A method and system for ecological environment correlation analysis based on graph neural network

Abstract

The present invention relates to the field of environmental analysis technology, and in particular to a method and system for ecological and environmental association analysis based on graph neural networks. The method comprises: obtaining ecological and environmental data; inputting the ecological and environmental data into a preset heterogeneous graph neural network model to obtain association patterns of the ecological and environmental data, wherein the heterogeneous graph neural network model is constructed based on graph convolutional neural networks and is used to extract node information, edge information, and spatiotemporal dependencies from the ecological and environmental data, and determine key nodes and association paths based on the node information, edge information, and spatiotemporal dependencies to obtain association patterns of the ecological and environmental data. The present invention can effectively mine complex association patterns in ecological and environmental data, providing support for ecological and environmental monitoring, early warning, and decision-making.

Description

Technical Field

[0001] The present invention relates to the field of environmental analysis technology, and in particular to an ecological environment association analysis method and system based on graph neural network. Background Art

[0002] In ecological and environmental correlation analysis, effectively utilizing graph neural network models for feature extraction and correlation mining faces numerous technical challenges. First, ecological and environmental data is highly heterogeneous and dynamic. Different types of nodes and edges contain distinct information that changes over time. This requires graph neural network models to flexibly handle different types of nodes and edges and adapt to dynamically changing network structures. Second, ecological and environmental correlations are multi-scale and multi-level, encompassing both local interactions and global synergistic effects. Graph neural network models must be able to simultaneously capture correlation patterns at different scales and levels and effectively integrate them. Third, ecological and environmental correlations often exhibit spatiotemporal dependencies. The state of a correlation at a given point in time is influenced by historical conditions, and the strength of correlations varies across regions. Incorporating spatiotemporal factors into graph neural network models and modeling spatiotemporal dependencies is an urgent issue. Finally, interpretability is also a crucial aspect of ecological and environmental correlation analysis. Simple correlation analysis results cannot directly guide decision-making; the underlying mechanisms of the correlation patterns need to be explained. However, most existing graph neural network models are black-box models and lack interpretability. Designing an interpretable graph neural network model to make the analysis results more transparent and credible is a direction worth exploring. Therefore, this paper proposes a method and system for ecological environment correlation analysis based on graph neural networks. Summary of the Invention

[0003] The purpose of the present invention is to address the above-mentioned technical problems and provide an ecological environment correlation analysis method and system based on graph neural networks, which can effectively mine complex correlation patterns in ecological environment data, provide support for ecological environment monitoring, early warning and decision-making, and has important theoretical significance and application value.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] A method for ecological environment association analysis based on graph neural network, comprising:

[0006] Obtain ecological and environmental data;

[0007] The ecological environment data is input into a preset heterogeneous graph neural network model to obtain the association pattern of the ecological environment data, wherein the heterogeneous graph neural network model is constructed based on graph convolutional neural networks, and the heterogeneous graph neural network model is used to extract node information, edge information, and spatiotemporal dependencies from the ecological environment data, and determine key nodes and association paths based on the node information, edge information, and spatiotemporal dependencies to obtain the association pattern of the ecological environment data.

[0008] Optionally, the heterogeneous graph neural network model includes: a heterogeneous graph construction unit, a spatiotemporal attention unit, an association feature extraction unit, and an association pattern analysis unit, wherein the heterogeneous graph construction unit is used to construct node embedding vectors and edge embedding vectors based on the ecological environment data to construct a heterogeneous graph; the spatiotemporal attention unit is used to perform time series encoding and spatial position encoding on the ecological environment data to obtain spatiotemporal dependencies; the association feature extraction unit is used to extract multi-level association features based on the heterogeneous graph and spatiotemporal dependencies; the association pattern analysis unit is used to calculate the association strength between nodes based on the association features, and determine the association pattern based on the association strength.

[0009] Optionally, the heterogeneous graph construction unit constructing a node embedding vector and an edge embedding vector based on the ecological environment data includes:

[0010] Constructing a node embedding matrix and an edge embedding matrix based on the ecological environment data;

[0011] The node embedding matrix and the edge embedding matrix are respectively mapped to a low-dimensional space through an embedding layer to obtain the node embedding vector and the edge embedding vector.

[0012] Optionally, the spatiotemporal attention unit performs time series encoding and spatial position encoding on the ecological environment data, and obtaining the spatiotemporal dependency includes:

[0013] Extracting time series information and spatial location information from the ecological environment data;

[0014] Using a time embedding layer to encode the time series information to obtain a time embedding vector;

[0015] Using a spatial embedding layer to encode the spatial position information to obtain a spatial embedding vector;

[0016] Fusing the temporal embedding vector and the spatial embedding vector to obtain a spatiotemporal embedding representation;

[0017] According to the spatiotemporal embedding representation, the attention weights between different spatiotemporal points are calculated through the spatiotemporal attention mechanism;

[0018] Based on the attention weights, the spatiotemporal embedding representations of different spatiotemporal points are weightedly aggregated to construct the spatiotemporal dependency relationship.

[0019] Optionally, the correlation feature extraction unit extracts multi-level correlation features based on the heterogeneous graph and the spatiotemporal dependency relationship, including:

[0020] A multi-scale graph convolution algorithm is used to extract features of the heterogeneous graph and spatiotemporal dependency by setting convolution kernels of different sizes, thereby obtaining local correlation features and global correlation features.

[0021] A graph pooling algorithm is used to fuse the local correlation features and the global correlation features to obtain multi-level correlation features.

[0022] Optionally, the association pattern analysis unit calculates the association strength between nodes based on the association features, and determining the association pattern based on the association strength includes:

[0023] Calculate the similarity between the node embedding vectors corresponding to the associated features, and use the similarity as the association strength measure;

[0024] Sort the association relationships between nodes based on the association strength metric to determine key nodes;

[0025] Taking the key node as a starting point, other nodes are connected based on the association strength metric value, an association path is determined, and the association pattern is obtained.

[0026] Optionally, the heterogeneous graph neural network model also includes a post-processing module, which is used to filter and optimize the association pattern according to the rules of the ecological environment field, remove redundant and invalid association patterns, and obtain the final association pattern.

[0027] To further achieve the above-mentioned purpose, the present invention also provides an ecological environment association analysis system based on graph neural network, including: an environmental data acquisition module and an environmental data analysis module, wherein the environmental data acquisition module is used to acquire ecological environment data; the environmental data analysis module is used to input the ecological environment data into a preset heterogeneous graph neural network model to obtain the association pattern of the ecological environment data, wherein the heterogeneous graph neural network model is constructed based on graph convolutional neural network, and the heterogeneous graph neural network model is used to extract node information, edge information, and spatiotemporal dependency from the ecological environment data, and determine key nodes and association paths based on the node information, edge information, and spatiotemporal dependency to obtain the association pattern of the ecological environment data.

[0028] The beneficial effects of the present invention are:

[0029] This paper addresses the heterogeneity and spatiotemporal dependencies of ecological and environmental data by constructing a heterogeneous graph neural network model. It uses multi-scale graph convolution and graph pooling algorithms to extract multi-level association patterns, and introduces a spatiotemporal attention mechanism to model spatiotemporal dependencies. Furthermore, the paper designs an association metric algorithm based on a graph neural network to calculate the strength of associations between nodes and optimizes the association patterns using a rule-based post-processing method. This paper can effectively mine complex association patterns in ecological and environmental data, providing support for ecological and environmental monitoring, early warning, and decision-making, and has important theoretical significance and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0031] Figure 1 This is a flow chart of an ecological environment association analysis method based on graph neural network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0033] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] Ecological environmental correlation analysis usually includes: (1) Identifying the relationship between environmental factors: analyzing the interactions between different environmental factors (such as climate change, water quality, land use, etc.); (2) Assessing the impact of environmental changes: predicting the impact of environmental changes on ecosystems (such as species diversity, water resources, soil health, etc.) through data analysis; (3) Supporting environmental decision-making: providing scientific basis for governments and relevant decision-makers to help them make more scientific environmental management decisions; (4) Optimizing ecological restoration and protection measures: providing data support for ecological protection and restoration, and optimizing the effectiveness of policies and measures.

[0035] Graph Neural Networks:

[0036] Graph neural networks (GNNs) are a class of neural network models that process graph-structured data. Graph data consists of nodes (vertices) and edges (edges), where nodes represent entities and edges represent relationships between them. In many practical applications, data can essentially be represented as graphs. The fundamental idea of GNNs is to learn node representations (embeddings) by passing information between a node and its neighbors. GNNs iteratively update the state (or features) of each node, aggregating information about its neighbors to obtain a high-level representation of the nodes in the graph.

[0037] A typical GNN calculation process includes the following steps:

[0038] Node feature initialization: Each node in the graph has an initial feature vector. These features can be node attributes (such as user information in a social network) or embedded representations obtained through pre-training.

[0039] Neighborhood information aggregation: Each node updates its feature representation at each layer by aggregating the features of its neighboring nodes (e.g., weighted average, maximum, or sum). Common aggregation methods include sum, average, and maximum.

[0040] Node feature update: A node updates its feature representation using a neural network (such as fully connected layers, activation functions, etc.) based on its aggregated neighbor information and its own current features.

[0041] Multi-layer iteration: The above process is usually repeated multiple times to form a multi-layer graph neural network structure. In each layer, the representation of the node gradually includes more and more information about its neighbors.

[0042] Output layer: The final node representation can be used for tasks such as node classification, graph classification, and link prediction.

[0043] Graph neural networks (GNNs) are a key branch of deep learning, specifically designed to process graph-structured data. They effectively capture the relationships between nodes and their neighbors in a graph through information transfer and aggregation between nodes. With the continuous development of GNNs, a growing number of variants and improved methods (such as GCN, GAT, and GraphSAGE) have emerged, enabling GNNs to achieve excellent results in various fields, such as social networks, chemical molecules, and recommendation systems.

[0044] Based on the above content, this embodiment provides an ecological environment association analysis method based on graph neural network, such as Figure 1 Shown, including:

[0045] Obtain ecological and environmental data;

[0046] The ecological environment data is input into a preset heterogeneous graph neural network model to obtain the association pattern of the ecological environment data, wherein the heterogeneous graph neural network model is constructed based on graph convolutional neural networks, and the heterogeneous graph neural network model is used to extract node information, edge information, and spatiotemporal dependencies from the ecological environment data, and determine key nodes and association paths based on the node information, edge information, and spatiotemporal dependencies to obtain the association pattern of the ecological environment data.

[0047] Specifically, this embodiment addresses the heterogeneity and spatiotemporal dependencies of ecological and environmental data by constructing a heterogeneous graph neural network model. It employs multi-scale graph convolution and graph pooling algorithms to extract multi-level association patterns, and introduces a spatiotemporal attention mechanism to model spatiotemporal dependencies. On this basis, the present invention designs an association metric algorithm based on a graph neural network to calculate the strength of associations between nodes and optimizes the association patterns using a rule-based post-processing method. This invention can effectively mine complex association patterns in ecological and environmental data, providing support for ecological and environmental monitoring, early warning, and decision-making, and has important theoretical significance and application value.

[0048] Furthermore, the heterogeneous graph neural network model includes: a heterogeneous graph construction unit, a spatiotemporal attention unit, an association feature extraction unit, and an association pattern analysis unit, wherein the heterogeneous graph construction unit is used to construct node embedding vectors and edge embedding vectors based on the ecological environment data to construct a heterogeneous graph; the spatiotemporal attention unit is used to perform time series encoding and spatial position encoding on the ecological environment data to obtain spatiotemporal dependencies; the association feature extraction unit is used to extract multi-level association features based on the heterogeneous graph and spatiotemporal dependencies; the association pattern analysis unit is used to calculate the association strength between nodes based on the association features, and determine the association pattern based on the association strength.

[0049] Furthermore, the heterogeneous graph construction unit constructs a node embedding vector and an edge embedding vector based on the ecological environment data, including: constructing a node embedding matrix and an edge embedding matrix based on the ecological environment data; mapping the node embedding matrix and the edge embedding matrix to a low-dimensional space through an embedding layer to obtain the node embedding vector and the edge embedding vector.

[0050] Furthermore, the spatiotemporal attention unit performs time series encoding and spatial position encoding on the ecological environment data, and obtaining the spatiotemporal dependency relationship includes: extracting the time series information and spatial position information in the ecological environment data; using the time embedding layer to encode the time series information to obtain a time embedding vector; using the spatial embedding layer to encode the spatial position information to obtain a spatial embedding vector; fusing the time embedding vector and the spatial embedding vector to obtain a spatiotemporal embedding representation; according to the spatiotemporal embedding representation, calculating the attention weights between different spatiotemporal points through the spatiotemporal attention mechanism; based on the attention weights, weighted aggregation is performed on the spatiotemporal embedding representations of different spatiotemporal points to construct the spatiotemporal dependency relationship.

[0051] Specifically, this embodiment needs to consider both temporal and spatial dimensions when processing complex spatiotemporal data. First, time series information and spatial location information are acquired. Time series information can be data for each node at different time points, while spatial location information includes the latitude and longitude coordinates of each node. The time embedding layer converts discrete time points into continuous vector representations. For example, sine and cosine functions can be used to encode time, mapping the 24 hours of a day into vectors in a high-dimensional space. This captures the cyclical nature of time. The spatial embedding layer converts the node's geographic location information into a vector representation, using grid encoding or learning-based methods to map two-dimensional coordinates into a high-dimensional space. This helps the model understand the spatial relationships between nodes. The fusion of spatiotemporal embedding representations can be achieved through simple concatenation or more complex attention mechanisms. The fused representation integrates information from both temporal and spatial dimensions, providing rich features for subsequent modeling. The core idea of the spatiotemporal attention mechanism is to enable the model to automatically learn the importance weights between different spatiotemporal points. The attention weights are obtained by calculating the similarity between the query vector and the key vector and then normalizing them using a softmax function. The weighted aggregation process performs a weighted summation of features at different spatiotemporal points based on the attention weights. This step enables the selective transfer of information, allowing the model to capture complex spatiotemporal dependencies.

[0052] Furthermore, the correlation feature extraction unit extracts multi-level correlation features based on the heterogeneous graph and spatiotemporal dependency, including: using a multi-scale graph convolution algorithm to extract features from the heterogeneous graph and spatiotemporal dependency by setting convolution kernels of different sizes to obtain local correlation features and global correlation features; using a graph pooling algorithm to fuse the local correlation features and global correlation features to obtain multi-level correlation features.

[0053] Furthermore, the association pattern analysis unit calculates the association strength between nodes based on the association features, and determines the association pattern based on the association strength, including: calculating the similarity between the node embedding vectors corresponding to the association features, and using the similarity as the association strength measurement value; sorting the association relationships between the nodes based on the association strength measurement value to determine the key nodes; taking the key node as the starting point, connecting other nodes based on the association strength measurement value, determining the association path, and obtaining the association pattern.

[0054] Specifically, we use cosine similarity to calculate the strength of association between nodes. For example, consider two nodes, A and B, whose vector representations are [0.2, 0.5, -0.1, 0.8] and [0.1, 0.6, -0.2, 0.7], respectively. The calculated cosine similarity is 0.98, indicating that the interests of these two users are highly similar. This method allows us to identify key nodes with strong associations. Starting from these key nodes, we can trace edges with strong associations, identify association paths, and form association patterns.

[0055] Furthermore, the heterogeneous graph neural network model also includes a post-processing module, which is used to filter and optimize the association pattern according to the rules of the ecological environment field, remove redundant and invalid association patterns, and obtain the final association pattern.

[0056] Specifically, in this embodiment, based on the expert knowledge and experience in the field of ecological environment, a set of rule-based post-processing rule base is established to optimize and screen the association patterns. The optimization and screening are as follows:

[0057] Obtain the initial association pattern results obtained by the association measurement algorithm as input data for post-processing optimization. Traverse each association pattern and determine whether the association pattern is redundant or invalid based on the rule conditions in the post-processing rule base. If the current association pattern meets any rule in the rule base, it will be marked as redundant or invalid and removed from the association result set. For the remaining valid association patterns, the association rule mining algorithm is used to further refine and summarize higher-level association rules. Combined with the knowledge of domain experts, the mined high-level association rules are manually reviewed and confirmed to remove associations that do not conform to the characteristics of the ecological environment. The screened and optimized association patterns and association rules are output as the final association analysis results for subsequent ecological and environmental applications and decision-making references.

[0058] For example, the rule base in this embodiment includes rules such as "The correlation strength between water quality indicators and meteorological conditions should not exceed a threshold." For example, if a correlation pattern shows a correlation of 0.9 between PM2.5 concentration and wind speed, this may be an anomalous correlation requiring further review. After obtaining the initial correlation pattern results, the system examines each correlation pattern one by one. Suppose a correlation pattern indicates that "industrial wastewater discharge is positively correlated with the fish population in nearby rivers." This clearly violates ecological common sense and should be marked as an invalid correlation and removed. For the remaining valid correlation patterns, association rule mining algorithms such as Apriori can be applied for in-depth analysis. For example, by analyzing environmental data from multiple cities, high-level association rules such as "When the area of an industrial park increases by 10% and the greening rate decreases by 5%, the probability of an 8% increase in the air quality index is 75%." During the manual review stage, environmental experts will review these automatically generated rules. For example, a rule may indicate that "wetland area is negatively correlated with bird diversity." This contradicts ecological theory and is therefore removed by experts. On the contrary, the rule that “urban green space ratio is positively correlated with residents’ happiness” may be retained because it is consistent with the research findings of environmental psychology. The final output of the correlation analysis results may include such as “within 10 km of the industrial park, for every additional air quality monitoring station, the annual average concentration of PM2.5 decreases by 2μg / m 3 Such a specific association pattern.

[0059] As another preferred embodiment, this method introduces a semi-supervised learning method into the training of the constructed heterogeneous graph neural network model.

[0060] Specifically, a semi-supervised learning approach is used to address the issue of sparse labels in ecological and environmental data. Through methods such as node and edge reconstruction, a self-supervised task for unlabeled data is constructed, and self-supervised learning is used to obtain the implicit feature representation of the data. The label information obtained from semi-supervised learning and the feature representation obtained from self-supervised learning are integrated to optimize the similarity calculation method in the association measurement algorithm. Based on the optimized similarity calculation method, association analysis is performed on ecological and environmental data to explore the inherent connections and patterns between the data. Utilizing the results of the association analysis, labels are completed and inferred for data with sparse labels, thereby expanding the data's label information. Techniques such as data augmentation and regularization are also introduced to improve the robustness of the model, reduce the risk of overfitting, and enhance the model's applicability to actual ecological and environmental data.

[0061] By combining machine learning with expert knowledge, we can more comprehensively and accurately understand complex ecological and environmental systems, providing strong support for the development of scientific environmental protection strategies. The resulting correlation analysis results can help decision-makers find the optimal balance between economic development and environmental protection, and achieve sustainable development.

[0062] In order to further optimize the above technical solution, this embodiment also provides an ecological environment association analysis system based on graph neural network, including: an environmental data acquisition module and an environmental data analysis module, wherein the environmental data acquisition module is used to acquire ecological environment data; the environmental data analysis module is used to input the ecological environment data into a preset heterogeneous graph neural network model to obtain the association pattern of the ecological environment data, wherein the heterogeneous graph neural network model is constructed based on graph convolutional neural network, and the heterogeneous graph neural network model is used to extract node information, edge information, and spatiotemporal dependency from the ecological environment data, and determine key nodes and association paths based on the node information, edge information, and spatiotemporal dependency to obtain the association pattern of the ecological environment data.

[0063] Furthermore, the heterogeneous graph neural network model includes: a heterogeneous graph construction unit, a spatiotemporal attention unit, an association feature extraction unit, and an association pattern analysis unit, wherein the heterogeneous graph construction unit is used to construct node embedding vectors and edge embedding vectors based on the ecological environment data to construct a heterogeneous graph; the spatiotemporal attention unit is used to perform time series encoding and spatial position encoding on the ecological environment data to obtain spatiotemporal dependencies; the association feature extraction unit is used to extract multi-level association features based on the heterogeneous graph and spatiotemporal dependencies; the association pattern analysis unit is used to calculate the association strength between nodes based on the association features, and determine the association pattern based on the association strength.

[0064] Furthermore, the heterogeneous graph construction unit constructs a node embedding vector and an edge embedding vector based on the ecological environment data, including: constructing a node embedding matrix and an edge embedding matrix based on the ecological environment data; mapping the node embedding matrix and the edge embedding matrix to a low-dimensional space through an embedding layer to obtain the node embedding vector and the edge embedding vector.

[0065] Furthermore, the spatiotemporal attention unit performs time series encoding and spatial position encoding on the ecological environment data, and obtaining the spatiotemporal dependency relationship includes: extracting the time series information and spatial position information in the ecological environment data; using the time embedding layer to encode the time series information to obtain a time embedding vector; using the spatial embedding layer to encode the spatial position information to obtain a spatial embedding vector; fusing the time embedding vector and the spatial embedding vector to obtain a spatiotemporal embedding representation; according to the spatiotemporal embedding representation, calculating the attention weights between different spatiotemporal points through the spatiotemporal attention mechanism; based on the attention weights, weighted aggregation is performed on the spatiotemporal embedding representations of different spatiotemporal points to construct the spatiotemporal dependency relationship.

[0066] Furthermore, the correlation feature extraction unit extracts multi-level correlation features based on the heterogeneous graph and spatiotemporal dependency, including: using a multi-scale graph convolution algorithm to extract features from the heterogeneous graph and spatiotemporal dependency by setting convolution kernels of different sizes to obtain local correlation features and global correlation features; using a graph pooling algorithm to fuse the local correlation features and global correlation features to obtain multi-level correlation features.

[0067] Furthermore, the association pattern analysis unit calculates the association strength between nodes based on the association features, and determines the association pattern based on the association strength, including: calculating the similarity between the node embedding vectors corresponding to the association features, and using the similarity as the association strength measurement value; sorting the association relationships between the nodes based on the association strength measurement value to determine the key nodes; taking the key node as the starting point, connecting other nodes based on the association strength measurement value, determining the association path, and obtaining the association pattern.

[0068] Furthermore, the heterogeneous graph neural network model also includes a post-processing module, which is used to filter and optimize the association pattern according to the rules of the ecological environment field, remove redundant and invalid association patterns, and obtain the final association pattern.

[0069] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A method for ecological environment association analysis based on graph neural network, characterized in that: include: Obtain ecological and environmental data; Inputting the ecological environment data into a preset heterogeneous graph neural network model to obtain a correlation pattern of the ecological environment data, wherein the heterogeneous graph neural network model is constructed based on graph convolutional neural networks, and the heterogeneous graph neural network model is used to extract node information, edge information, and spatiotemporal dependencies from the ecological environment data, and determine key nodes and correlation paths based on the node information, edge information, and spatiotemporal dependencies to obtain the correlation pattern of the ecological environment data; The heterogeneous graph neural network model includes: a heterogeneous graph construction unit, a spatiotemporal attention unit, an association feature extraction unit, and an association pattern analysis unit, wherein the heterogeneous graph construction unit is used to construct a node embedding vector and an edge embedding vector based on the ecological environment data to construct a heterogeneous graph; the spatiotemporal attention unit is used to perform time series encoding and spatial position encoding on the ecological environment data to obtain spatiotemporal dependencies; the association feature extraction unit is used to extract multi-level association features based on the heterogeneous graph and spatiotemporal dependencies; the association pattern analysis unit is used to calculate the association strength between nodes based on the association features, and determine the association pattern based on the association strength.

2. The ecological environment association analysis method based on graph neural network according to claim 1 is characterized in that: The heterogeneous graph construction unit constructs a node embedding vector and an edge embedding vector based on the ecological environment data, including: Constructing a node embedding matrix and an edge embedding matrix based on the ecological environment data; The node embedding matrix and the edge embedding matrix are respectively mapped to a low-dimensional space through an embedding layer to obtain the node embedding vector and the edge embedding vector.

3. The ecological environment association analysis method based on graph neural network according to claim 1 is characterized in that: The spatiotemporal attention unit performs time series encoding and spatial position encoding on the ecological environment data to obtain spatiotemporal dependency relationships, including: Extracting time series information and spatial location information from the ecological environment data; Using a time embedding layer to encode the time series information to obtain a time embedding vector; Using a spatial embedding layer to encode the spatial position information to obtain a spatial embedding vector; Fusing the temporal embedding vector and the spatial embedding vector to obtain a spatiotemporal embedding representation; According to the spatiotemporal embedding representation, the attention weights between different spatiotemporal points are calculated through the spatiotemporal attention mechanism; Based on the attention weights, the spatiotemporal embedding representations of different spatiotemporal points are weightedly aggregated to construct the spatiotemporal dependency relationship.

4. The ecological environment association analysis method based on graph neural network according to claim 1 is characterized in that: The correlation feature extraction unit extracts multi-level correlation features based on the heterogeneous graph and the spatiotemporal dependency relationship, including: A multi-scale graph convolution algorithm is used to extract features of the heterogeneous graph and spatiotemporal dependency by setting convolution kernels of different sizes, thereby obtaining local correlation features and global correlation features. A graph pooling algorithm is used to fuse the local correlation features and the global correlation features to obtain multi-level correlation features.

5. The ecological environment association analysis method based on graph neural network according to claim 1 is characterized in that: The association pattern analysis unit calculates the association strength between nodes based on the association features, and determining the association pattern based on the association strength includes: Calculate the similarity between the node embedding vectors corresponding to the associated features, and use the similarity as the association strength measure; Sort the association relationships between nodes based on the association strength metric to determine key nodes; Taking the key node as a starting point, other nodes are connected based on the association strength metric value, an association path is determined, and the association pattern is obtained.

6. The ecological environment association analysis method based on graph neural network according to claim 1 is characterized in that: The heterogeneous graph neural network model also includes a post-processing module, which is used to filter and optimize the association pattern according to the rules of the ecological environment field, remove redundant and invalid association patterns, and obtain the final association pattern.

7. An ecological environment association analysis system based on graph neural network, characterized in that: include: An environmental data acquisition module and an environmental data analysis module, wherein the environmental data acquisition module is used to acquire ecological environment data; the environmental data analysis module is used to input the ecological environment data into a preset heterogeneous graph neural network model to obtain the association pattern of the ecological environment data, wherein the heterogeneous graph neural network model is constructed based on graph convolutional neural networks, and the heterogeneous graph neural network model is used to extract node information, edge information, and spatiotemporal dependencies from the ecological environment data, and determine key nodes and association paths based on the node information, edge information, and spatiotemporal dependencies to obtain the association pattern of the ecological environment data; The heterogeneous graph neural network model includes: a heterogeneous graph construction unit, a spatiotemporal attention unit, an association feature extraction unit, and an association pattern analysis unit, wherein the heterogeneous graph construction unit is used to construct a node embedding vector and an edge embedding vector based on the ecological environment data to construct a heterogeneous graph; the spatiotemporal attention unit is used to perform time series encoding and spatial position encoding on the ecological environment data to obtain spatiotemporal dependencies; the association feature extraction unit is used to extract multi-level association features based on the heterogeneous graph and spatiotemporal dependencies; the association pattern analysis unit is used to calculate the association strength between nodes based on the association features, and determine the association pattern based on the association strength.

Images