multiplied by the feature transformation matrix W to obtain the projection vector. This projection vector maps the coastal zone samples from the original high-dimensional index space to a low-dimensional feature space. Then, according to the formula... Calculate the similarity vector f i The d-th element, the numerator of the formula Represents the projection vector The absolute value of the d-th component, i.e., the absolute value of the projection of the coastal zone sample x onto the d-th column vector of the feature transformation matrix W, with the denominator being the projection vector. The sum of the absolute values ​​of each component, this formula is equivalent to ,in The L2 norm of vector Wx represents the square root of the sum of the squares of its components. Physically, similarity vectors f... i The d-th element This reflects a normalized measure of the projection intensity of the i-th coastal zone sample in the d-th feature subspace direction relative to the total projection intensity. A larger value indicates a stronger correlation between the coastal zone sample and the d-th feature subspace. The similarity matrix F is solved based on closed simplex constraints, which require a similarity vector f... i The constraint that the sum of all elements of the similarity vector equals 1 and all elements are non-negative ensures that the similarity vector has a probability distribution property, which can be interpreted as the probability that a coastal zone sample belongs to each feature subspace. Arranging the similarity vectors of all coastal zone samples column-wise constitutes the similarity matrix F. This matrix measures the correlation strength between each coastal zone sample and different risk feature subspaces, enabling different types of coastal zone samples to be associated with different feature subspaces. This overcomes the limitations of the single subspace assumption in traditional methods and more accurately captures the differentiated correlations between spatially heterogeneous coastal zones and various risk assessment indicators.

[0029] In one possible implementation of the present invention, the similarity vector f i The solution also includes: based on the vertical cluster center matrix B, the discrete encoding matrix Z, and the similarity vector f i The feature transformation matrix W is obtained by minimizing the objective function, where the objective function is: ,in This represents finding the minimum value of the characteristic transformation matrix W. Represents the similarity vector f i The vector formed by taking the square root of each element. This indicates element-wise multiplication, which involves multiplying corresponding elements of two vectors. Let W be the transpose of the characteristic transformation matrix. Let B be the index vector of the i-th coastal zone sample, B be the cluster center matrix, and Z be the discrete encoding matrix. Let β represent the square of the L2 norm, and β be a positive real parameter that controls the strength of regularization.

[0030] In one possible implementation of the present invention, calculating the similarity matrix F further includes: The similarity vector f of the i-th coastal zone sample is solved using the quadratic optimization formula. i The quadratic optimization formula is: Where min represents finding the minimum value, and the constraint conditions are... Represents the similarity vector f i The sum of all elements equals 1, constraint condition express All elements are non-negative. For similarity vectors f i The j-th element, w j Let J be the j-th column vector of the characteristic transformation matrix W. This is the transpose of the column vector. Let be the index vector of the i-th coastal zone sample. express and The absolute value of the inner product result, where η is a positive real number parameter controlling the strength of regularization, and r is the projection dimension. S103, when the termination condition is met, sort the row vectors in the feature transformation matrix W according to their L2 norm, select the indicators corresponding to the first k row vectors as key risk assessment indicators, and conduct ecological disaster reduction risk assessment on the coastal zone to be assessed based on the key risk assessment indicators.

[0031] It should be noted that in this embodiment, after completing the iterative updates of the cluster center matrix B, discrete encoding matrix Z, feature transformation matrix W, and similarity matrix F, it is necessary to determine whether the termination condition is met to decide whether to continue iterating. The termination condition includes two scenarios: first, the number of iterations reaches a preset number, which can be set according to actual computing resources and accuracy requirements; second, the difference between the objective function values ​​of two adjacent iterations is less than a preset threshold, indicating that the objective function has converged and further iteration is unlikely to yield significant performance improvements.

[0032] The iteration process stops when any of the above termination conditions are met, at which point the feature transformation matrix W has fully learned the feature structure information of the coastal zone sample data. Subsequently, the row vectors in the feature transformation matrix W are sorted in descending order based on their L2 norm. The L2 norm is the square root of the sum of the squares of the elements in each row vector, and its value reflects the importance of the corresponding coastal zone risk assessment indicator in the feature transformation process. A larger L2 norm indicates a greater contribution of the indicator to distinguishing different types of coastal zone areas. The indicators corresponding to the first k row vectors with the largest L2 norm are selected as key risk assessment indicators. These k indicators constitute the optimal feature subset for coastal zone ecological disaster reduction risk assessment, where k is the number of key risk assessment indicators selected, which can be determined based on the cumulative feature contribution rate or actual assessment needs. The first k row vectors form the optimal risk assessment feature subset W'. Based on this feature subset, ecological disaster reduction risk assessment is performed on the coastal zone area to be assessed, with the specific calculation formula being Y = W'. T X, where X is the raw indicator data of the coastal zone region to be evaluated, and W' T Let Y be the transpose of the optimal risk assessment feature subset W', and let Y be the dimensionality-reduced risk assessment feature data of the coastal zone to be assessed. This formula projects the high-dimensional original index data of the coastal zone to be assessed onto a low-dimensional feature space spanned by key risk assessment indicators. The resulting risk assessment feature data Y eliminates redundant and noisy information, retaining the most discriminative key features for ecological disaster reduction risk assessment, thereby achieving efficient and accurate assessment of ecological disaster reduction risks in the coastal zone.

[0033] In one possible implementation of the present invention, the method further includes a quality assessment of the selection results of coastal zone ecological disaster reduction risk assessment features, using cluster identification accuracy for evaluation. The formula for calculating the cluster identification accuracy is as follows: Where n is the number of coastal zone samples. This represents the true risk level label for the i-th coastal zone region. Let h(·,·) be the cluster label for the i-th coastal zone region, h(·,·) be an indicator function that returns 1 if the two inputs are equal and 0 otherwise, and map(·) be a label mapping function implemented by the Kuhn-Munkres algorithm to map the cluster label set to the real label set.

[0034] In one possible implementation of the present invention, the quality assessment further employs standard mutual information entropy as an evaluation index for the effectiveness of coastal zone risk classification, and the formula for calculating the standard mutual information entropy is as follows: ,in For mutual information, t represents the set of true risk levels for coastal areas, and t' represents the set of cluster labels. Let be the larger of the entropy E(t) of set t and the entropy E(t') of set t'; the formula for calculating the mutual information MI(t,t') is as follows: , where p(x) is the probability that any sample belongs to the true risk level x, p(x') is the probability that any sample belongs to the cluster label x', and p(x,x') is the joint probability that any sample belongs to both the true risk level x and the cluster label x'. It represents the logarithm to the base 2.

[0035] Please see Figure 2 The second embodiment of the present invention provides a coastal zone ecological disaster reduction assessment device based on random subspace, comprising: Initialization unit 201 is used to acquire the original indicator data of coastal zone ecological disaster reduction risk assessment, construct sample matrix X, and initialize feature transformation matrix W, discrete coding matrix Z and cluster center matrix B; The update unit 202 is used to iteratively update the cluster center matrix B, the discrete encoding matrix Z, and the feature transformation matrix W based on the sample matrix X, and to calculate the similarity matrix F of the set of column vectors to which each sample belongs; Risk assessment unit 203 is used to sort the row vectors in the feature transformation matrix W according to the L2 norm when the termination condition is met, select the indicators corresponding to the first k row vectors as key risk assessment indicators, and conduct ecological disaster reduction risk assessment on the coastal zone to be assessed based on the key risk assessment indicators.

[0036] The third embodiment of the present invention provides a coastal zone ecological disaster reduction assessment device based on random subspace, including a memory and a processor. The memory stores a computer program, which can be executed by the processor to implement a coastal zone ecological disaster reduction assessment method based on random subspace as described in any of the above embodiments.

[0037] The fourth embodiment of the present invention provides a computer-readable storage medium, characterized in that it stores a computer program, which can be executed by the processor of the device where the computer-readable storage medium is located, to implement the coastal zone ecological disaster reduction assessment method based on random subspace as described in any of the above claims.

[0038] Based on the coastal zone ecological disaster reduction assessment method, apparatus, equipment, and storage medium provided by this invention, a sample matrix is ​​constructed and a feature transformation matrix, a discrete coding matrix, and a cluster center matrix are initialized. During the iterative update process, the similarity matrix of each sample belonging to the column vector set of the feature transformation matrix is ​​calculated, so that different types of coastal zone samples can be associated with different feature subspaces, thereby breaking through the limitation of the single subspace assumption. Finally, key risk assessment indicators are selected by sorting the L2 norm of each row vector of the feature transformation matrix, so as to realize targeted ecological disaster reduction risk assessment for coastal zone areas with spatial heterogeneity.

[0039] Exemplary examples show that the computer program described in the third and fourth embodiments of the present invention can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in implementing a coastal zone ecological disaster reduction assessment device based on a random subspace. For example, the apparatus described in the second embodiment of the present invention.

[0040] The processor referred to can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the proposed coastal zone ecological disaster reduction assessment method based on random subspace, connecting various parts of the method through various interfaces and lines.

[0041] The memory can be used to store the computer program and / or modules. The processor, by running or executing the computer program and / or modules stored in the memory, and by calling the data stored in the memory, implements various functions of a coastal zone ecological disaster reduction assessment method based on random subspace. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, text conversion function, etc.), etc.; the data storage area may store data created based on the use of the mobile phone (such as audio data, text message data, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0042] If the implemented module is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0043] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0044] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A coastal zone ecological disaster reduction assessment method based on random subspace, characterized in that, include: Obtain the raw indicator data for coastal zone ecological disaster reduction risk assessment, construct sample matrix X, and initialize feature transformation matrix W, discrete coding matrix Z, and cluster center matrix B; Based on the sample matrix X, iteratively update the cluster center matrix B, the discrete encoding matrix Z, and the feature transformation matrix W, and calculate the similarity matrix F of the column vector set to which each sample belongs to the feature transformation matrix W; When the termination condition is met, the row vectors in the feature transformation matrix W are sorted according to their L2 norm, and the indicators corresponding to the first k row vectors are selected as key risk assessment indicators. Based on the key risk assessment indicators, an ecological disaster reduction risk assessment is carried out on the coastal zone to be assessed.

2. The coastal zone ecological disaster reduction assessment method based on random subspace according to claim 1, characterized in that, The raw data include annual average significant wave height, isobath distance from shore, coastline type, coastal lowlands, coastline slope, built-up area, vegetation cover, population density, tourist arrivals, and coastline change rate.

3. The coastal zone ecological disaster reduction assessment method based on random subspace according to claim 1, characterized in that, The feature transformation matrix W is an m-row, r-column real matrix, initialized as an arbitrary matrix, where m is the feature dimension of the coastal zone risk assessment index and r is the projection dimension. The cluster center matrix B is an r-row, c-column real matrix, initialized as a column orthogonal matrix, where c is the number of clusters in the coastal zone region; The discrete coding matrix Z is a c-row, n-column real matrix, initialized as a column-orthogonal and non-negative matrix, where n is the number of coastal zone samples.

4. The coastal zone ecological disaster reduction assessment method based on random subspace according to claim 1, characterized in that, Updating the cluster center matrix B includes: According to the formula R=ZX T W calculates the auxiliary variable matrix R, where Z is the discrete coding matrix, and X... T Let X be the transpose of the sample matrix X, and W be the feature transformation matrix; Perform singular value decomposition on the auxiliary variable matrix R to obtain the left singular matrix U. R Singular value diagonal matrix Σ R and the right singular matrix V R , among which, U R Each column of V is a left singular vector of R. R Each column is a right singular vector of R, Σ R It is a diagonal matrix whose diagonal elements are singular values ​​of R; Based on the cyclic permutation invariance of the trace function, an equivalent expression for the objective function is calculated, specifically by summing σ over i from 1 to c. ii Multiply by ψ ii , where σ ii Σ is a singular valued diagonal matrix R The i-th diagonal element, ψ ii Let Ψ be the i-th diagonal element of the intermediate matrix Ψ, which is given by the formula Ψ=V R T BU R Calculations show that V R T V is a right singular matrix R The transpose of the matrix; The objective function reaches its maximum value when the intermediate matrix Ψ is a diagonal identity matrix, thus obtaining the cluster center matrix used to characterize the clustering characteristics of different types of coastal zone regions. Its update formula is B=V R *I r*c *U R T , where I r*c Let U be an r-row, c-column identity matrix. R T For the left singular matrix U R The transpose of .

5. The coastal zone ecological disaster reduction assessment method based on random subspace according to claim 1, characterized in that, Updating the discrete coding matrix Z includes: Construct an auxiliary variable matrix Q, transform the objective function into a least-squares optimization problem with square F-norm regularization, and separate the orthogonal constraints and non-negativity constraints of the discrete encoding matrix Z in a decoupling manner, where the F-norm is the square root of the sum of the squares of all elements of the matrix; The auxiliary variable matrix Q is updated based on the non-negative least squares problem, so that all elements of Q are non-negative. Perform singular value decomposition on the matrix containing the auxiliary variable matrix Q and the regularization term hyperparameter λ to obtain the left singular matrix U. R and the right singular matrix V R , where λ is a positive real number that controls the strength of regularization; Based on the left singular matrix U R and the right singular matrix V R Calculate the discrete encoding matrix Z that satisfies the Stifel manifold constraint, which requires that the column vectors of Z be mutually orthogonal and be unit vectors. The discrete encoding matrix Z is used to characterize the membership relationship between each coastal zone sample and the cluster center.

6. The coastal zone ecological disaster reduction assessment method based on random subspace according to claim 1, characterized in that, Updating the feature transformation matrix W includes: Construct an auxiliary variable matrix A, and use the augmented Lagrange method to decouple the L2,1 norm regularization term. The L2,1 norm is calculated by first calculating the square root of the sum of squares of the elements in each row of the matrix to obtain the L2 norm of that row, and then summing the L2 norms of all rows. The L2,1 norm is used to induce row sparsity of the feature transformation matrix W to achieve the screening of coastal zone risk assessment indicators. The optimization problem of L1 norm regularization is transformed into an equivalent objective function based on the iterative reweighting algorithm, where the L1 norm is the sum of the absolute values ​​of the elements of the vector; The scalar terms are transformed into matrix form using the trace function, and the feature transformation matrix W is updated based on the alternating minimization algorithm. The trace function is the sum of the diagonal elements of the matrix. The Lagrange multiplier matrix is ​​updated based on the difference between the auxiliary variable matrix A and the feature transformation matrix W. The feature transformation matrix W is used to map the high-dimensional coastal zone index data to a low-dimensional risk feature space.

7. The coastal zone ecological disaster reduction assessment method based on random subspace according to claim 1, characterized in that, Calculating the similarity matrix F includes: for the i-th coastal zone sample, according to the formula... Calculate the similarity vector f i , in Let x represent the projection of the coastal zone sample x onto the d-th column vector of the feature transformation matrix W. The L1 norm measure of the projection. The L2 norm measure of the projection is used to solve the similarity matrix F based on the closed simplex constraint. The similarity matrix F is used to measure the correlation strength between each coastal zone sample and different risk feature subspaces.

8. A coastal zone ecological disaster reduction assessment device based on random subspace, characterized in that, include: The initialization unit is used to acquire the original indicator data for coastal zone ecological disaster reduction risk assessment, construct the sample matrix X, and initialize the feature transformation matrix W, discrete coding matrix Z, and cluster center matrix B. The update unit is used to iteratively update the cluster center matrix B, the discrete encoding matrix Z, and the feature transformation matrix W based on the sample matrix X, and to calculate the similarity matrix F of the set of column vectors to which each sample belongs; The risk assessment unit is used to sort the row vectors in the feature transformation matrix W according to the L2 norm when the termination condition is met, select the indicators corresponding to the first k row vectors as key risk assessment indicators, and conduct ecological disaster reduction risk assessment on the coastal zone to be assessed based on the key risk assessment indicators.

9. A coastal zone ecological disaster reduction assessment device based on random subspace, characterized in that, The system includes a memory and a processor. The memory stores a computer program that can be executed by the processor to implement a coastal zone ecological disaster reduction assessment method based on a random subspace as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The device contains a computer program that can be executed by a processor of the device on which the computer-readable storage medium is located, to implement the coastal zone ecological disaster reduction assessment method based on random subspace as described in any one of claims 1 to 7.

Images