Method for establishing ecological resource intelligent credible data element model

Abstract

The application provides an ecological resource intelligent credible data element model establishment method, including: obtaining natural ecological basic data; obtaining natural ecological derived data, the natural ecological derived data including ecological tourism description metadata and leisure and health care description metadata; respectively performing controlled feature mining on the natural ecological basic data and the natural ecological derived data to generate ecological basic data elements and ecological derived data elements; respectively performing key feature extraction on the ecological basic data elements and the ecological derived data elements to obtain basic ecological key features and derived ecological key features; obtaining green finance data elements, and establishing an association relationship between the green finance data elements and the basic ecological key features and the derived ecological key features to generate an ecological resource credible data element model. The technical scheme in the application at least solves the problem that green finance data elements and ecological resource data lack effective association.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent processing, and in particular to a method for establishing an intelligent and credible data element model of ecological resources. BACKGROUND

[0002] In today's era of digitalization and sustainable development, effective management and rational utilization of ecological resources are increasingly critical. Ecological resources not only maintain the balance of the earth's ecology, but also serve as an important support for green economic and social development. Natural ecosystems encompass a variety of components such as forests, rivers, lakes, grasslands, and farmland, and the rich data they contain have immeasurable value for scientific decision-making, ecological protection, and green industry development. At the same time, with the increasing attention to the ecological environment, the fields of ecological tourism, leisure and health care, and other ecological derivatives are thriving, and related data are also increasing.

[0003] However, there are many problems in the field of ecological resource data management. On the one hand, the description metadata of natural ecological basic data, such as forests and rivers, are often scattered in different departments and systems, lacking effective integration and unified management, resulting in poor data consistency and usability. Moreover, when analyzing these basic data, it is difficult to comprehensively and accurately extract their controlled features, and it is difficult to provide strong data element support for scientific assessment of ecological resources. On the other hand, for natural ecological derivative data such as ecological tourism, leisure and health care, there are also problems of data fragmentation and uneven quality, and there is a lack of effective technical means for key feature extraction, making it difficult to accurately extract derivative ecological key features that are important for the comprehensive utilization and industrial development of ecological resources. In addition, green finance, as an important force to promote ecological protection and sustainable development, lacks an effective correlation mechanism between its data elements and ecological resource data, and cannot fully support the development and protection of ecological resources.

[0004] In summary, the existing ecological resource data management technology cannot meet the current needs of intelligent and credible management of ecological resources and the coordinated development of green industries, and an innovative method is needed to solve these problems. SUMMARY

[0005] To solve the above technical problems, the present application provides a method for establishing an intelligent and credible data element model of ecological resources to at least solve or alleviate the problems existing in the prior art.

[0006] To achieve the above purpose, according to one aspect of the present application, a method for establishing an intelligent and credible data element model of ecological resources is provided, which comprises:

[0007] acquire natural ecological basic data, the natural ecological basic data including at least one of forest description metadata, river description metadata, lake description metadata, grassland description metadata, farmland description metadata;

[0008] acquire natural ecological derived data, the natural ecological derived data including ecological tourism description metadata and leisure and health care description metadata;

[0009] respectively controlled feature mining on natural ecological basic data and natural ecological derived data to generate ecological basic data elements and ecological derived data elements;

[0010] respectively key feature extraction on ecological basic data elements and ecological derived data elements to obtain basic ecological key features and derived ecological key features;

[0011] acquire green finance data elements, and establish the correlation between the green finance data elements and the basic ecological key features and the derived ecological key features to generate ecological resource credible data element model.

[0012] The technical solution in the application has at least the following technical benefits:

[0013] (1) The present scheme acquires natural ecological basic data, covering at least one of forest description metadata, river description metadata, lake description metadata, grassland description metadata, and farmland description metadata. This operation breaks the status quo of data being scattered in different departments and systems, realizes the centralized acquisition of basic data, and lays a foundation for subsequent integration and unified management. Controlled feature mining on the acquired natural ecological basic data can comprehensively and accurately extract ecological basic data elements from these basic data, thereby providing strong data element support for scientific assessment of ecological resources and solving the problem of difficult effective controlled feature mining.

[0014] (2) The present application acquires natural ecological derived data, including ecological tourism description metadata and leisure and health care description metadata, and collects originally fragmented ecological derived data. Key feature extraction on these natural ecological derived data can accurately extract derived ecological key features, providing important guidance for comprehensive utilization of ecological resources and industrial development, and making up for the lack of effective key feature extraction techniques in this field.

[0015] (3) The present scheme acquires green finance data elements and establishes the correlation between the green finance data elements and the basic ecological key features and the derived ecological key features. Through the establishment of this correlation mechanism, the support of green finance for ecological resource development and protection can be fully exerted, the problem of lack of effective correlation between green finance data elements and ecological resource data is solved, and the demand for intelligent and credible management of ecological resources and coordinated development of green industry is met. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 This is a schematic diagram of a method for establishing an intelligent and trustworthy data element model of ecological resources according to an embodiment of this application. Detailed Implementation

[0017] Figure 1 This is a schematic diagram illustrating a method for establishing an intelligent and trustworthy data element model for ecological resources, as described in an embodiment of this application. Figure 1 As shown, it includes:

[0018] Acquire basic natural ecological data, which includes at least one of forest description metadata, river description metadata, lake description metadata, grassland description metadata, and cultivated land description metadata;

[0019] Acquire natural ecology-derived data, including ecotourism descriptive metadata and leisure and wellness descriptive metadata;

[0020] Controlled feature mining is performed on both basic natural ecological data and derived natural ecological data to generate basic ecological data elements and derived ecological data elements.

[0021] Key features were extracted from both basic ecological data elements and derived ecological data elements to obtain basic ecological key features and derived ecological key features.

[0022] Acquire green finance data elements and establish their correlation with basic ecological key features and derived ecological key features to generate a credible data element model of ecological resources.

[0023] Optionally, the acquisition of basic natural ecological data includes:

[0024] Based on the constructed multi-source data collection rules, carriers carrying different types of basic natural ecological data are analyzed to obtain basic natural ecological data.

[0025] Preferably, in a specific application scenario, the specific implementation process of the above solution is as follows:

[0026] Preferably, in the natural ecological basic data acquisition stage, there are n types of natural ecological basic data carriers, denoted as set V = {v1, v2, ..., v...} n}. Each carrier v i (1≤i≤n) carries a high-dimensional vector of basic natural ecological data. Where m i Let f represent the dimension of the i-th carrier data, reflecting the data complexity. The multi-source data acquisition rules are constructed as a non-linear mapping function f. c : It is based on an improved autoencoder structure. Autoencoder consists of an encoder e and a decoder d, the encoder e maps the input data D bi to a low-dimensional representation z i = e(D bi ), the decoder d reconstructs z i to The autoencoder is trained by minimizing the reconstruction error where λ is a regularization parameter that controls the sparsity of the low-dimensional representation. The trained encoder e is the nonlinear mapping function f c in the multi-source data acquisition rule. The processed data D bi ′ = f c (D bi ) is obtained by analyzing the carrier data. The final natural ecological basic data D b is: where ω i is the weight coefficient determined by the analytic hierarchy process (AHP), reflecting the importance of different carrier data. In the AHP method, a judgment matrix A = (a ij ) is constructed, where a ij represents the importance comparison of carrier i relative to carrier j, and the maximum eigenvalue λ max and its corresponding eigenvector of the judgment matrix are calculated by the eigenvalue method, and the weight coefficient ω i is obtained by normalizing the eigenvector.

[0027] The actual meaning of each parameter in the above formula is described in detail as follows: n: the number of types of natural ecological basic data carriers. v i : the i-th type of natural ecological basic data carrier. D bi : the high-dimensional vector of natural ecological basic data carried by the i-th carrier, with dimension m i varying with the type of carrier, reflecting the data richness. m i′ : the dimension of the data after processing by the multi-source data acquisition rule, which may be different from m i , determined by the autoencoder structure. f c : the nonlinear mapping function trained based on the autoencoder, used to analyze the carrier data, remove noise and redundant information, and extract key features. e: the encoder of the autoencoder, which maps high-dimensional data to a low-dimensional space. d: the decoder of the autoencoder, which reconstructs the original data from the low-dimensional representation. z i : the low-dimensional representation of the i-th carrier data in the autoencoder. L: the training loss function of the autoencoder, consisting of reconstruction error and regularization term, ensuring reconstruction accuracy and sparsity of low-dimensional representation. λ: regularization parameter, adjusting the balance between reconstruction error and sparsity of low-dimensional representation. ω i: Weight coefficient determined by AHP method, reflecting the contribution degree of the ith carrier data in the final natural ecological basic data. A: Judgment matrix in AHP, used to compare the relative importance of different carrier data. ij : Element in judgment matrix A, indicating the importance comparison of carrier i relative to carrier j. max : The largest eigenvalue of judgment matrix A.

[0028] Preferably, in the natural ecological derivation data acquisition link, the natural ecological derivation description text is a sequence T = [t1, t2, …, t N ], where t j (1≤j≤N) is the jth word unit in the text, and N is the length of the text. First, structured processing is performed, and the present application utilizes a model g s based on conditional random field (CRF). The CRF is defined on an undirected graph G = (V, E), where the vertices V correspond to the word units in the text, and the edges E represent the relationship between the words. Let x = [x1, x2, …, x N ] be the feature vector sequence of the input text, and y = [y1, y2, …, y N ] be the output structured label sequence. The joint probability distribution of the CRF is: where Z(x) is a normalization factor, λ k is a weight parameter of the feature function f k , which is trained by maximum likelihood estimation. After the CRF model processing, the structured text S = g s (T) is obtained.

[0029] Preferably, semantic analysis is performed on the structured text S, and the present application adopts a method based on deep semantic embedding (DSE). A deep neural network model g sa is constructed, the input of which is the structured text S, which is mapped to a high-dimensional semantic space through multiple hidden layers to obtain the natural ecological derivation feature F = g sa (S). The model is trained by minimizing the contrastive loss function (where M is the number of training samples, sim is the cosine similarity function, and [·] + represents taking the positive value), so that texts with similar semantics are closer in the high-dimensional space, and texts with different semantics are farther apart. Based on the natural ecological derivation feature F, the natural ecological derivation data D d is generated. A variant of the generative adversarial network (GAN), the conditional generative adversarial network (cGAN), is adopted. The generator G inputs the noise vector z and the natural ecological derivation feature F to generate the data D d =G(z,F); the discriminator D inputs the real natural ecological derivation data and the generated data Dd , to judge its authenticity. The cGAN minimizes the objective function through adversarial training:

[0030] The creative physical meaning of each parameter in the above formula is explained as follows: T: natural ecological derivative description text sequence. j : the jth word unit in the text. N: text length. s : structured processing model based on conditional random field. G: undirected graph in CRF, used to describe the relationship between text words. x: feature vector sequence of input text, containing word itself, part of speech, context, etc. y: output structured label sequence, such as sentence component label, etc. Z(x): normalization factor of CRF, to ensure the legality of probability distribution. k : weight parameter of feature function f k in CRF, which is learned through training to reflect the influence of different features on structured labels. f k : feature function of CRF, describing the relationship between labels and input text features. S: structured text after processing. sa : semantic analysis model based on deep semantic embedding. F: natural ecological derivative feature, which is the representation of structured text in high-dimensional semantic space. L dse : contrastive loss function of deep semantic embedding model, used to optimize the model to better capture semantic information. M: number of training samples. sim: cosine similarity function, measuring the similarity of two feature vectors in semantic space. G: generator of cGAN, generating natural ecological derivative data from noise vector and natural ecological derivative feature. z: noise vector, introducing randomness to generated data. D: discriminator of cGAN, judging the authenticity of generated data. : real natural ecological derivative data. L cgan : objective function of cGAN, improving the quality of generated data through adversarial training of generator and discriminator.

[0031] Preferably, in the controlled feature mining link, the data set composed of natural ecological basic data D b and natural ecological derivative data D d is D = {D b , D d}. The business rules and feature mining constraints are represented as a constraint function set where c l : is a real-valued function, for example, c lThe constraints can be constraints on statistical features of data (such as mean, variance range of data) or semantic relationships (such as logical relationships between certain ecological indicators). The controlled feature mining model is constructed as a hybrid model h based on a variational autoencoder-reinforcement learning (VAE-RL). The variational autoencoder part, for input data x e D, the encoder E maps it to the latent space z = E(x), where (μ(x) and σ 2 (x) are the mean and variance, respectively), and the decoder D generates reconstructed data from the latent space by minimizing the variational lower bound loss function (where D KL is the Kullback-Leibler divergence, which measures the difference between two probability distributions). The reinforcement learning part, define an agent A, whose state space S is the latent space z, and the action space A is the transformation operation on the latent space (such as scaling, translation, etc.). The agent selects an action a e A according to the current state s e S, and gets a reward r, the reward function R is designed to be related to the constraint condition, that is (where a l is the weight coefficient, representing the importance of different constraint conditions). The agent learns the optimal strategy π(s) by maximizing the long-term cumulative reward (where γ is the discount factor, which determines the importance of future rewards). Through the VAE-RL hybrid model h, the natural ecological basic data and natural ecological derived data are subjected to controlled feature mining, generating ecological basic data elements E b and ecological derived data elements E d :

[0032] E b = h(D b ), E d = h(D d ).

[0033] In the above formula, the creative physical meaning of each parameter is explained as follows: D: the data set composed of natural ecological basic data and natural ecological derived data. The constraint set composed of multiple constraint functions c l ensures that the feature mining conforms to business rules and specific constraints. h: controlled feature mining model based on variational autoencoder-reinforcement learning. E: the encoder of the variational autoencoder, which maps the input data to the latent space. D: the decoder of the variational autoencoder, which reconstructs data from the latent space. z: latent space variable, subject to normal distribution μ(x): mean of input data x in latent space. σ 2 (x): variance of input data x in latent space. L vae: Variational lower bound loss function of variational autoencoder, balancing reconstruction accuracy and reasonability of latent space distribution. KL : Kullback-Leibler divergence, used to measure the difference between latent space distribution and standard normal distribution. A: Reinforcement learning agent. S: State space of agent, which is latent space z. A: Action space of agent, containing transformation operations on latent space. r: Reward obtained by agent after performing action. R: Reward function, related to constraint condition, guiding agent learning by punishing actions violating constraints. a l : Constraint condition c l : Weight coefficient in reward function, reflecting the importance of different constraints. G t : Long-term cumulative reward of agent starting at time step t. γ: Discount factor, determining the influence of future rewards on current decision-making. π(s): Optimal policy of agent, selecting optimal action according to current state. E b : Generated ecological basic data elements. E d : Generated ecological derived data elements.

[0034] Preferably, in the key feature extraction link, the ecological basic data elements E b and the ecological derived data elements E d form an element set E = {E b , E d}. The key feature extraction direction is set as a high-dimensional vector field where x ∈ E, with dimension d θ , and its value changes with data point x to adapt to the feature extraction needs of different data regions. Multi-scale transformation is performed on E, using a combination method based on multi-resolution analysis (MRA) and discrete wavelet transform (DWT). Let be the scale parameter, for data sample x ∈ E, the approximate coefficient A s (x) and the detail coefficient D s (x) are obtained under scale s through DWT. Preferably, in order to combine the key feature extraction direction, the coefficients are weighted to obtain the weighted approximate coefficient and the weighted detail coefficient where, and are the weight vectors related to approximate coefficients and detail coefficients extracted from the vector field . After weighting, the feature information of ecological basic data elements and ecological derived data elements under different scales s is obtained

[0035] Preferably, the feature information at different scales The fusion analysis is performed, and the model f based on a graph neural network (GNN) is adopted in the present application f A graph structure G=(V, E) is constructed, where the vertices V are the feature information at different scales (i.e. and The edges E represent the relationship between the features (such as spatial adjacency relationship, semantic association relationship, etc.). Through the propagation and learning process of the GNN, the basic ecological key features K b and the derived ecological key features K d are obtained: wherein and are the feature information of the ecological basic data elements and the ecological derived data elements at the scale s, respectively.

[0036] In the above formula, the creative physical meaning of each parameter is explained as follows: E: the element set composed of the ecological basic data elements and the ecological derived data elements, which integrates the data after controlled feature mining, and provides a rich data basis for key feature extraction. Different types of data elements play different roles in subsequent analysis, and together help to extract representative key features. The high-dimensional vector field of the key feature extraction direction, with a dimension d θ determines the complexity and diversity of feature extraction. The vector field dynamically changes according to different data points x, and can sensitively capture the unique features of different regional data. For example, in the ecological basic data elements, for different attributes such as forest area and river flow, the vector field can adjust the direction weight to highlight the features related to the stability of the ecological system; in the ecological derived data elements, for data such as ecological tourism visitor flow and leisure and health service satisfaction, the vector field can focus on the key factors affecting the development of the industry.s: scale parameter, taking the value set as Different scales s represent different observation perspectives of different resolutions of data. Smaller scales focus on local details of data, such as the distribution details of specific tree species in a forest; larger scales focus on the overall trend of data, such as the trend of forest coverage area in a larger geographical area. Through multi-scale transformation, feature information of data at different levels can be comprehensively obtained. s (x): the approximation coefficient of the data sample x at the scale s obtained by discrete wavelet transform (DWT). It retains the low-frequency component of the data at this scale, reflecting the overall trend and main features of the data. For example, in the analysis of river description metadata, the approximation coefficient can reflect the average trend of river flow over a period of time. s(x): the detail coefficient of the data sample x obtained by DWT at scale s. It contains the high-frequency components of the data at this scale, revealing the local changes and detailed information of the data. For example, in the analysis of grassland metadata, the detail coefficient can highlight the local changes of grassland vegetation in a particular season or region, such as the sudden increase or decrease of vegetation in some areas. and are the weight vectors related to the approximation coefficient and the detail coefficient extracted from the vector field . They adjust the weights of the approximation coefficient and the detail coefficient according to the key feature extraction direction. For example, if the key feature extraction direction focuses on highlighting the long-term stability characteristics of the ecosystem, then will give higher weight to the part of the approximation coefficient that reflects the long-term trend; if the focus is on the short-term fluctuations of the ecosystem, will emphasize the information in the detail coefficient that reflects the short-term changes. The feature information set of the ecological base data element and the ecological derived data element at different scales s is composed of the weighted approximation coefficient and the weighted detail coefficient . This set comprehensively summarizes the feature representation of the data at different scales and key feature extraction directions, providing a rich variety of information for subsequent fusion analysis. G: the graph structure constructed by the graph neural network (GNN) model f f , the vertices V are the feature information at different scales (i.e. and ), and the edges E represent the relationship between the features. Through this graph structure, GNN can capture complex associations between features, such as spatial adjacency relationships (ecological region features that are adjacent in geographical space) and semantic association relationships (such as the similarity and difference between ecological tourism and leisure health care in semantics). b and K d : the base ecological key features and the derived ecological key features, respectively. They are the results obtained by GNN after fusion analysis of the feature information at different scales, representing the features in the ecological base data element and the ecological derived data element that are of key significance to the evaluation, development, and protection of ecological resources. These key features can provide core basis for the subsequent correlation analysis of green finance data elements and the construction of ecological resource credible data element models.

[0037] Preferably, in the process of establishing the correlation relationship and generating the model, the green finance data element is where p represents the dimension of the green finance data, reflecting the complexity and number of features of the green finance data. The green finance data element is calculated with the base ecological key features K b and the derived ecological key features K dThe correlation between the variables, the present application adopts a complex similarity measure function based on the extension of mutual information (Mutual Information) Mutual information is used to measure the degree of dependence between two random variables, and the extended function Consider the high-order dependence relationship between multiple variables. Let X = [K b ,K d ,D g ] be a joint variable set containing basic ecological key features, derived ecological key features and green financial data elements. By calculating the mutual information matrix M mi between each variable in the joint variable set, and after a series of matrix transformations (such as singular value decomposition, etc.), the correlation matrix M r : M mi (i,j) = I(X i ; X j ), M r = Transform(M mi ), where I(X i ; X j ) represents the mutual information between variables X i and X j , and Transform represents a series of complex matrix transformation operations, including but not limited to matrix normalization, adjustment of singular values after singular value decomposition, and recombination matrix, etc. to highlight the correlation characteristics between variables.

[0038] To calculate the causality between green financial data elements and basic ecological key features, derived ecological key features, the present application adopts a complex inference algorithm based on structural causal model (SCM) combined with Bayesian network (Bayesian Network) A Bayesian network B = (G, Θ) is constructed, where G is a directed acyclic graph, nodes represent variables (i.e. K b , K d and D g ), and edges represent the causal relationship between variables; Θ is a set of conditional probability distribution parameters. Through learning on a large amount of data, the maximum a posteriori estimation (MAP) method is used to determine the parameters Θ, and according to the inference rules of Bayesian network, the causal effect between variables is calculated, and the causal relationship graph Based on the correlation matrix M r and the causal relationship graph , an ecological resource credible data element model is constructed The method is based on knowledge graph (Knowledge Graph) fusion and deep learning. The correlation matrix and causal relationship diagram are part of the knowledge graph, and the nodes represent different data elements (green finance data elements, basic ecological key features, derived ecological key features), and the edges represent their correlation and causal relationship. Through a deep learning model based on graph convolution network (GCN) Learn and reason about the knowledge graph to generate an ecological resource credible data element model

[0039] The creative physical meaning of each parameter in the above formula is explained as follows: D g : Green finance data elements, with dimension p representing the richness and complexity of green finance data, covering various financial indicators such as green investment amount, green credit interest rate, etc. Based on the similarity measure function of mutual information expansion, by considering high-order dependence, it can more accurately capture the complex correlation between green finance data elements and ecological key features. X: The joint variable set containing basic ecological key features, derived ecological key features and green finance data elements, used to calculate mutual information. M mi : Mutual information matrix, recording the mutual information value between each variable in the joint variable set, reflecting the degree of dependence between variables. Transform: A series of matrix transformation operations to convert the mutual information matrix into a correlation matrix, highlighting the correlation characteristics between variables. M r : Correlation matrix, showing the correlation between green finance data elements and basic ecological key features, derived ecological key features, providing important association information for subsequent model construction. Based on the structural causal model combined with the causal inference algorithm of Bayesian network, the causal relationship between variables can be analyzed in depth. B: The constructed Bayesian network consists of a directed acyclic graph G and a set of conditional probability distribution parameters Θ, used to describe the causal structure and probability relationship between variables. G: The directed acyclic graph in the Bayesian network, the nodes represent variables, and the edges represent causal relationships, whose structure is determined by data learning. Θ: The set of conditional probability distribution parameters of the Bayesian network, which is learned and determined according to the data by the maximum posterior estimation method. Causal relationship diagram, which intuitively displays the causal relationship between green finance data elements and basic ecological key features, derived ecological key features, providing causal logic support for the ecological resource credible data element model. Deep learning model based on graph convolution network, used to learn and reason about the knowledge graph (including correlation matrix and causal relationship diagram) to generate an ecological resource credible data element model. The finally generated ecological resource credible data element model integrates the correlation and causality between green financial data elements and ecological key characteristics, and can provide comprehensive and credible data support for the scientific management and evaluation of ecological resources and the development of green industries.

[0040] Therefore, based on the above scheme, the embodiment at least has the following technical benefits:

[0041] 1. By constructing a multi-source data acquisition rule based on an improved autoencoder, the natural ecological basic data of different carriers can be effectively parsed and fused. Traditional methods are difficult to handle the complex structure of multi-source data, while this method uses an autoencoder to reduce and reconstruct high-dimensional tensor data, removing noise and redundant information, so that different carrier data can be processed in a unified framework. For example, when integrating multi-source data such as forests and rivers, key features can be accurately extracted to avoid data loss and improve data quality, providing a solid foundation for subsequent analysis. The use of the analytic hierarchy process to determine the weight of different carrier data can more scientifically reflect the importance of each carrier data to the overall natural ecological basic data compared to subjective judgment. In ecological resource management, different carrier data has different contributions to ecological assessment, such as forest cover data may have more influence than local data of some obscure wetlands when assessing regional ecological stability. The weight determined by AHP can reasonably reflect this difference, making the integrated data more consistent with the actual ecological significance.

[0042] 2. The structured processing model based on conditional random field and the semantic analysis method based on deep semantic embedding can effectively process natural ecological derivative description text. Traditional text processing methods are difficult to deeply mine the semantic relationship and potential structure in the text, while CRF can use the relationship between text words for structured labeling, and the DSE model minimizes the contrast loss function to make similar semantic texts closer in high-dimensional space, accurately extracting natural ecological derivative feature elements. When analyzing ecological tourism and leisure and health-related texts, key information such as tourist experience and service quality can be accurately extracted to provide high-quality features for the generation of ecological derivative data. The use of conditional generative adversarial networks to generate natural ecological derivative data can generate more diverse and realistic scenario data compared to traditional generation methods. When generating ecological tourism visitor flow prediction data, it can combine tourism resource characteristics, seasonal factors, and other conditions to generate more realistic and reliable data, providing strong support for ecological tourism planning and management.

[0043] 3. A hybrid model based on variational autoencoder-reinforcement learning can perform feature mining under the constraints of complex business rules and feature mining. Traditional feature mining methods are difficult to balance constraints and feature optimization, while the model adjusts the feature representation with the constraints as the guide through the exploration of the reinforcement learning agent in the latent space. In ecological resource data mining, such as in the mining of ecological basic data elements, it is ensured that the data features meet the logical relationship and statistical constraints of ecological indicators, while optimizing the features to better reflect the characteristics of the ecological system. Variational autoencoder maps data to latent space, providing a new perspective for feature mining. The agent's operation in the latent space can discover feature combinations and change patterns that traditional methods cannot detect. For example, in exploring ecological derivative data elements, it can mine the potential collaborative development features between ecological tourism and leisure and health care, providing a basis for the innovative development of ecological industries.

[0044] 4. A combination method based on multi-resolution analysis and discrete wavelet transform, combined with key feature extraction direction vector field, can comprehensively obtain data features from different scales. Traditional feature extraction methods often only focus on single scale or fixed direction features, while this method can dynamically adjust the feature extraction direction according to the data points and capture the overall trend and local details of the data at different scales. In analyzing forest ecological data, it can grasp the overall trend of forest coverage area changes and focus on the details of tree species distribution in specific areas, providing more comprehensive information for ecological resource assessment. Using graph neural networks to analyze and integrate multi-scale feature information can effectively capture the complex relationships between features. Traditional methods are difficult to handle high-dimensional and complex feature relationships, while GNN can deeply mine the spatial adjacency and semantic association between features by constructing a graph structure, taking different scale features as nodes and feature relationships as edges. In analyzing the correlation between ecological resources and green finance data, it can discover the potential influence path between key ecological features and green finance indicators, providing decision-making basis for green finance supporting ecological resource development.

[0045] 5. Based on a similarity measurement function and structural causal model combined with Bayesian network inference algorithms, this method can accurately analyze the correlation and causality between green finance data elements and key ecological features. Traditional analysis methods struggle to accurately measure higher-order dependencies and causal relationships among multiple variables. This method determines correlations by calculating the mutual information matrix and performing complex transformations, and uses Bayesian network learning and inference to determine causal relationships. When studying the role of green finance in ecological resource protection, it can clarify the causal link between green investment and improved ecosystem stability, as well as the correlation between green credit interest rates and ecotourism development, providing a scientific basis for policy formulation. A knowledge graph-integrated deep learning method generates a credible data element model for ecological resources, integrating correlations and causal relationships to produce a comprehensive and reliable data model. Traditional models struggle to comprehensively consider multiple complex relationships, while this method integrates the correlation matrix and causal relationship graph into the knowledge graph, using graph convolutional networks for learning and inference, enabling the model to reflect the inherent logic of ecological resource data. In ecological resource management, this model can provide unified and accurate data support for ecological resource assessment, development, and protection, improving the scientific rigor and effectiveness of ecological resource management.

[0046] Optionally, the acquisition of natural ecological derived data includes:

[0047] Structured text is obtained by structuring descriptive text derived from natural ecology.

[0048] Semantic analysis is performed on the structured text to extract natural ecology-derived feature elements;

[0049] Data derived from natural ecology features.

[0050] Preferably, in one embodiment, in the structured processing step of the natural ecology derived descriptive text, the natural ecology derived descriptive text is T = [t1, t2, ..., t N ], where t j Let j represent the j-th word unit in the text, and N be the text length. The specific steps include the following:

[0051] (1) Lexical embedding and positional encoding steps

[0052] To convert the text into a vector representation, first, for each word t... j Embedding operations are performed. A Transformer-based relative positional encoding embedding method is used, with the vocabulary embedding matrix set as follows: Where d vocab It is the size of the vocabulary, d emb It's an embedded dimension. (Vocabulary t) j Embedded vector e j =V[t j ], here V[tj ] indicates taking the word t from the word embedding matrix V. j The corresponding row vector. Simultaneously, to capture the positional information of words in the text, a positional encoding vector p is introduced. j Position encoding is generated using a sine and cosine function: Where j is the position index of the word, k is the dimension index of the embedding vector, and p j,k It is the position encoding vector p j The k-th element. The final vocabulary feature vector x j =e j +p j .

[0053] (2) Structured processing model

[0054] Preferably, in one embodiment, a model based on an improved Graph Convolutional Recurrent Network (GCRN) is used for structured processing. A text graph G = (V, E) is constructed, where vertices V = {v1, v2, ..., v...} N The edges E represent the relationships between words in the corresponding text. Here, the weights of the edges are determined by calculating the semantic similarity between words, using an improved method based on cosine similarity. Where, x i and x j It is the word t i and t j The feature vector is denoted by |ij|, where |ij| is the positional distance of the word in the text, and σ is a parameter that controls distance decay. When s(t i ,t j When the value is greater than a certain threshold τ, at vertex v i and v j Add an edge between them, with the weight s(t) i ,t j ).

[0055] In the GCRN model, each vertex v j The hidden state h j The updated formula is as follows: Where l represents the number of GCRN layers, and GRU stands for Gated Recurrent Unit, used to process sequence information. After L layers of GCRN processing, the final hidden state of each word is obtained. These hidden states constitute the representation of structured text.

[0056] Preferably, in one embodiment, the semantic analysis of structured text includes the following steps:

[0057] (1) Semantic analysis model construction

[0058] The structured text S is subjected to semantic analysis by using a model based on a deep semantic fusion network (DSFN). The DSFN model is composed of multiple semantic fusion modules, each of which contains a self-attention mechanism and a multi-layer perceptron (MLP). In the self-attention mechanism, for the input structured text S, a query vector Q = W Q S, a key vector K = W K S, and a value vector V = W V S are calculated, where is a learnable weight matrix, d att is the dimension of the attention mechanism. The self-attention output A is calculated as follows: Then, the self-attention output A is input into the multi-layer perceptron for further feature extraction. Let the input of the multi-layer perceptron be A, which is processed through two fully connected layers and an activation function (such as ReLU): F = MLP(A) = W2ReLU(W1A + b1) + b2, where W1, W2 are the weight matrices of the fully connected layers, and b1, b2 are the bias vectors. After processing by M semantic fusion modules, the natural ecological derivative feature element F M is obtained.

[0059] Preferably, in an embodiment, in the step of generating natural ecological derivative data based on the natural ecological derivative feature element, a generative model based on the fusion of a generative adversarial network (GAN) and a variational autoencoder (VAE) is used to generate the natural ecological derivative data. Let the natural ecological derivative feature element be F M , which is concatenated with a random noise vector to obtain an input vector I = [F M ; z], where is a normal distribution with a mean of 0 and a covariance matrix of the identity matrix I, and specifically includes the following technical process:

[0060] (1) Generator construction

[0061] The generator G is a multi-layer neural network, the input of which is I, and the output of which is the generated natural ecological derivative data D g . Let the weight matrices of the layers of the generator be and the bias vectors be Then: where σ is an activation function (such as tanh).

[0062] (2) Construction of discriminator

[0063] The discriminator D is also a multi-layer neural network, the input of which is the real natural ecological derivative data Dr or generated data D g , the output is a scalar representing the probability that the data is real data. Let the weight matrix of each layer of the discriminator be The bias vector is Then: During training, the model is optimized by minimizing the adversarial loss of the generator and the discriminator. The goal of the generator is to maximize the probability that the discriminator will judge the generated data as real data, and the goal of the discriminator is to maximize the probability that it will judge real data as real data and generated data as fake data. The specific loss function is as follows:

[0064] By alternating optimization of the generator and the discriminator, the generated natural ecological derivative data D g is increasingly close to the distribution of real data.

[0065] In the above formula, the specific physical meaning of each parameter is as follows: T: natural ecological derivative description text sequence, is the original text data, containing description information about ecological tourism, leisure and health care, etc. t j : the jth word unit in the text, is the basic element that constitutes the text. N: text length, reflects the size of the text, affects the complexity of subsequent processing. V: word embedding matrix, maps each word in the vocabulary to a d emb dimensional vector space, the dimension d vocab ×d emb determines the richness of word representation and computational complexity. vocab : vocabulary size, i.e. the number of different words in the text, which affects the number of rows in the word embedding matrix. d emb : embedding dimension, determines the length of each word vector, a larger embedding dimension can capture richer semantic information of words. p j : position encoding vector, used to add position information to words, so that the model can distinguish the same words in different positions, its dimension is the same as the word embedding vector dimension, i.e. d emb . G: text graph, used to describe the relationship between words in the text, by constructing a graph structure, the model can use the semantic association between words for structured processing. V: vertex set of the text graph, corresponding to the words in the text. E: edge set of the text graph, the weight of the edge is determined by the semantic similarity between words, reflecting the closeness between words. s(t i ,t j ): words t i and t jThe semantic similarity between words is calculated by considering both the vector representation and the position distance. The parameter σ controls the distance decay and determines the influence of the position distance on the semantic similarity. A larger σ makes the position distance have less influence on the semantic similarity. The threshold τ is used to add edges in the text graph when the semantic similarity between words is greater than the threshold. The value of τ affects the sparsity and structure of the text graph. GRU is a Gated Recurrent Unit used to process text sequence information and update the hidden state of the vertex in GCRN. The internal parameters (such as weight matrix and bias vector) are learned through training. The number of layers l determines the depth of the model in extracting text structure information. Increasing the number of layers can learn more complex text structure features. L is the total number of layers in GCRN. After L layers of processing, the final representation of the structured text is obtained.

[0066] The physical meaning of each parameter in the semantic analysis step is as follows: S: The representation of the structured text, which is the sequence of word hidden states obtained after GCRN processing, contains the structure and partial semantic information of the text. Q ,W K ,W V The weight matrix in the self-attention mechanism is used to map the structured text S to the query, key, and value vector space, with dimension d att ×d emb determines the calculation and representation ability of the attention mechanism. d att The dimension of the attention mechanism affects the complexity of the self-attention calculation and the feature representation ability. A larger dimension can capture more rich semantic relationships, but also increases the computational load. A: The output of the self-attention, which is obtained by calculating the query, key, and value vectors, reflects the attention distribution between words in the text. W1, W2: Weight matrices in the multi-layer perceptron, used for further feature extraction of the self-attention output, with dimensions determined by the input and output feature quantities. b1, b2: Bias vectors in the multi-layer perceptron, used to adjust the output of the neural network. ReLU: Activation function, used to introduce nonlinearity, allowing the model to learn more complex semantic features. M: The number of semantic fusion modules, determines the depth and complexity of the model in extracting semantic features, increasing the number of modules can learn more advanced semantic representations, but may also lead to increased training time and overfitting. M The natural ecological derived feature vector obtained after M semantic fusion modules is the final result of the semantic analysis of the structured text, containing key semantic features related to ecological tourism, leisure and health care, etc.

[0067] In the data generation step, the physical meaning of each parameter is as follows: F M: Natural ecological derived feature elements, as part of the input of the generation model, provide semantic constraints for the generated data, making it conform to the characteristics of natural ecological derived data. z: Random noise vector, subject to normal distribution For introducing randomness for generated data, so that the generated data has diversity. I: The input vector of the generator, which is spliced by natural ecological derived feature elements and random noise vectors, has a dimension equal to the sum of the dimensions of natural ecological derived feature elements and noise vectors. G: The generator, which is a multi-layer neural network, is used to generate natural ecological derived data according to the input vector, and its internal parameters (such as weight matrix and bias vector) are learned through training. The weight matrix of each layer of the generator determines the transformation of the input vector by the generator. The bias vector of each layer of the generator is used to adjust the output of each layer of the generator. P: The number of layers of the generator determines the processing complexity of the input vector by the generator and the quality of the generated data. σ: Activation function (such as tanh), used to introduce nonlinearity, so that the generator can generate more abundant data distribution. g : The generated natural ecological derived data is the output result of the generator. D: The discriminator, which is a multi-layer neural network, is used to judge whether the input data is real data or generated data, and its internal parameters (such as weight matrix and bias vector) are learned through training. The weight matrix of each layer of the discriminator determines the feature extraction and judgment method of the input data by the discriminator. The bias vector of each layer of the discriminator is used to adjust the output of each layer of the discriminator. Q: The number of layers of the discriminator determines the discrimination ability of the data by the discriminator, and increasing the number of layers can improve the discrimination accuracy. p: The probability value output by the discriminator, indicating the probability of the input data being real data. G : The loss function of the generator, used to measure the difference between the data generated by the generator and the real data distribution.

[0068] To this end, based on the above scheme, in an embodiment, the following technical benefits are provided:

[0069] 1. The relative position encoding embedding method of the Transformer is used. The vocabulary embedding matrix maps the vocabulary to a high-dimensional vector space, giving each vocabulary a rich semantic representation. Position encoding cleverly incorporates the position information of the vocabulary in the text into the vector representation through unique sine and cosine functions. In ecotourism texts, the two words "scenic area entrance" and "tourist center" are not only distinguished by semantic embedding, but also reflect their order in describing the tour route through position encoding, allowing the model to accurately capture the spatial and semantic structure of the text and provide a solid foundation for subsequent analysis. The semantic similarity between words is calculated using an improved cosine similarity formula, taking into account both the word vector representation and the position distance to assign edge weights to the text graph. In leisure and health care texts, "forest bathing" and "meditation activities" have similar semantics and may be adjacent in descriptions, so a higher semantic similarity promotes stronger connections between the corresponding vertices in the text graph. This text graph constructed based on semantic association uses the Graph Convolutional Recurrent Network (GCRN) to update the vertex hidden state, effectively mining the context relationships and semantic dependency structures of the words in the text. Compared to traditional sequence models, it can better handle complex relationships between words in long texts and improve the accuracy of structured processing.

[0070] 2. The self-attention mechanism in the Deep Semantic Fusion Network (DSFN) generates query, key, and value vectors through a learnable weight matrix to calculate the attention distribution between words in the text. When analyzing ecotourism and leisure and health care-related texts, the self-attention mechanism allows the model to focus on the semantic relationships between "hiking" and "exploring" and related modifiers in "hiking exploration in ecotourism projects," highlighting key semantic information and avoiding local information loss. This improves semantic analysis accuracy and provides a comprehensive and in-depth understanding of text semantics. The self-attention output is further processed by a multi-layer perceptron, which performs nonlinear transformation and feature extraction on the word semantics through two fully connected layers and a ReLU activation function. In the ecological derivative data scenario, it can extract high-level semantic features such as "the positive impact of ecotourism on physical and mental health" from the original word semantics, integrate related semantic information scattered throughout the text, and provide rich and accurate semantic information for generating natural ecological derivative features. This helps to uncover the internal relationships between ecological resources and tourism and health care industries.

[0071] 3. A generative model based on the fusion of Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs) uses concatenated natural ecological derivative features and random noise vectors as input to the generator. When generating ecotourism visitor flow prediction data, natural ecological derivative features (such as semantic features of scenic area ecological resources and seasonal factors) provide semantic constraints to ensure the data conforms to the actual ecotourism scenario. The random noise vector introduces randomness, making the generated visitor flow data diverse within a reasonable range, meeting the prediction needs of different scenarios. Compared with traditional data generation methods, it can generate more realistic and diverse data. The generator and discriminator are alternately optimized through adversarial loss. The generator strives to generate natural ecological derivative data that the discriminator might mistake for real, while the discriminator continuously improves its ability to distinguish between real and fake data. When generating evaluation data for leisure and wellness services, as training progresses, the data generated by the generator gradually approaches the distribution of real evaluation data, and the data quality continuously improves. This provides reliable data support for ecological resource development and service improvement, effectively solving the problem of insufficient data authenticity and diversity in traditional generative models.

[0072] Optionally, the controlled feature mining of basic natural ecological data and derived natural ecological data to generate basic ecological data elements and derived ecological data elements includes:

[0073] Based on the established business rules and feature mining constraints, a controlled feature mining model is constructed.

[0074] Based on the controlled feature mining model, controlled feature mining is performed on basic natural ecological data and derived natural ecological data to generate basic ecological data elements and derived ecological data elements.

[0075] Preferably, in one embodiment, constructing the controlled feature mining model includes the following technical processing steps:

[0076] (1) Formalization of business rules and constraints

[0077] Let the set of business rules be Where R i It is a Boolean function R i : n represents the dimension of the data features. In the context of ecological resource management, if R1 represents "the forest area ratio should be between 0 and 1", then for the input natural ecological data feature vector x = (x1, x2, ..., x... n When the proportion of forest area in x meets the range requirement, R1(x) = 1; otherwise, R1(x) = 0.

[0078] Feature mining constraints are represented by constraint function Φ(x), which is a real-valued function Φ: For example, in eco-tourism data mining, the constraint condition can be "the difference between the number of tourists in the tourism peak season and the carrying capacity of tourism facilities should be within a reasonable range", at this time Φ(x) can be defined as the absolute value of the difference, and whether the constraint is met is judged by setting a threshold.

[0079] (2) Construction of a variational autoencoder-reinforcement learning hybrid model

[0080] Variational autoencoder (VAE): input natural ecological data The encoder maps it to a latent space. Let the encoder network be a multilayer perceptron (MLP), then the mean μ(x) and log variance logσ 2 (x) of the latent space are calculated as follows: μ(x) = MLP μ (x; θ μ ), where MLP μ and are multilayer perceptrons with parameters θ μ and .

[0081] Sample z from the latent space: z = μ(x) + σ(x) ⊙ ∈, where ⊙ denotes element-wise multiplication.

[0082] The decoder is also a multilayer perceptron that reconstructs z into

[0083] The loss function L VAE of the VAE is: where is the posterior distribution, is the prior distribution, and p(x|z) is the conditional probability of the decoder generating x. It can be further expanded as: Here k is the dimension of the latent space.

[0084] (3) Reinforcement learning (RL)

[0085] The state s of the reinforcement learning agent is defined as s = [z, Φ(z)], and the action is the adjustment to the latent variable z.

[0086] The reward function R(s, a) is: R(s, a) = αR rec (s, a) + βR con (s, a) + γR rule (s, a)

[0087] where: is the reconstruction reward, which encourages the agent to generate data that can be well reconstructed. is the constraint reward, δ is the constraint threshold. is the business rule reward, λ i is the rule weight.

[0088] The agent selects actions using a policy network π(a|s; θ π ) and updates the policy network parameters θ π using the proximal policy optimization (PPO) algorithm. π The PPO algorithm updates θ π by maximizing the following objective function: where is the importance sampling ratio, is the advantage estimate, and ∈ is the clipping parameter.

[0089] Preferably, in a specific embodiment, feature mining based on the controlled feature mining model includes the following technical steps:

[0090] (1) Data input and preprocessing

[0091] The natural ecological basic data and the natural ecological derived data are normalized to obtain and The normalization function is

[0092] (2) Feature mining process

[0093] For each normalized data sample (from basic data or derived data):

[0094] 1. Obtain the latent variable z through the VAE encoder. 2. The agent selects actions a according to the state s = [z, Φ(z)] using the policy network π(a|s; θ π ) 3. Update the latent variable z' = z + a. 4. Check the constraint condition Φ(z') and the business rule R i (z'), if not satisfied, the agent reselects actions until it is satisfied. 5. Reconstruct z' to 6. Denormalize the reconstructed data to obtain the final feature mining result

[0095] Preferably, in a specific embodiment, the generation of ecological basic data elements and ecological derived data elements is performed through the above feature mining process to obtain ecological basic data elements and ecological derived data elements

[0096] The physical meaning of each parameter in the above process is as follows: the business rule set A set of Boolean functions are included to define the business requirements that the data needs to satisfy in ecological resource management, such as data range, logical relationship, etc. Constraint function Φ(x): measures the degree of deviation of data feature vector x from the constraint condition, and ensures the rationality of data in ecological data mining. μ , and are the parameters of the encoder and decoder multi-layer perceptron, which determine the mapping and reconstruction of data in the latent space. VAE : balance data reconstruction error and latent space distribution, so that the model learns an effective representation of the data. α, β and γ: weight coefficients of reconstruction reward, constraint reward and business rule reward in the reward function, adjusting them can control the agent behavior. π : policy network parameters, optimized by PPO algorithm to maximize cumulative reward. ∈: clipping parameter of PPO algorithm, to prevent the policy from updating too much. base and X derived : are the sample sets of natural ecological basic data and derived data respectively. N and M: the number of samples of basic data and derived data respectively. δ: constraint threshold, used to judge whether the constraint condition is met. i : the weight of business rule R i , reflecting the importance of different rules.

[0097] Preferably, in an embodiment, the above-mentioned controlled feature mining on natural ecological basic data and natural ecological derived data respectively to generate ecological basic data elements and ecological derived data elements has the following technical advantages:

[0098] 1. Traditional methods usually lack systematic integration of business rules and constraint conditions. In the context of ecological resource management, it may only be manually screened after data collection, or use simple conditional statements to preliminarily filter the data. For example, when processing forest area ratio data, it may only be manually checked during data entry, and it is impossible to ensure that the data meets the rules in real time and comprehensively during feature mining. For complex constraint conditions, such as the dynamic relationship between the number of tourists and the carrying capacity of facilities in the context of ecological tourism, traditional methods are difficult to effectively cope with, and often cannot integrate such constraints into the core process of data mining.

[0099] The present scheme formalizes business rules into a set of Boolean functions and the feature mining constraint condition is represented as a real-valued function Φ(x), which can accurately define and quantify various rules and constraints. Throughout the feature mining process, these rules and constraints are strictly followed from the data input. For example, in the ecological basic data element mining, for business rules such as forest area proportion and river flow range, the model can judge in real time whether the data feature vector meets the conditions. In ecological tourism data mining, the constraint function Φ(x) can continuously monitor the difference between the number of tourists and the carrying capacity of facilities, and dynamically adjust according to the set threshold, ensuring that the mined data elements fully meet the actual business requirements and constraint conditions, greatly improving the accuracy and reliability of the data.

[0100] 2. Traditional data feature extraction and dimension reduction methods, such as principal component analysis (PCA), can only extract linear features and cannot well handle the probability distribution and generation of data. In ecological resource data, data often has complex nonlinear features, and traditional methods are difficult to capture the internal relationship between these features. For example, when analyzing forest ecosystem data in natural ecological basic data, there is a complex nonlinear relationship between forest species diversity, vegetation coverage, and soil quality. PCA cannot effectively extract this information. At the same time, traditional methods have limited ability to generate new data samples, making it difficult to meet the needs of data expansion and prediction in ecological resource management.

[0101] The VAE in this scheme uses a multi-layer perceptron as the encoder and decoder, which can learn the nonlinear feature representation of the data. By mapping the data to the latent space, VAE not only extracts the key features of the data, but also generates new data samples by manipulating the latent space. For example, when processing natural ecological derived data such as ecological tourism visitor flow prediction, VAE can learn the distribution rules of the latent space based on historical data, and then generate visitor flow prediction data in different scenarios by sampling in the latent space. The loss function L VAE balances the reconstruction error and latent space distribution, enabling the model to learn an effective representation of the data, providing a solid foundation for subsequent feature mining and generation tasks.

[0102] 3. Traditional feature mining methods are usually based on fixed algorithms and parameters, lacking the ability to adapt to environmental changes and constraint conditions. In the context of ecological resource management, business rules and constraint conditions may change over time, season, policy, etc. Traditional methods cannot adjust the mining strategy in a timely manner. For example, in the peak and off-season of ecological tourism, the constraint conditions of the number of tourists and the carrying capacity of facilities may be different, and traditional methods are difficult to dynamically adjust the feature mining process according to such changes. In addition, traditional methods often cannot fully utilize the feedback information between data, resulting in the mined features may not be optimal.

[0103] The reinforcement learning agent in this solution can perceive the latent variable z and the satisfaction of the constraint condition in real time by defining the state as s = [z, Φ(z)]. The reward function R(s, a) considers the reconstruction reward, constraint reward, and business rule reward, enabling the agent to generate not only well-reconstructed data but also data that meets various constraint conditions and business rules when exploring the latent space. For example, when mining ecological basic data elements, the agent can adjust the action a according to the current latent variable and constraint condition to generate data features that meet business rules such as forest area proportion and river flow range. By updating the policy network parameters θ π using the Proximal Policy Optimization (PPO) algorithm, the agent can continuously learn and optimize the policy to improve feature mining capability in complex environments, thereby generating more valuable ecological basic and derived data elements.

[0104] 4. Traditional feature mining processes are often single and linear, lacking effective interaction and feedback between each link. In ecological resource management scenarios, different types of data (such as natural ecological basic data and natural ecological derived data) may require different processing methods, but traditional methods are difficult to comprehensively process these data. For example, when processing ecological tourism data, traditional methods may only focus on single data such as tourist flow, ignoring the association with natural ecological basic data (such as scenic ecological environment quality). In addition, traditional methods are inefficient when processing large-scale data, making it difficult to meet real-time requirements. The controlled feature mining model constructed in this solution combines data input and preprocessing, feature mining process, and ecological data element generation. By uniformly processing natural ecological basic data and natural ecological derived data, it can mine the potential association between different types of data. For example, when generating ecological basic data elements and ecological derived data elements, the model can consider forest ecosystem data and ecological tourism data simultaneously, and mine key information such as the impact of ecological tourism on forest ecosystems. At the same time, through the synergistic effect of VAE and RL, the model can efficiently perform feature mining while meeting business rules and constraint conditions, improving data processing efficiency and accuracy, and providing more comprehensive and valuable data support for ecological resource management.

[0105] Optionally, the key features of the ecological basic data elements and the ecological derived data elements are extracted respectively to obtain basic ecological key features and derived ecological key features, including:

[0106] Based on the set key feature extraction direction, multi-scale transformation is performed on the ecological basic data elements and the ecological derived data elements to obtain feature information of the ecological basic data elements and the ecological derived data elements at different scales;

[0107] The characteristic information of the ecological base data elements and the ecological derived data elements at different scales is fused and analyzed to obtain the base ecological key characteristics and the derived ecological key characteristics.

[0108] Preferably, in an embodiment, the set of ecological base data elements is The set of ecological derived data elements is where N and M are the number of ecological base data elements and ecological derived data elements, respectively. The multi-scale transformation based on the set key characteristic extraction direction includes the following steps:

[0109] (1) Definition of the key characteristic extraction direction

[0110] The set key characteristic extraction direction is a high-dimensional vector field where x can be or The dimension d of the vector field is d θ , and its value changes with the data point x to adapt to the characteristic extraction requirements of different data regions. In the ecological resource management scenario, for ecological base data elements such as forest area, river flow, etc., the direction can be adjusted according to the functional zoning of the ecological system (such as core protection area, buffer zone, etc.), highlighting the key characteristics related to ecological protection; for ecological derived data elements such as ecological tourism visitor flow, leisure and health care service satisfaction, etc., the direction can be adjusted according to market demand and industry development trend, focusing on key factors affecting the sustainable development of ecological industry.

[0111] (2) Multi-scale transformation

[0112] A combined method based on non-subsampled contourlet transform (NSCT) and fractional Fourier transform (FrFT) is used for multi-scale transformation. Let be the scale parameter, for ecological base data elements the approximate coefficients and the detail coefficients are obtained by NSCT at scale s. Therefore, since NSCT is a multi-scale, multi-directional image decomposition method, it can effectively capture the edge and texture information of the data.

[0113] In order to combine the key characteristic extraction direction, the NSCT coefficients are weighted to obtain the weighted approximate coefficients and the weighted detail coefficients where and are the vector fields The weight vectors related to the approximation coefficients and the detail coefficients are extracted. These weight vectors are used to adjust the approximation coefficients and the detail coefficients according to the key feature extraction direction, so as to highlight the information related to the key feature.

[0114] Meanwhile, the fractional Fourier transform is performed on the , which is defined as:

[0115] wherein, a is the order of the fractional Fourier transform, K α (t, u) is a kernel function, which is defined as:

[0116] Here, n is an integer, and δ is the Dirac function. The fractional Fourier transform can analyze the data in the time-frequency domain, and by adjusting the order a, the characteristic information of different time-frequency characteristics can be obtained.

[0117] After the fractional Fourier transform, the coefficients under the scale s are obtained. wherein is the weight vector related to the fractional Fourier transform coefficients extracted from the vector field .

[0118] For the ecological derivative data element , the above NSCT and fractional Fourier transform and their weighting processing are also performed to obtain

[0119] After the weighting processing, the characteristic information of the ecological basic data element under different scales s is obtained The characteristic information of the ecological derivative data element under different scales s is obtained

[0120] Preferably, in an embodiment, in the fusion analysis link of the characteristic information under different scales, the following steps are included:

[0121] (1) Construct a fusion analysis model

[0122] A fusion analysis model based on quantum neural network (QNN) and graph attention network (GAT) is adopted. For the characteristic information of the ecological basic data element under different scales , a graph structure G base = (V base , E base ) is constructed, wherein the vertex V base is the characteristic information under different scales (i.e. ), and the edge E baserepresents the relationship between features. Here the weight of the edge is determined by calculating the quantum correlation between features, which is calculated by a method based on quantum mutual information. Let the quantum states of two features x and y be ρ x and ρ y , and their quantum mutual information I Q (x;y) is defined as: I Q (x;y) = S(ρ x )+S(ρ y )-S(ρ xy ), where S(ρ) is the von Neumann entropy of quantum state ρ, defined as S(ρ) = -tr(ρlog2ρ), ρ xy is the joint quantum state of x and y. When I Q (x;y) is greater than a certain threshold τ base , an edge is added between vertices x and y, and the weight of the edge is I Q (x;y).

[0123] The graph structure G base is input into the graph attention network, which calculates the importance weight of each vertex through the attention mechanism. For a vertex v∈V base , its attention weight α v is calculated as follows: where, is a learnable attention vector, W is a weight matrix, and are the feature vectors of vertex v and its adjacent vertex u, is the set of adjacent vertices of vertex v, || represents vector splicing, and LeakyReLU is a leaky rectified linear unit. Then, the output of the graph attention network is input into the quantum neural network. The quantum neural network is composed of quantum neurons, and the state update of the quantum neurons adopts quantum gate operation. Let the initial state of the quantum neuron be |ψ0>, and after a series of quantum gate operations U1, U2, …, U K , the final state |ψ K > = U K U K-1 …U1|ψ0> is obtained. The quantum gate U k may be a rotation gate, a phase gate, etc., and its parameters are learned through training. After the fusion analysis of the quantum neural network and the graph attention network, the basic ecological key features K base are obtained. For the feature information of the ecological derivative data element at different scales , a graph structure G derived =(V derived ,E derivedThe weights of the edges are calculated (based on quantum mutual information). After fusion analysis using graph attention networks and quantum neural networks, the key features K of the derived ecosystem are obtained. derived .

[0124] The creative physical meaning of each parameter in the above technical processing is explained as follows: E base : A collection of ecological foundation data elements, containing data elements related to the natural ecological foundation obtained through controlled feature mining, such as quantitative representations of features like forest area and river flow. E derived : Ecological derived data elements set, including data elements related to derivative fields such as ecotourism, leisure and wellness, etc., such as quantitative representations of data like ecotourism visitor flow and leisure and wellness service satisfaction. N: The number of ecological basic data elements, reflecting the richness and diversity of ecological basic data. M: The number of ecological derived data elements, reflecting the scale and complexity of ecological derived data. A high-dimensional vector field with dimension d for key feature extraction directions. θ This determines the complexity and flexibility of feature extraction. In ecological resource management scenarios, this vector field can dynamically adjust the focus and direction of feature extraction according to different data types and application requirements. They are from vector fields The weight vectors extracted are related to the NSCT approximation coefficients, detail coefficients, and fractional Fourier transform coefficients. These weights are applied to the coefficients of different transforms based on the key feature extraction direction, highlighting information relevant to the key features. The scale parameter set, where 's' is the scale parameter, represents different perspectives on the data at different resolutions. Smaller scales focus on local details of the data, while larger scales emphasize the overall trend of the data. Ecological basic data elements The approximation coefficients and detail coefficients obtained by NSCT at scale s reflect the low-frequency and high-frequency feature information of the data at that scale. The weighted approximation coefficients and detail coefficients are weighted by a weight vector related to the direction of key feature extraction to highlight key features. Ecological basic data elements The fractional Fourier transform results can be used to obtain characteristic information of different time-frequency characteristics by adjusting the order α of the fractional Fourier transform. Ecological basic data elements The coefficients are obtained at scale s after fractional Fourier transform. The weighted fractional Fourier transform coefficients are weighted using a weight vector related to the key feature extraction direction to highlight key features. For parameters related to ecologically derived data elements (such as...) ), with similar meaning as the ecological base data elements, but corresponding transformations and weighting are applied to the ecological derived data. G base , G derived : Graph structures of feature information construction at different scales for ecological base data elements and ecological derived data elements respectively, describing the associations between features through the relationship of vertices and edges. base , V derived : Vertex set of graph structures G base and G derived , containing feature information at different scales. base , E derived : Edge set of graph structures G base and G derived , the weight of the edge is calculated by quantum mutual information, reflecting the correlation between features. Q (x; y): Quantum mutual information, used to measure the quantum correlation between two features x and y, calculated by von Neumann entropy. S(ρ): Von Neumann entropy of quantum state ρ, used to calculate quantum mutual information. base , τ derived : Threshold value, used to determine the addition of edges in the graph structure, when the quantum mutual information of two features is greater than the threshold value, the edge is added. Learnable attention vector in graph attention network, used to calculate the attention weight of vertex. W: Weight matrix in graph attention network, used for linear transformation of vertex feature vector. α v : Attention weight of vertex v in graph attention network, reflecting the importance of the vertex in feature fusion. |ψ0>: Initial state of quantum neuron in quantum neural network. U1, U2, …, U K : Quantum gate in quantum neural network, through the operation of quantum neuron state, realizes the processing of information and the extraction of features. K base : Basic ecological key features obtained through fusion analysis, containing feature information in ecological base data elements that is of key significance to ecological resource assessment, management, etc. K derived : Derived ecological key features obtained through fusion analysis, containing feature information in ecological derived data elements that is of key significance to ecological industry development, ecological service assessment, etc.

[0125] Preferably, in an embodiment, the above-mentioned key feature extraction of ecological base data elements and ecological derived data elements respectively, to obtain basic ecological key features and derived ecological key features, has the following technical benefits:

[0126] 1. Traditional key feature extraction methods usually use fixed algorithms and preset feature directions, lacking adaptability to different data regions and application scenarios. For example, when analyzing ecological basic data, a unified feature extraction direction based on spatial distribution may be used, but the data differences between different functional divisions (such as core protection areas and buffer zones) cannot be addressed specifically, making it difficult to highlight key features related to ecological protection. In the processing of ecological derivative data, such as ecological tourism data, the focus of feature extraction cannot be dynamically adjusted according to market demand and industry development trends.

[0127] This scheme sets up a high-dimensional vector field As the key feature extraction direction, its dimension d θ flexibility and the characteristics of changing with data points x can accurately adapt to the feature extraction needs of different data regions. In the processing of ecological basic data, the distribution characteristics of rare species can be highlighted in core protection areas, and the characteristics of ecological transition zones can be emphasized in buffer zones. In terms of ecological derivative data, the correlation between tourist flow and service quality can be focused on during the tourist season, and the cost control and resource maintenance characteristics can be focused on during the off-season, effectively focusing on key factors affecting the sustainable development of ecological industries.

[0128] 2. Traditional multi-scale transformation techniques rely on a single transformation method, such as simple wavelet transform, which cannot fully combine the advantages of different transformations. In the processing of ecological data, it is difficult to consider the spatial information such as edges and textures of data, as well as time-frequency domain information. For complex ecological system structures in ecological basic data, such as multi-level structures of forest vegetation, a single transformation cannot fully capture features at different scales; in ecological derivative data, such as the time distribution and spatial layout characteristics of ecological tourism activities, traditional methods cannot effectively analyze them in both time-frequency domain and spatial domain.

[0129] This scheme uses a combination of non-subsampled contourlet transform (NSCT) and fractional Fourier transform (FrFT). NSCT can effectively capture edge and texture information of data, providing support for spatial feature analysis of ecological basic data, such as clearly showing river boundaries and forest patch edges. FrFT analyzes data in the time-frequency domain, revealing the variation rules of ecological data in time and frequency, such as the seasonal and periodic changes of ecological tourism tourist flow. At the same time, through weighted processing, different transformation coefficients are adjusted according to the key feature extraction direction, enhancing the highlighting effect of key features. For example, when analyzing forest ecological basic data, the weight of the corresponding coefficient is increased for areas related to key ecological protection features, making the key features more prominent at different scales.

[0130] 3. Traditional feature fusion analysis is mostly based on simple linear combination or correlation analysis, which cannot effectively capture the complex nonlinear relationship and quantum level correlation between features. In ecological data fusion, it is difficult to build an accurate feature relationship model. For example, when analyzing the correlation between ecological basic data elements and ecological derived data elements, traditional methods may only consider the surface linear correlation, ignoring the complex interactions within the ecological system, such as the nonlinear impact of ecological tourism activities on the ecological environment.

[0131] The present scheme uses a fusion analysis model based on quantum neural network (QNN) and graph attention network (GAT). The quantum correlation between features is calculated by quantum mutual information to determine the edge weight of the graph structure, which can reveal the deep quantum level correlation between features, which traditional correlation analysis cannot do. Graph attention network uses attention mechanism to calculate vertex importance weight, which can highlight the role of key features in fusion, such as accurately identifying features that play a key role in ecological resource assessment in ecological basic data feature fusion. Quantum neural network processes feature information through quantum gate operation, and its quantum characteristics give the model stronger information processing and feature extraction capabilities, which can mine feature patterns that traditional neural networks cannot find, providing more comprehensive and in-depth key feature information for ecological resource management and ecological industry development.

[0132] Optionally, the green finance data elements are obtained, and the correlation between the green finance data elements and the basic ecological key features and the derived ecological key features is established to generate an ecological resource credible data element model, comprising:

[0133] The correlation between the green finance data elements and the basic ecological key features and the derived ecological key features is calculated to construct a correlation matrix;

[0134] The causality between the green finance data elements and the basic ecological key features and the derived ecological key features is calculated to construct a causal relationship graph;

[0135] Based on the correlation matrix and the causal relationship graph, an ecological resource credible data element model is constructed.

[0136] Preferably, in a specific application scenario, the set of green finance data elements is wherein represents the kth green finance data element, and P is the number of green finance data elements. The set of basic ecological key features is The set of derived ecological key features is Q and R are the number of basic ecological key features and derived ecological key features, respectively. A hybrid method based on maximum information coefficient (MIC) and kernel canonical correlation analysis (KCCA) is used to calculate the correlation.

[0137] First, the maximum information coefficient between green finance data elements and the basic ecological key features The maximum information coefficient measures the maximum information transmission between two variables by gridding the data, and its calculation process is as follows: let x and y be the data sample sets of and , respectively, and divide the value range of x and y into m and n intervals, respectively, to construct an m x n grid. Let p ij be the probability of data point (x i , y j ) falling into the i-th row and j-th column grid, and p i and p ·j be the marginal probability of the i-th row and j-th column, respectively. Then the mutual information I(x; y) is:

[0138] The maximum information coefficient is defined as the maximum value of the ratio of mutual information I(x; y) to log2min(m, n) among all possible grid divisions, that is: where N is the number of data samples, and B is a preset bandwidth parameter, usually taking 0.6. At the same time, the kernel canonical correlation analysis is used to calculate the canonical correlation coefficient between green finance data elements and the basic ecological key features . Let X and Y be the data matrices of and , respectively, and map the data to a high-dimensional feature space through the kernel function κ to obtain Φ(X) and Φ(Y). Kernel canonical correlation analysis aims to find projection vectors a and b such that the correlation of the projected variables u = a T Φ(X) and v = b T Φ(Y) is maximum. Define the covariance matrix Then the canonical correlation coefficient ρ is the solution to the following generalized eigenvalue problem:

[0139] Finally, the correlation between green finance data elements and the basic ecological key features is measured by combining the maximum information coefficient and the canonical correlation coefficient where λ1 and λ2 are weight coefficients determined by cross-validation.

[0140] Similarly, the correlation between green finance data elements and the derived ecological key features is measured by

[0141] ​Constructing the correlation matrix M r , whose elements are where k = 1, …, P, q = 1, …, Q, r = 1, …, R.

[0142] Preferably, in a specific application scenario, a method based on structural causal model (SCM) and Bayesian structure learning is used to calculate the causality, and when constructing the structural causal model, there is a causal relationship between the green finance data elements D g , the basic ecological key features K b , and the derived ecological key features K d , which can be expressed as: D g = f g (K b , ∈ d , ∈ g ), K b = f b (∈ b ), K d = f d (K b , ∈ d ), where f g , f b , f d are unknown causal functions, ∈ g , ∈ b , ∈ d are independent noise variables.

[0143] A Bayesian structure learning algorithm is used to determine the causal structure. Let be a directed acyclic graph (DAG) representing the causal relationships between variables, with nodes representing variables in D g , K b , K d , and edges representing causal directions. The goal of Bayesian structure learning is to find an optimal that maximizes the posterior probability , where D is the observed data. According to Bayes' theorem, the prior probability can be set according to some prior knowledge or uniform distribution. The likelihood function is obtained by modeling the conditional probability distribution of each node. For each node X, its conditional probability distribution P(X|Pa(X)) (Pa(X) is the parent node set of X) can be represented by a parameterized model, such as a Gaussian distribution (for continuous variables) or a multinomial distribution (for discrete variables). By maximizing the posterior probability , the optimal causal structure , i.e., the causal relationship graph, can be obtained.

[0144] Preferably, in a specific embodiment, the ecological resource credible data element model is constructed based on the correlation matrix and the causal diagram, and a method based on tensor decomposition and deep belief network (DBN) is used to construct the ecological resource credible data element model, which is specifically as follows:

[0145] The correlation matrix M r is regarded as a three-order tensor , in which one dimension corresponds to green financial data elements, one dimension corresponds to basic ecological key features, and the other dimension corresponds to derived ecological key features. High-order singular value decomposition (HOSVD) is used to decompose the tensor :

[0146] , in which is a core tensor, U 1 , U 2 , and U 3 are orthogonal matrices, and x i represents a tensor product along the i-th mode.

[0147] The decomposed core tensor and the causal diagram are used as inputs of a deep belief network (DBN). The deep belief network is stacked by multiple restricted Boltzmann machines (RBM). Let the input of the first RBM be x 1 , the number of visible layer nodes of which is equal to the number of elements of the core tensor , and the number of hidden layer nodes is H1. The energy function of the RBM is: , in which h 1 is a hidden layer state, is a weight between a visible layer node i and a hidden layer node j, and are biases of the visible layer and the hidden layer, respectively.

[0148] The RBM is trained by a contrastive divergence algorithm to obtain a weight matrix W 1 and bias vectors b 1 and c 1 . The hidden layer output of the first RBM is used as the input of the next RBM, and multiple RBMs are trained in sequence to finally obtain parameters of the deep belief network.

[0149] The output of the trained deep belief network is the ecological resource credible data element model . The model can output an evaluation and a prediction of the ecological resource state according to input green financial data elements, basic ecological key features, and derived ecological key features.

[0150] In the above embodiment, the creative physical meanings of various parameters are as follows: D gThis is a collection of green finance data elements, including data related to green finance such as green investment amounts and green loan interest rates. These data reflect the support and impact of the financial sector on ecological resources. b K: A set of key basic ecological features, containing key features related to the natural ecological foundation obtained after key feature extraction, such as biodiversity indicators of forest ecosystems, water flow and key water quality parameters, etc. These features play a crucial role in the stability and function of ecosystems. d The first set of key ecological features represents the set of key characteristics related to ecotourism, leisure and wellness, and other eco-derived fields. Examples include key factors for ecotourism visitor satisfaction and demand characteristics for leisure and wellness services. These features reflect the value and application of ecological resources in derivative industries. The second set of key ecological features represents the number of green finance data elements, reflecting the richness and dimensionality of the data. The third set of key ecological features represents the scale of key features after screening and extraction from basic ecological data. The fourth set of key ecological features represents the number of key features derived from the derived ecological data, showcasing the diversity of key features within the derived ecological data. Green finance data elements Key characteristics of basic ecology The maximum information coefficient, used to measure the maximum amount of information transferred between two variables, is obtained by gridding the data and calculating mutual information. m, n: the number of intervals into which the ranges of data samples x and y are divided when calculating the maximum information coefficient, affecting the accuracy and complexity of the calculation. p ij p i ., p .j In calculating mutual information, the probability and marginal probability of a data point falling into the grid are fundamental parameters. N: Number of data samples, used to calculate the maximum information coefficient and the covariance matrix in kernel canonical correlation analysis. B: Bandwidth parameter in the maximum information coefficient calculation, controlling the grid density, typically set to 0.6, affecting the calculated maximum information coefficient. X, Y: Green finance data elements in kernel canonical correlation analysis. and key characteristics of basic ecology The data matrix is ​​the input data for calculating the canonical correlation coefficient. κ: Kernel function, used to map the data to a high-dimensional feature space, enhancing the separability and correlation analysis capabilities of the data. Common kernel functions include Gaussian kernel and polynomial kernel. C xx C yy C xy: Covariance matrix in kernel canonical correlation analysis, used to compute canonical correlation coefficients. p: Canonical correlation coefficients obtained from kernel canonical correlation analysis, measuring the correlation between two variables after projection into a high-dimensional feature space. l1, l2: Weight coefficients in computing correlation measures by combining maximum information coefficient and canonical correlation coefficients, determined through cross-validation to balance the contributions of both methods to the correlation measure. Green finance data elements Integrated correlation measures with underlying ecological key features , combining the results of maximum information coefficient and canonical correlation coefficient. Green finance data elements Integrated correlation measures with derived ecological key features , computed similarly to r kb . M r : Correlation matrix storing the correlation measure values between green finance data elements and underlying ecological key features, derived ecological key features, which is an important input for subsequent model construction. f g , f b , f d : Unknown causal functions in structural causal models, describing the causal relationships between green finance data elements, underlying ecological key features, and derived ecological key features. e g , e b , e d : Independent noise variables in structural causal models, reflecting random factors not explained in the model. Directed acyclic graph representing causal relationships between variables, with nodes as variables in green finance data elements, underlying ecological key features, and derived ecological key features, and edges representing causal directions. In Bayesian structure learning, given observed data D, the posterior probability of causal structure is the objective function for finding the optimal causal structure. Likelihood function obtained by modeling the conditional probability distribution of each node, reflecting the generation probability of observed data D given causal structure . Prior probability of causal structure , which can be set according to prior knowledge or uniform distribution, affects the results of Bayesian structure learning. Optimal causal structure obtained through Bayesian structure learning, i.e., causal relationship graph, showing the causal relationships between green finance data elements, underlying ecological key features, and derived ecological key features. Regarding correlation matrix M r as a third-order tensor, used for high-order singular value decomposition to extract the underlying structure and features in the data. The core tensor obtained by high-order singular value decomposition retains the main information and characteristics of the tensor . 1 , U 2 , U 3 : orthogonal matrix in high-order singular value decomposition, used for tensor transformation and decomposition. i : tensor product along the i-th mode, is an operation in high-order singular value decomposition. 1 : input of the first restricted Boltzmann machine (RBM) in deep belief network, whose dimension is equal to the number of elements of the core tensor . H1: the number of nodes in the first RBM hidden layer, which affects the learning ability and feature extraction ability of the RBM. E(x 1 , h 1 ): energy function of the RBM, used to define the joint probability distribution of the visible layer and hidden layer states in the RBM.

[0151] Optionally, the above obtaining green finance data elements and establishing the correlation between the elements and the basic ecological key characteristics and the derived ecological key characteristics to generate an ecological resource credible data element model has a specific technical processing process. In the ecological resource management and green finance correlation analysis scene, it has significant advantages compared to traditional technologies:

[0152] 1. Traditional methods often rely on simple Pearson correlation coefficients, which can only measure linear relationships and are powerless against complex nonlinear relationships between ecological and financial data. For example, when analyzing green investment and forest ecosystem biodiversity, it is impossible to dig out the complex dependence relationship hidden behind the data, which may miss key information and lead to one-sided evaluation of the correlation between the two.

[0153] The present scheme combines the maximum information coefficient (MIC) and kernel canonical correlation analysis (KCCA). MIC calculates mutual information by gridding data, which can capture various types of correlations between variables, including nonlinear relationships, and comprehensively measure the amount of information transmission between green finance and ecological key characteristics. KCCA uses kernel functions to map data to high-dimensional space and dig out deep relationships between data. By integrating both, it can accurately identify complex relationships when evaluating the relationship between green credit interest rates and water resource flow and water quality key parameters, providing a more comprehensive and accurate basis for correlation for ecological resource management and green finance decision-making.

[0154] 2. Traditional causal analysis is mostly based on regression models, which are difficult to distinguish between causal and correlational relationships, and are prone to draw incorrect conclusions in complex ecological financial systems. For example, it may misjudge the correlation between the increase in tourist satisfaction and the increase in green investment as a causal relationship, and cannot accurately reveal the internal causal logic between ecological industries and financial support.

[0155] The application is based on structural causal model (SCM) and Bayesian structure learning, and explicitly determines the causal relationship between variables. The optimal causal structure is determined by maximizing the posterior probability through Bayes theorem. It can consider various noise factors and establish appropriate conditional probability distribution for different types of variables. When analyzing the causal relationship between green finance data elements, basic ecological key features and derived ecological key features, it can clearly present the causal paths such as how green investment directly or indirectly affects the development of eco-tourism industry, and the feedback of ecological foundation change to green finance policy, providing a scientific basis for policy making and resource management.

[0156] 3. The traditional model construction method has weak processing ability for high-dimensional and multi-source data, and it is difficult to integrate ecological foundation, derived features and green finance data. When constructing an ecological resource evaluation model, only a single data source or simple association may be considered, resulting in poor model prediction and evaluation ability, and the model cannot adapt to the complex and variable actual situation of ecological finance.

[0157] The application uses tensor decomposition and deep belief network (DBN). The correlation matrix is regarded as a third-order tensor for high-order singular value decomposition, and the potential structural features of the data are extracted, and the complex relationship between multi-dimensional data is preserved. DBN is stacked by multiple restricted Boltzmann machines, which can effectively learn the input data pattern. Taking the construction of an ecological resource credible data element model as an example, it can fully integrate green finance, ecological foundation and derived key feature information, accurately evaluate the state of ecological resources and predict future trends, and provide strong model support for ecological resource management and green finance investment strategy formulation.

[0158] The above description is only the preferred embodiment of the application and the explanation of the applied technical principles. Those skilled in the art should understand that the scope of the application involved in the present application is not limited to the technical solutions formed by the specific combination of the above technical features, and also covers other technical solutions formed by the combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with the technical features disclosed in the present application (but not limited to) having similar functions to form a technical solution.

Claims

1. A method for establishing an intelligent and reliable data element model for ecological resources, characterized in that, include: Acquire basic natural ecological data, which includes at least one of forest description metadata, river description metadata, lake description metadata, grassland description metadata, and cultivated land description metadata; Acquire natural ecology-derived data, including ecotourism descriptive metadata and leisure and wellness descriptive metadata; Controlled feature mining is performed on both basic natural ecological data and derived natural ecological data to generate basic ecological data elements and derived ecological data elements. This includes: constructing a controlled feature mining model based on set business rules and feature mining constraints; and performing controlled feature mining on both basic natural ecological data and derived natural ecological data based on the controlled feature mining model to generate basic ecological data elements and derived ecological data elements. Key features were extracted from both basic ecological data elements and derived ecological data elements to obtain basic ecological key features and derived ecological key features. Acquire green finance data elements and establish their correlation with basic ecological key features and derived ecological key features to generate a credible data element model of ecological resources. This includes: calculating the correlation between green finance data elements and basic ecological key features and derived ecological key features to construct a correlation matrix; calculating the causality between green finance data elements and basic ecological key features and derived ecological key features to construct a causal relationship diagram; and constructing a credible data element model of ecological resources based on the correlation matrix and the causal relationship diagram.

2. The method according to claim 1, characterized in that, The acquisition of basic natural ecological data includes: based on the constructed multi-source data collection rules, parsing carriers carrying different types of basic natural ecological data to obtain the basic natural ecological data from them.

3. The method according to claim 1, characterized in that, The acquisition of natural ecological derived data includes: Structured text is obtained by structuring descriptive text derived from natural ecology. Semantic analysis is performed on the structured text to extract natural ecology-derived feature elements; Natural ecology-derived feature elements are used to generate natural ecology-derived data.

4. The method according to claim 1, characterized in that, The process involves extracting key features from both basic ecological data elements and derived ecological data elements to obtain basic ecological key features and derived ecological key features, including: Based on the established key feature extraction direction, multi-scale transformation is performed on ecological basic data elements and ecological derived data elements to obtain feature information of ecological basic data elements and ecological derived data elements at different scales. Key basic ecological features are obtained by fusing and analyzing the characteristic information of ecological basic data elements at different scales. Key features of the derived ecosystem are obtained by fusing and analyzing the characteristic information of ecological derivative data elements at different scales.

Images