Green infrastructure planning method, system and equipment based on AI drive

Abstract

The invention discloses a green infrastructure planning method based on AI driving. The green infrastructure planning method comprises the following steps: S1, generating a training sample set X and an actual label set y; s2, selecting a weight vector omega and a bias b; s3, generating a sample (xi, yi); s4, calculating a linear combination z = omegaxi + b of (xi, yi), omega and b; s5, obtaining a prediction label # imgabs0 # of the sample (xi, yi) through f (z), S6, comparing # imgabs1 # and yi, if # imgabs2 # and yi are the same, executing a step S7, and if # imgabs3 # and yi are different, executing a step S8; s7, generating another sample (xi, yi) according to the training sample set X and the actual label set y, and repeating S4, S5 and S6; s8, updating omega and b, generating another sample (xi, yi) according to the training sample set X and the actual label set y, and repeating S4, S5 and S6; and S9, after traversing all samples (xi, yi), generating a prediction model z = omegaxi + b and # imgabs4 # by using the current omega and b, thereby being capable of combining GIS data, utilizing machine learning to identify a spatial mode, scientifically planning green infrastructure, automatically iterating and optimizing the model, and rapidly providing an optimal decision scheme.

Description

Technical Field

[0001] The present invention relates to the field of land planning, and in particular to an AI-driven green infrastructure planning method, system and device. Background Art

[0002] In the current urbanization process, the planning and construction of green infrastructure are of great significance for maintaining and enhancing the ecological functions of the city, improving the quality of life of residents and promoting the sustainable development of the city; green infrastructure usually includes natural and semi-natural spaces such as parks, green spaces, water bodies, greenways, etc., which can not only provide places for leisure and entertainment, but also improve the urban microclimate, purify the air, reduce noise pollution, and protect biodiversity.

[0003] Existing green infrastructure planning methods are based on a comprehensive consideration of factors such as urban spatial structure, land use patterns, population distribution and environmental quality. However, faced with the rapidly changing urban environment and increasingly complex social needs, traditional methods have limitations when processing large-scale multi-dimensional data, and it is difficult to comprehensively and systematically consider all influencing factors. In addition, with the expansion of urban scale and diversification of functions, the demand for green infrastructure is also increasing, which puts higher requirements on planning work.

[0004] In recent years, with the rapid development of artificial intelligence (AI) technology, the application of AI in urban planning and management has gradually increased. AI technology can process and analyze large amounts of data, identify patterns and trends, and provide a scientific basis for decision-making. Therefore, in green infrastructure planning, AI technology can be used to analyze multi-dimensional data such as urban spatial structure, land use patterns, population distribution, and environmental quality, predict future trends, optimize planning schemes, and improve the scientificity and adaptability of planning.

[0005] Although AI technology has great potential in green infrastructure planning, there is currently a lack of mature AI-driven planning methods, systems and equipment. Existing research and practice mainly focus on single data analysis and model construction, lacking a systematic planning methodology and integrated planning platform. In addition, the application of AI technology in green infrastructure planning also faces problems such as difficulty in data acquisition, insufficient model generalization ability, and poor interpretability of planning results. Summary of the invention

[0006] The present invention aims to solve the technical problems in the above-mentioned technology at least to some extent.

[0007] To this end, the present invention discloses, in a first aspect, a green infrastructure planning method based on AI driving, comprising the following steps:

[0008] S1: Generate training sample set X={x1,x2,…,xn} and the actual label set y = {y1,y2,…,y n}, where x i is the feature vector of the ith plot, y i is the actual label of the green infrastructure of the ith plot; S2: randomly select a weight vector ω = (ω1, ω2, …, ω m ) and bias b, where m is the feature vector x i S3: According to the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1, y2, ..., y n}Generate a sample (x i ,y i ); S4: Calculate the sample (x i ,y i ), the weight vector ω=(ω1,ω2,…,ω m ) and the linear combination of the bias b, specifically, z = ωx i +n; S5: Apply activation function f(z) to obtain the sample (x i ,y i ) S6: Comparing predicted labels and the actual label y i , if the predicted label and the actual label y i If the predicted label is the same, execute step 7. and the actual label y i If they are different, execute step 8; S7: according to the training sample set X={x1,x2,…,x n} and the actual label set y = {y1, y2, ..., y n} Generate another sample (x i ,y i ) and repeat S4, S5, S6 until all the samples (x i ,y i ); S8: Update the weight vector ω=(ω1,ω2,…,ω m ) and the bias b, according to the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1, y2, ..., y n} Generate another sample (x i ,y i ) and repeat S4, S5, S6 until all the samples (x i ,y i ); S9: When traversing all the samples (xi ,y i ), using the current weight vector ω=(ω1,ω2,…,ω m ) and the bias b generate a prediction model, specifically, z = ωx i +b and

[0009] The AI-driven green infrastructure planning method disclosed in the present invention has the following beneficial effects:

[0010] (1) Improving the scientific nature of planning. By combining data generated by the Geographic Information System (GIS), the present invention can use machine learning algorithms to analyze and process large amounts of multi-dimensional data and identify potential spatial patterns and trends, making green infrastructure planning more scientific.

[0011] (2) Optimizing the decision-making process: The model training module provided by the invention can automatically and iteratively update the weight vector and bias parameters, thereby continuously optimizing the prediction model. This automated process helps to quickly obtain the optimal solution.

[0012] In addition, the AI-driven green infrastructure planning method disclosed in the present invention may also have the following additional technical features:

[0013] In one embodiment of the present invention, the application of the activation function f(z) to obtain the sample (x i ,y i ) Specifically, the activation function The value of the activation function f(z) is the predicted label The value of .

[0014] In one embodiment of the present invention, the updating of the weight vector ω=(ω1,ω2,…,ω m ) and the bias b, specifically, Among them, η is the learning rate.

[0015] In one embodiment of the present invention, the training sample set X={x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n}, where x i is the feature vector of the ith plot, y i is the actual label of the green infrastructure of the ith plot, including the following steps:

[0016] S1.1: Collect GIS layer data related to green infrastructure; S1.2: Perform at least data cleaning, standardization and spatial resampling preprocessing on the GIS layer data; S1.3: Divide the study area into multiple land units i; S1.4: Extract the feature vector x related to green infrastructure in each land unit i i And based on the feature vector x i Generate training sample set X = {x1, x2, ..., x n}, where the feature vector x i It is composed of the attribute values ​​of the GIS layer data preprocessed in S1.2; S1.5: Determine the actual label y of each land unit i i And generate the actual label set y = {y1,y2,…,y n}.

[0017] In one embodiment of the present invention, the training sample set X={x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n}, where x i is the feature vector of the ith plot, y i is the actual label of the green infrastructure of the ith plot, including the following steps:

[0018] S1.6: Verify the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1, y2, ..., y n}, specifically, |X|=|y|=n, where |X| is the training sample set X={x1,x2,…,x n}, |y| is the actual label set y = {y1, y2, ..., y n}; S1.7: Verify the feature vector x i The dimensions are, Where m is the eigenvector x i Dimension.

[0019] In one embodiment of the present invention, the AI-driven green infrastructure planning method further includes the following steps:

[0020] S10: Generate a test sample set V = {v1, v2, ..., v n}, where v j is the feature vector of the jth plot; S11: Use the prediction model to process the test sample set V = {v1, v2, ..., v n} and generate a test result. If the test result reaches a preset test threshold, the prediction model is used as the final model. If the test result does not reach the preset test threshold, the learning rate η is adjusted until the test result reaches the preset test threshold, and the prediction model is used as the final model.

[0021] In a second aspect, the present invention discloses a green infrastructure planning system based on AI driving, comprising:

[0022] A data processing module is used to generate the training sample set X={x1, x2,…, x n} and the actual label set y = {y1, y2, ..., y n}; Model training module, used to randomly generate the weight vector ω=(ω1,ω2,…,ω m ) and the bias b, and iteratively update the weight vector ω=(ω1,ω2,…,ω m ) and the bias b; a verification module for using the prediction model to predict the test sample set V = {v1, v2, ..., v n}Processing and generating a prediction result; an evaluation and adjustment module, used to evaluate whether the prediction result reaches the preset test threshold, and adjust the learning rate η when the preset test threshold is not reached, until the prediction result reaches the preset test threshold.

[0023] The present invention discloses a device in a third aspect, comprising:

[0024] At least one processor and at least one memory, wherein when the processor executes the readable instructions in the memory, the AI-driven green infrastructure planning method as described in any one of claims 1 to 6 can be implemented.

[0025] Additional contents and advantages of the present invention will be given in the following description or may be understood through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The technical solutions and beneficial effects of the present invention will become apparent and easily understood from the following contents in conjunction with the accompanying drawings, wherein:

[0027] Figure 1 A flowchart of a green infrastructure planning method based on AI drive according to an embodiment of the present invention;

[0028] Figure 2 A flowchart of an AI-driven green infrastructure planning method according to another embodiment of the present invention;

[0029] Figure 3A flowchart of an AI-driven green infrastructure planning method according to another embodiment of the present invention;

[0030] Figure 4 A flowchart of an AI-driven green infrastructure planning method according to another embodiment of the present invention;

[0031] Figure 5 This is an architecture diagram of an AI-driven green infrastructure planning system according to another embodiment of the present invention. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0033] The following will describe the AI-driven green infrastructure planning method, system and device disclosed in the present invention with reference to the accompanying drawings.

[0034] like Figure 1 As shown, an AI-driven green infrastructure planning method includes the following steps:

[0035] S1: Generate training sample set X={x1,x2,…,x n} and the actual label set y = {y1,y2,…,y n}, where x i is the feature vector of the ith plot, y i is the actual label of the green infrastructure of the ith plot;

[0036] S2: Randomly select a weight vector ω=(ω1,ω2,…,ω m ) and bias b, where m is the feature vector x i The number of features in ;

[0037] S3: According to the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n}Generate a sample (x i ,y i );

[0038] S4: Calculate the sample (x i ,y i ), weight vector ω=(ω1,ω2,…,ω m ) and the linear combination of the bias b, specifically, z = ωx i +b;

[0039] S5: Apply activation function f(z) to get sample (x i ,y i )

[0040] S6: Comparing predicted labels and the actual label y i , if the predicted label and the actual label y i If the predicted label is the same, execute S7. and the actual label y i If they are different, execute S8;

[0041] S7: According to the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n} Generate another sample (x i ,y i ) and repeat S4, S5, S6 until all samples (x i ,y i );

[0042] S8: Update weight vector ω=(ω1,ω2,…,ω m ) and bias b, according to the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n} Generate another sample (x i ,y i ) and repeat S4, S5, S6 until all samples (x i ,y i );

[0043] S9: When traversing all samples (x,y i ), use the current weight vector ω=(ω1,ω2,…,ω m ) and bias b to generate a prediction model, specifically, z = ωx i +b and

[0044] In the embodiment of the present invention, for GIS coating data, relevant staff can obtain land use layers, soil type layers, population density layers or traffic flow layers, etc.; relevant staff divides the study area into 100 plots of 500m×500m, extracts land use type (such as residential area, commercial area or industrial area, etc.), soil type, population density and traffic flow for each plot i, and marks the actual label y of each plot i , where y i =1 means that the land is suitable for building green infrastructure,i =-1 means that the land is not suitable for the construction of green infrastructure;

[0045] The relevant staff randomly selects the weight vector ω=(ω1,ω2,…,ω m ) and bias b, and then from the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n}Select a sample (x i ,y i ), assuming that the feature vector x of this sample i =(0.7, 0.6, 0.3, 08), the four values ​​correspond to land use type, soil type, population density and traffic flow, respectively, the actual label y i =1 means that the plot i is suitable for the construction of green infrastructure;

[0046] Assuming the current weight vector ω=(0.2,0.3,-0.1,0.4) and the bias b=0.1, the linear combination is calculated, specifically, z=0.2×0.7+0.3×0.6+(-0.1)×0.3+0.4×0.8+0.1=0.71; using the activation function f(z) to get the sample (x i ,y i ) Assume that if z>1, then If z < 1, then Obviously at this time

[0047] Because the predicted label The same as the actual label y = 1, and then execute S7, from the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n}Select the next sample (x i ,y i ), then repeat S4, S5, S6 until all samples (x i ,y i );

[0048] If in the above steps, the predicted label If it is different from the actual label y, then update the weight vector ω=(ω1,ω2,…,ω m ) and bias b;

[0049] Finally, when all samples (x i ,y i ) is traversed, using the current weight vector ω=(ω1,ω2,…,ω m ) and bias b complete the prediction model: z = ωxi +b and

[0050] According to the AI-driven green infrastructure planning method disclosed in the present invention, on the one hand, the scientific nature of planning can be improved. By combining the data generated by the geographic information system (GIS), the present invention can use machine learning algorithms to analyze and process large amounts of multi-dimensional data and identify potential spatial patterns and trends, which makes green infrastructure planning more scientific. On the other hand, it can optimize the decision-making process. The model training module provided by the invention can automatically iterate and update the weight vector and bias parameters, thereby continuously optimizing the prediction model. This automated process helps to quickly obtain the optimal solution.

[0051] In order to better explain the present invention, in one embodiment of the present invention, the activation function f(z) is applied to obtain the sample (x i ,y i ) Specifically, the activation function The value of the activation function f(z) is the predicted label The value of .

[0052] In the embodiment of the present invention, the relevant staff can use a step function as the activation function f(z). In addition, the relevant staff can also use a Sigmoid function, specifically, The linear combination z = ωx is transformed by the Sigmoid function i +b is mapped to (0,1), and this value can be understood as the probability that the land parcel is suitable for green infrastructure;

[0053] Taking z=0.71 in the above embodiment as an example, the corresponding This means that there is a 67% probability that the land is suitable for building green infrastructure;

[0054] In addition, due to And 0.67 is slightly close to y i =1, so the prediction of the model can also be considered correct, but for y i = -1, obviously we cannot think that 0.67 is close, so will be considered an incorrect prediction.

[0055] For other technical contents, please refer to the above embodiments, which will not be described in detail in this embodiment.

[0056] In order to better explain the present invention, in one embodiment of the present invention, the weight vector ω is updated to be (ω1, ω2, ..., ω m ) and bias b, specifically, Among them, η is the learning rate.

[0057] In the embodiment of the present invention, the learning rate η is used to control the weight vector ω=(ω1,ω2,…,ω m ) and the update step size of the parameters of bias b. The relevant staff can set the learning rate η to a small positive number according to the actual situation, such as η = 0.01;

[0058] In addition, relevant staff can also dynamically adjust the value of the learning rate η, for example, in the weight vector ω=(ω1,ω2,…,ω m ) and the number of iterations of the bias b reaches a certain number, the learning rate η is multiplied by any number between (0,1), and the new learning rate η is used to continue the next weight vector ω=(ω1,ω2,…,ω m ) and the iteration of bias b.

[0059] For other technical contents, please refer to the above embodiments, which will not be described in detail in this embodiment.

[0060] In order to better explain the present invention, in one embodiment of the present invention, as Figure 2 As shown, according to GIS, the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n}, where x i is the feature vector of the ith plot, y i is the actual label of the green infrastructure of the ith plot, including the following steps:

[0061] S1.1: Collect GIS layer data related to green infrastructure;

[0062] Relevant staff can select corresponding layers according to actual conditions, such as land use type layer, soil type layer, population density layer, traffic flow layer, hydrology layer and environmental quality layer;

[0063] S1.2: Perform at least data cleaning, standardization and spatial resampling preprocessing on GIS layer data;

[0064] First, the incomplete or erroneous data of the above layers are cleaned, then all layer data are standardized to ensure that the scale of all layer data is consistent, and all are converted to between (0,1), and finally the layer data is spatially resampled to keep its spatial resolution consistent;

[0065] S1.3: Divide the study area into multiple land units i;

[0066] In this step, the study area is generally divided into smaller plots as neatly as possible, but the area of ​​the plots should not be too small, otherwise the increase in the number of plots will lead to an exponential increase in the subsequent calculation amount;

[0067] S1.4: Extract the feature vector x related to green infrastructure in each land unit i i And based on the feature vector x i Generate training sample set X = {x1, x2, ..., x n}, where the eigenvector x i It consists of the attribute values ​​of the GIS layer data preprocessed in S1.2;

[0068] S1.5: Determine the actual label y for each land unit i i And generate the actual label set y = {y1,y2,…,y n}.

[0069] For other technical contents, please refer to the above embodiments, which will not be described in detail in this embodiment.

[0070] In order to better explain the present invention, in one embodiment of the present invention, as Figure 3 As shown, according to GIS, the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n}, where x i is the feature vector of the ith plot, y i is the actual label of the green infrastructure of the ith plot, including the following steps:

[0071] S1.6: Verify the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n}, specifically, |X|=|y|=n, where |X| is the training sample set X={x1,x2,…,x n}, |y| is the actual label set y = {y1,y2,…,y n} the number of elements in;

[0072] In this step, it is mainly used to ensure that the training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n}, the number of elements in the same, the relevant staff can use the comparison function to implement this step;

[0073] S1.7: Verify feature vector x i The dimensions are, Where m is the eigenvector x i Dimensions;

[0074] In this step, it is mainly used to ensure that each eigenvector x i The dimensions in are consistent, and the relevant staff can traverse all feature vectors x i And check its length to achieve;

[0075] In S1.6 and S1.7, if data mismatch or missing occurs, inconsistent data points are automatically deleted or missing values ​​are automatically filled.

[0076] For other technical contents, please refer to the above embodiments, which will not be described in detail in this embodiment.

[0077] In order to better explain the present invention, in one embodiment of the present invention, as Figure 4 As shown, the AI-driven green infrastructure planning method also includes the following steps:

[0078] S10: Generate a test sample set V = {v1, v2, ..., v n}, where v j is the feature vector of the jth plot;

[0079] S11: Use the prediction model to process the test sample set V = {v1, v2, ..., v n} and generate test results. If the test results reach the preset test threshold, the prediction model is used as the final model. If the test results do not reach the preset test threshold, the learning rate η is adjusted until the test results reach the preset test threshold, and the prediction model is used as the final model.

[0080] In the embodiment of the present invention, generally speaking, the training sample set X = {x1, x2, ..., x n} and the test sample set V = {v1,v2,…,v n The ratio between them is 8:2. For example, if there are 1000 plots, 800 of them are used for training sample set X = {x1, x2, …, x n}, 200 for testing sample set V = {v1, v2, …, v n};

[0081] The preset test threshold can be set to 95%, that is, the test sample set V = {v1, v2, ..., v n}There are at least 190 (v i ,y i ) and the actual label y i Correspondingly, this means that the prediction model is good enough to be used as the final model, otherwise it cannot be used as the final model.

[0082] For other technical contents, please refer to the above embodiments, which will not be described in detail in this embodiment.

[0083] Moreover, it should be noted that Figure 5 As shown, an AI-driven green infrastructure planning system 100 includes:

[0084] Data processing module 101, used to generate a training sample set X = {x1, x2, ..., x n} and the actual label set y = {y1,y2,…,y n};

[0085] Model training module 102, used to randomly generate a weight vector ω=(ω1,ω2,…,ω m ) and bias b, and iteratively updates the weight vector ω=(ω1,ω2,…,ω m ) and bias b;

[0086] Verification module 103, used to use the prediction model to test the sample set V = {v1, v2, ..., v n}Process and generate prediction results;

[0087] The evaluation and adjustment module 104 is used to evaluate whether the prediction result reaches a preset test threshold, and adjust the learning rate η when the preset test threshold is not reached, until the prediction result reaches the preset test threshold.

[0088] In addition, it should be noted that a device includes:

[0089] At least one processor and at least one memory, wherein when the processor executes the readable instructions in the memory, the AI-driven green infrastructure planning method as claimed in any one of claims 1 to 6 can be implemented.

[0090] In summary, the AI-driven green infrastructure planning method disclosed in the present invention can, on the one hand, improve the scientific nature of planning. By combining the data generated by the geographic information system (GIS), the present invention can use machine learning algorithms to analyze and process large amounts of multi-dimensional data and identify potential spatial patterns and trends, which makes green infrastructure planning more scientific. On the other hand, it can optimize the decision-making process. The model training module provided by the invention can automatically iterate and update the weight vector and bias parameters, thereby continuously optimizing the prediction model. This automated process helps to quickly obtain the optimal solution.

[0091] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A green infrastructure planning method based on AI, characterized in that: The following steps are involved: S1: Generate training sample set X={x1,x2,…,x n } and the actual label set y = {y1,y2,…,y n }, where x i is the feature vector of the ith plot, y i is the actual label of the green infrastructure of the ith plot; S2: Randomly select a weight vector ω=(ω1,ω2,…,ω m ) and bias b, where m is the feature vector x i The number of features in ; S3: According to the training sample set X={x1, x2, ..., x n } and the actual label set y = {y1, y2, ..., y n }Generate a sample (x i ,y i ); S4: Calculate the sample (x i ,y i ), the weight vector ω=(ω1,ω2,…,ω m ) and the linear combination of the bias b, specifically, z = ωx i +b; S5: Apply activation function f(z) to obtain the sample (x i ,y i ) S6: Comparing predicted labels and the actual label y i , if the predicted label and the actual label y i If the predicted label is the same, execute S7. and the actual label y i If they are different, execute S8; S7: According to the training sample set X={x1, x2, ..., x n } and the actual label set y = {y1, y2, ..., y n } Generate another sample (x i ,y i ) and repeat S4, S5, S6 until all the samples (x i ,y i ); S8: Update the weight vector ω=(ω1,ω2,…,ω m ) and the bias b, according to the training sample set X = {x1, x2, ..., x n } and the actual label set y = {y1, y2, ..., y n } Generate another sample (x i ,y i ) and repeat S4, S5, S6 until all the samples (x i ,y i ); S9: When traversing all the samples (x i ,y i ), using the current weight vector ω=(ω1,ω2,…,ω m ) and the bias b generate a prediction model, specifically, z = ωx i +b and 2. The AI-driven green infrastructure planning method according to claim 1, characterized in that: The application of the activation function f(z) obtains the sample (x i ,y i ) Specifically, the activation function The value of the activation function f(z) is the predicted label The value of .

3. The AI-driven green infrastructure planning method according to claim 1, characterized in that: The updating weight vector ω=(ω1,ω2,…,ω m ) and the bias b, specifically, Among them, η is the learning rate.

4. The AI-driven green infrastructure planning method according to claim 1, characterized in that: The training sample set X={x1, x2, ..., x n } and the actual label set y = {y1,y2,…,y n }, where x i is the feature vector of the ith plot, y i is the actual label of the green infrastructure of the ith plot, including the following steps: S1.1: Collect GIS layer data related to green infrastructure; S1.2: performing at least data cleaning, standardization and spatial resampling preprocessing on the GIS layer data; S1.3: Divide the study area into multiple land units i; S1.4: Extract the feature vector x related to green infrastructure in each land unit i i And based on the feature vector x i Generate training sample set X = {x1, x2, ..., x n }, where the feature vector x i It is composed of the attribute values ​​of the GIS layer data preprocessed in S1.2; S1.5: Determine the actual label y of each land unit i i And generate the actual label set y = {y1,y2,…,y n }.

5. The AI-driven green infrastructure planning method according to claim 4, characterized in that: The training sample set X={x1, x2, ..., x n } and the actual label set y = {y1,y2,…,y n }, where x i is the feature vector of the ith plot, y i is the actual label of the green infrastructure of the ith plot, including the following steps: S1.6: Verify the training sample set X = {x1, x2, ..., x n } and the actual label set y = {y1, y2, ..., y n }, specifically, |X|=|y|=n, where |X| is the training sample set X={x1,x2,…,x n }, |y| is the actual label set y = {y1, y2, ..., y n } the number of elements in; S1.7: Verify the feature vector x i The dimensions are, dim(x i )=m, where m is the eigenvector x i Dimension.

6. The AI-driven green infrastructure planning method according to claim 1, characterized in that: The following steps are also included: S10: Generate a test sample set V = {v1, v2, ..., v n }, where v j is the feature vector of the jth plot; S11: Use the prediction model to process the test sample set V = {v1, v2, ..., v n } and generate a test result. If the test result reaches a preset test threshold, the prediction model is used as the final model. If the test result does not reach the preset test threshold, the learning rate η is adjusted until the test result reaches the preset test threshold, and the prediction model is used as the final model.

7. An AI-driven green infrastructure planning system, characterized in that: The method for executing the AI-driven green infrastructure planning method according to claim 1 comprises: A data processing module is used to generate the training sample set X={x1, x2,…, x n } and the actual label set y = {y1, y2, ..., y n }; Model training module, used to randomly generate the weight vector ω=(ω1,ω2,…,ω m ) and the bias b, and iteratively update the weight vector ω=(ω1,ω2,…,ω m ) and the bias b; A verification module is used to use the prediction model to verify the test sample set V = {v1, v2, ..., v n }Process and generate prediction results; An evaluation and adjustment module is used to evaluate whether the prediction result reaches the preset test threshold, and to adjust the learning rate η when the preset test threshold is not reached, until the prediction result reaches the preset test threshold.

8. A device, characterized in that include: At least one processor and at least one memory, wherein when the processor executes the readable instructions in the memory, the AI-driven green infrastructure planning method as described in any one of claims 1 to 6 can be implemented.

Images