A food production ecological efficiency calculation system and calculation method

Abstract

The present invention discloses a food production ecological efficiency calculation system and a calculation method thereof, which relates to the field of ecological efficiency assessment technology, including: constructing an input-output indicator system, collecting basic food production data of different regional blocks in the target area, using the SBM-Undesirable model to establish a food production ecological efficiency calculation model, and calculating and outputting the food production ecological efficiency of different regional blocks; extracting the temporal characteristics, spatial characteristics and spatiotemporal characteristics of the food production ecological efficiency of the target area; constructing a prediction model to analyze the temporal evolution of the food production ecological efficiency and the spatial mobility analysis results of the target area. The present invention incorporates environmental costs into the analysis framework, measures the production efficiency level more comprehensively, and uses deep learning methods to obtain the temporal evolution and spatial mobility of the food production ecological efficiency, providing data basis for improving the regional food production layout and regional cooperation.

Description

Technical Field

[0001] The present invention relates to the technical field of eco-efficiency assessment, and more specifically, to a food production eco-efficiency calculation system and a calculation method thereof. Background Art

[0002] As the main industry of food production, agriculture provides necessary support for the daily needs of residents. In recent years, the continuous innovation of agricultural technology and the transformation and upgrading of agricultural product processing have greatly increased the food supply and the total agricultural output value. However, while the agricultural economy has achieved rapid transformation and upgrading, a series of agricultural environmental pollution problems have also emerged: the unscientific use of pesticides, fertilizers, and agricultural films has brought about the problem of chemical residues, which not only pollutes the ecological environment, but also poses a challenge to people's life, health and safety; the popularization of large agricultural machinery has greatly increased the emission of polluting gases such as tail gas during agricultural production, resulting in an increase in agricultural carbon emissions. In order to solve these agricultural ecological and environmental problems, it is urgent to conduct in-depth research on the development level of regional agricultural ecological efficiency and its main influencing factors, and accelerate the solution of food supply and agricultural ecological problems.

[0003] The traditional food production efficiency evaluation system only includes the input and desired output of food production, including the impact of social factors such as capital, labor and land in the agricultural production process, but does not take into account the impact of food production on the ecological environment, and ignores the natural environment factors closely related to agricultural production, resulting in performance evaluation that is too one-sided and unable to scientifically reflect the true situation of food production. Therefore, it is necessary to further enrich and improve the efficiency evaluation index system, refine the research perspective to analyze the development trend, spatial distribution characteristics and influencing factors of food production ecological efficiency, and conduct a systematic exploration of food ecological efficiency. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a food production ecological efficiency measurement system and a measurement method thereof, the purpose of which is to measure the ecological efficiency of food production through the super-efficiency SBM-Undesirable model, and to conduct a scientific and accurate evaluation of the food ecological efficiency to provide a reference for measuring the coordination between food production and resources and the environment.

[0005] The present invention provides a food production ecological efficiency calculation system, including: a food production basic data collection module, an input-output indicator system module, a food production ecological efficiency calculation module, a trend prediction module and a data analysis and visualization module;

[0006] The basic data collection module for grain production collects basic data on grain production in different regions of the target area and performs preprocessing on the basic data on grain production;

[0007] The input-output indicator system module selects output variables and input variables, constructs an input-output indicator system, and determines indicator parameters from basic grain production data according to the input-output indicator system;

[0008] The food production ecological efficiency calculation module uses the SBM-Undesirable model to establish a food production ecological efficiency calculation model, imports the indicator parameters into the food production ecological efficiency calculation model, and calculates and outputs the food production ecological efficiency of different regional blocks;

[0009] The trend prediction module extracts the temporal characteristics, spatial characteristics and spatiotemporal characteristics of the ecological efficiency of food production in different regional blocks of the target area, and constructs a prediction model to predict the ecological efficiency of food production in different regional blocks;

[0010] The data analysis and visualization module performs liquidity analysis on the food production eco-efficiency prediction data, and visualizes the food production eco-efficiency prediction data and the liquidity analysis results.

[0011] In this solution, the basic data collection module for grain production is specifically:

[0012] Obtaining basic data on grain production in the target area, dividing the target area into regional blocks according to preset regional division standards, matching the regional blocks with the basic data on grain production in the region, standardizing the basic data on grain production, using PCA dimensionality reduction processing to reduce data dimensionality, and clustering data according to grain production input conditions and grain production output conditions;

[0013] The food production input status data set and the food production output status data set are obtained through clustering, and sliding window processing is performed on different data sets to increase the dimensional correlation and time correlation of the basic data of food production in the data set;

[0014] Variable keyword extraction is used to perform variable extraction in the food production input status data set and the food production output status data set to extract input variables, expected output variables and unexpected output variables.

[0015] In this solution, the input-output indicator system module is specifically:

[0016] Obtain the input variables, expected output variables and unexpected output variables of different regional blocks in the target area, and count the input variables, expected output variables and unexpected output variables shared by different regional blocks;

[0017] Obtain food output variables from the common expected output variables to construct expected output indicators, select relevant variables from the common non-expected output variables based on the perspective of non-point source pollution and carbon emissions, and use the selected relevant variables and specific variables in the common input variables to retrieve and obtain the participating eco-efficiency evaluation examples;

[0018] Based on SHAP, the eco-efficiency evaluation example is explained, the characteristic variables involved in the eco-efficiency example are extracted, and the Shapley value of each characteristic variable is generated by SHAP local attribution analysis to characterize the contribution of the characteristic variable, obtain the local explanation of each eco-efficiency example, and construct a local explanation matrix;

[0019] Each row of the local explanation matrix represents an instance, and each column represents the Shapley value of each feature variable. The average value of the absolute value of the Shapley value of each feature variable is calculated based on the local explanation matrix to generate the global importance of the feature variable.

[0020] The characteristic variables are grouped according to the non-expected output variables and input variables, and the characteristic variables are sorted in descending order according to the global importance of the characteristic variables in the input variable grouping, a preset number of characteristic variables are obtained to construct input indicators, and non-point source pollution factors and carbon emission factors are constructed according to the global importance of the characteristic variables in the non-expected output variable grouping, and non-expected output indicators are constructed based on the constructed non-point source pollution factors and carbon emission factors;

[0021] Obtain the specific variables and variable descriptions corresponding to the expected output indicators, non-expected output indicators, and input indicators, build an input-output indicator system, and use the constructed input-output indicator system to determine the indicator parameters in the basic data of grain production.

[0022] In this solution, the food production ecological efficiency calculation module is specifically:

[0023] A food production ecological efficiency calculation model is constructed based on the super efficiency SBM-Undesirable model, and non-point source pollution and carbon emissions are taken as undesirable outputs. The food production ecological efficiency calculation model includes n decision units, each of which is constructed by an expected output index, an undesirable output index, and an input index. The index parameters are imported into the food production ecological efficiency calculation model to calculate and output the food production ecological efficiency of different regional blocks;

[0024] The food production eco-efficiency calculation model is expressed as:

[0025]

[0026] in, represents the ecological efficiency of food production, Respectively represent the total number of expected output indicators, undesired output indicators, and input indicators, The number of items representing input indicators, expected output indicators, and undesired output indicators, They represent the input index, expected output and unexpected output index parameters respectively. represents the weight vector, represents the number of decision-making units, represents the input redundant variable, represents the expected output deficit variable, Represents the variable of undesired excess output.

[0027] In this solution, the trend prediction module is specifically:

[0028] Obtaining the food production eco-efficiency sequences of different regional blocks in the target area at a preset time step, spatially annotating the food production eco-efficiency sequences based on the spatial distribution of the different regional blocks, and performing time convolution on the food production eco-efficiency sequences of the different regional blocks through the first time convolution stage;

[0029] Obtaining examples of food production eco-efficiency driving factor analysis through big data retrieval methods, extracting influencing factors of food production eco-efficiency driving factor analysis examples to construct data samples, evaluating the importance scores of influencing factors based on the mRMR algorithm, and determining the influencing factors of food production eco-efficiency in the target area based on the importance scores;

[0030] Determine the influencing factor characteristics of different regional blocks according to the influencing factors, calculate the correlation between regional blocks using the influencing factor characteristics, food production ecological efficiency sequence and geographic information, and determine the correlation between regional blocks based on the Pearson correlation coefficient;

[0031] Constructing a spatial graph of the ecological efficiency of food production in regional blocks, wherein the regional blocks are used as nodes in the spatial graph, the influencing factor characteristics, the ecological efficiency sequence of food production and the geographical information of each regional block are used as node attributes, the edge structure between the nodes is determined according to the association between the regional blocks, and an adjacency matrix is ​​constructed to characterize the spatial graph;

[0032] The spatial graph is represented and learned by using a graph attention network, and the regional block nodes and the adjacency matrix are introduced into the graph attention network. The attention coefficient of each edge in the spatial graph is calculated by the graph attention mechanism to weight the regional block nodes and extract the spatial characteristics of the ecological efficiency of food production in different regional blocks.

[0033] The spatial features of the graph attention network output are obtained for calculation in the second time convolution stage, and the outputs of the two time convolution stages are spliced ​​to obtain the time characteristics of the ecological efficiency of food production in different regional blocks in the target area. The first time convolution stage and the second time convolution stage are both composed of a two-dimensional convolution module and a GLU module.

[0034] In this solution, the temporal features and spatial features are fused to obtain spatiotemporal features, specifically:

[0035] The acquired temporal features and spatial features are concatenated as the input of the Transformer architecture, and the input is embedded. A gated multi-head attention mechanism is introduced into the Transformer architecture, and the linear regression of each attention head in the gated multi-head attention mechanism is replaced by a graph convolution to improve the spatial correlation of the mapped features.

[0036] The output of the gated multi-head attention mechanism is obtained and concatenated with the original input processed by the activation function, and the time attention representation of each attention head is determined by using an update gate. All attention heads are aggregated to obtain the output of the Transformer architecture, and the spatiotemporal characteristics of the ecological efficiency of food production in different regional blocks in the target area are obtained;

[0037] The temporal features, spatial features and spatiotemporal features are introduced into a feedforward neural network, and a flattening layer and a linear head are used to obtain a prediction of the ecological efficiency of food production in different regional blocks at a preset future time.

[0038] In this solution, the data analysis and visualization module is specifically:

[0039] According to the food production ecological efficiency of different regional blocks in the target area at the current time stamp, the regional blocks are clustered using the food production ecological efficiency to obtain the horizontal distribution of different food production ecological efficiency levels;

[0040] The prediction model and observation model are constructed by iterative Kalman filtering algorithm, and the state vector is constructed according to the ecological efficiency of food production in the corresponding regional blocks with the same level distribution. The prediction model is used to track the estimated value corresponding to the level of ecological efficiency of food production at different levels of distribution, and the predicted value of ecological efficiency of food production in each regional block is used for correction in the observation model.

[0041] The level distribution is updated according to the estimated values ​​corresponding to the levels of ecological efficiency of food production, and the liquidity analysis results are obtained through the differences in level distribution.

[0042] Another aspect of the present invention provides a method for calculating the ecological efficiency of food production, comprising the following steps:

[0043] Construct an input-output indicator system, collect basic data on grain production in different regions of the target area, and calculate indicator parameters based on the input-output system;

[0044] Using the SBM-Undesirable model to establish a food production ecological efficiency calculation model, importing the indicator parameters into the food production ecological efficiency calculation model, and calculating and outputting the food production ecological efficiency of different regional blocks;

[0045] Obtain the time series and distribution of food production ecological efficiency in different blocks of the target area, and extract the time series, spatial characteristics and spatiotemporal characteristics of food production ecological efficiency;

[0046] A prediction model is constructed based on the extracted temporal features, spatial features and spatiotemporal features, and the prediction model is used to predict the ecological efficiency of food production in different regional blocks. The liquidity analysis is performed through the food production ecological efficiency prediction data, and the food production ecological efficiency prediction data and liquidity analysis are visualized.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The present invention measures the ecological efficiency of food production through the super-efficiency SBM-Undesirable model, incorporates environmental costs into the analysis framework, avoids the overestimation of the true efficiency value, measures the production efficiency level more comprehensively, and conducts a scientific and accurate assessment of the ecological efficiency of food to provide a reference for measuring the coordination between food production and resources and the environment.

[0049] This paper uses deep learning methods to obtain the temporal evolution and spatial fluidity of the ecological efficiency of food production, explores its spatiotemporal evolution laws and fluidity characteristics through analysis from temporal and spatial perspectives, and provides data reference for improving regional food production layout and regional cooperation. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments or exemplary embodiments of the present invention, the drawings required for use in the embodiments or exemplary descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained according to the drawings without paying creative work.

[0051] Figure 1 The architecture diagram of the food production eco-efficiency calculation system is shown;

[0052] Figure 2 A flowchart of constructing an input-output indicator system according to an embodiment is shown;

[0053] Figure 3A flow chart showing an embodiment of extracting temporal and spatial characteristics of food production eco-efficiency in a target area;

[0054] Figure 4 A flow chart showing the method for calculating the eco-efficiency of food production. DETAILED DESCRIPTION

[0055] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0056] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0057] Figure 1 The architecture diagram of the food production eco-efficiency calculation system is shown;

[0058] like Figure 1 As shown, the embodiment of the present invention provides a food production ecological efficiency calculation system 1, including a food production basic data collection module 101, an input-output indicator system module 102, a food production ecological efficiency calculation module 103, a trend prediction module 104 and a data analysis and visualization module 105;

[0059] The basic data collection module for grain production collects basic data on grain production in different regions of the target area and performs preprocessing on the basic data on grain production;

[0060] The input-output indicator system module selects output variables and input variables, constructs an input-output indicator system, and determines indicator parameters from basic grain production data according to the input-output indicator system;

[0061] The food production ecological efficiency calculation module uses the SBM-Undesirable model to establish a food production ecological efficiency calculation model, imports the indicator parameters into the food production ecological efficiency calculation model, and calculates and outputs the food production ecological efficiency of different regional blocks;

[0062] The trend prediction module extracts the temporal characteristics, spatial characteristics and spatiotemporal characteristics of the ecological efficiency of food production in different regional blocks of the target area, and constructs a prediction model to predict the ecological efficiency of food production in different regional blocks;

[0063] The data analysis and visualization module performs time series evolution and liquidity analysis on the food production eco-efficiency prediction data, and visualizes the food production eco-efficiency prediction data, time series evolution and liquidity analysis results.

[0064] The basic data collection module for grain production collects basic data on grain production in different regions of the target area and performs preprocessing on the basic data on grain production;

[0065] The input-output indicator system module selects output variables and input variables, constructs an input-output indicator system, and determines indicator parameters from basic grain production data according to the input-output indicator system;

[0066] The food production ecological efficiency calculation module uses the SBM-Undesirable model to establish a food production ecological efficiency calculation model, imports the indicator parameters into the food production ecological efficiency calculation model, and calculates and outputs the food production ecological efficiency of different regional blocks;

[0067] The trend prediction module extracts the temporal characteristics, spatial characteristics and spatiotemporal characteristics of the ecological efficiency of food production in different regional blocks of the target area, and constructs a prediction model to predict the ecological efficiency of food production in different regional blocks;

[0068] The data analysis and visualization module performs liquidity analysis on the food production eco-efficiency prediction data, and visualizes the food production eco-efficiency prediction data and the liquidity analysis results.

[0069] It should be noted that the basic data collection module for grain production is responsible for acquiring the basic data on grain production in the target area, reflecting the basic conditions for grain production in the target area in terms of location, topography, climate and agricultural technology, dividing the target area into regional blocks according to a preset regional division standard, preferably dividing it according to administrative location or gridding the target area, and dividing the grid according to the terrain conditions to obtain regional blocks; matching the regional blocks with the basic data on grain production in the region, standardizing the basic data on grain production, using PCA dimensionality reduction processing to reduce data dimensionality, and clustering data according to grain production input status and grain production output status; obtaining a grain production input status data set and a grain production output status data set through clustering, the grain production input status data set includes data such as agricultural mechanization, application of agricultural materials, farmland facilities, grain planting area and number of farmers, and the grain production output status data set includes data such as grain output, comprehensive index of agricultural non-point source pollution and agricultural carbon emissions. Sliding window processing is performed on different data sets to increase the dimensional correlation and time correlation of the basic data of grain production in the data set; variable keyword extraction is used to extract variables in the grain production input status data set and the grain production output status data set to extract input variables, expected output variables and unexpected output variables.

[0070] Figure 2 A flow chart of constructing an input-output indicator system according to an embodiment is shown.

[0071] According to an embodiment of the present invention, the input-output indicator system module constructs an input-output indicator system, specifically:

[0072] S202, obtaining input variables, expected output variables and undesired output variables of different regional blocks in the target area, and counting the input variables, expected output variables and undesired output variables shared by different regional blocks;

[0073] S204, obtaining food output variables from the common expected output variables to construct expected output indicators, selecting relevant variables from the common non-expected output variables based on the perspectives of non-point source pollution and carbon emissions, and using the selected relevant variables and specific variables in the common input variables to retrieve and obtain the participating eco-efficiency evaluation instance;

[0074] S206, interpreting the eco-efficiency evaluation instance based on SHAP, extracting characteristic variables involved in the eco-efficiency instance, generating Shapley values ​​of each characteristic variable using SHAP local attribution analysis, characterizing the contribution of the characteristic variable, obtaining a local explanation of each eco-efficiency instance, and constructing a local explanation matrix;

[0075] S208, each row of the local explanation matrix represents an instance, each column represents the Shapley value of each feature variable, and the average value of the absolute value of the Shapley value of each feature variable is calculated based on the local explanation matrix to generate the global importance of the feature variable;

[0076] S210, grouping the characteristic variables according to the non-desired output variables and input variables, sorting the characteristic variables in descending order according to the global importance of the input variable group, obtaining a preset number of characteristic variables to construct an input index, sorting the characteristic variables in descending order according to the global importance of the characteristic variables in the non-desired output variable group to construct non-point source pollution factors and carbon emission factors, and constructing non-desired output indicators based on the constructed non-point source pollution factors and carbon emission factors;

[0077] S212, obtaining specific variables and variable descriptions corresponding to expected output indicators, non-expected output indicators, and input indicators, constructing an input-output indicator system, and using the constructed input-output indicator system to determine indicator parameters in the basic data of grain production.

[0078] It should be noted that among the common non-desired output variables, relevant variables are selected from the perspectives of non-point source pollution and carbon emissions. The relevant variables from the perspective of non-point source pollution include fertilizer application, pesticide application, agricultural film use, total nitrogen and total phosphorus loss, etc. The relevant variables from the perspective of carbon emissions include fertilizer, pesticide, agricultural film, diesel, agricultural cultivation, tillage and other carbon emission sources. The selected relevant variables and specific variables in the common input variables are used to retrieve the participating eco-efficiency evaluation examples, and the SHAP algorithm is used for interpretation. The SHAP algorithm is based on cooperative game theory and local interpretation. According to the marginal contribution of each feature in the instance, the contribution of each feature to the prediction result is calculated to achieve the interpretation of each instance. By training the SHAP algorithm, a Shapley value is generated for any feature variable in the instance to represent the contribution to the prediction result. For any feature variable, the weighted sum of all possible feature value combinations is taken to obtain the Shapley value of the feature variable. During the training process, the Shapley value of each feature variable and the Shapley mean of all training samples are obtained. When the Shapley value of the feature variable is greater than the Shapley mean, a positive impact is generated, otherwise a negative impact is generated. In order to perform a global interpretation to determine the overall contribution of the feature variable, the average of the absolute values ​​of the Shapley values ​​of each feature variable is used to construct the global importance. The global importance The calculation formula is: ,in Indicates the total number of instances, Indicates In the example The Shapley value of the characteristic variable.

[0079] In the input-output indicator system, the expected output variable is the grain yield data, in kilograms per mu. The global importance is used to screen the specific variables of the input indicators. Preferably, the input indicators include labor, land, and capital. Among them, the labor input selects the number of working days per mu as a proxy variable; the land selects the cost of land per mu; and the capital includes the input of machinery and pesticides, as well as the application amount of seeds and fertilizers. Labor input, the average number of working days per mu of grain; land input, the average land cost per mu of grain, yuan / mu; machinery input, the average machinery operation cost per mu of grain, yuan / mu; seed input, the average seed usage per mu of grain, kilograms / mu; fertilizer input, the average fertilizer (pure) application per mu of grain, kilograms / mu. And use the global importance to determine the relevant variables for calculating non-point source pollution and carbon emission factors. Preferably, the undesired output variables are the carbon dioxide emissions from food production caused by fertilizers, pesticides, agricultural films, and tillage, in kilograms per ton; the non-point source pollution emissions from food production, the use of fertilizers, pesticides, and agricultural films multiplied by their non-point source pollution coefficients, the non-point source pollution emitted per ton of food produced, in kilograms per ton.

[0080] It should be noted that the food production ecological efficiency calculation module, based on the determination of expected output and unexpected output of food production, constructs a random preamble model, and selects the super-efficiency SBM model including undesirable output to calculate the ecological efficiency of food production. The SBM-Undesirable model not only overcomes the shortcomings of the traditional DEA model's preset production function and the distribution form of key technical variables, but also improves the explanatory power of the ordinary SBM model. It can also introduce undesirable output into the model to measure the production ecological efficiency and calculate the ecological efficiency of food production. The super-efficiency SBM-Undesirable model not only overcomes the shortcomings of the traditional DEA model's preset production function and the distribution form of key technical variables, but also improves the explanatory power of the ordinary SBM model. It can also introduce undesirable output into the model to measure green production efficiency.

[0081] A food production ecological efficiency calculation model is constructed based on the super efficiency SBM-Undesirable model, and non-point source pollution and carbon emissions are taken as undesirable outputs. The food production ecological efficiency calculation model includes n decision units, each of which is constructed by an expected output index, an undesirable output index, and an input index. The index parameters are imported into the food production ecological efficiency calculation model to calculate and output the food production ecological efficiency of different regional blocks;

[0082] The food production eco-efficiency calculation model is expressed as:

[0083]

[0084] in, represents the ecological efficiency of food production, Respectively represent the total number of expected output indicators, undesired output indicators, and input indicators, The number of items representing input indicators, expected output indicators, and undesired output indicators, They represent the input index, expected output and unexpected output index parameters respectively. represents the weight vector, represents the number of decision-making units, represents the input redundant variable, represents the expected output deficit variable, Represents the variable of undesired excess output.

[0085] The ecological efficiency of food production is in the range of 0-1. When the ecological efficiency of food production is 1, the ecological efficiency of food is effective, there is no redundant input factors, undesirable excess output and insufficient expected output. When the ecological efficiency of food production is less than 1, the ecological efficiency of food has not reached an effective state, and the input-output indicator system needs to be adjusted.

[0086] Figure 3 A flow chart of an embodiment for extracting temporal and spatial characteristics of the ecological efficiency of food production in a target area is shown.

[0087] According to an embodiment of the present invention, the trend prediction module extracts the temporal and spatial characteristics of the ecological efficiency of food production in the target area, specifically:

[0088] S302, obtaining the food production eco-efficiency sequences of different regional blocks in the target area at a preset time step, spatially annotating the food production eco-efficiency sequences based on the spatial distribution of the different regional blocks, and performing time convolution on the food production eco-efficiency sequences of the different regional blocks through a first time convolution stage;

[0089] S304, obtaining an analysis example of driving factors of food production eco-efficiency through a big data retrieval method, extracting influencing factors of the analysis example of driving factors of food production eco-efficiency to construct a data sample, evaluating the importance scores of the influencing factors based on the mRMR algorithm, and determining the influencing factors of food production eco-efficiency in the target area according to the importance scores;

[0090] S306, determining the influencing factor characteristics of different regional blocks according to the influencing factors, calculating the correlation between the regional blocks using the influencing factor characteristics, the food production eco-efficiency sequence and the geographic information, and determining the correlation between the regional blocks based on the Pearson correlation coefficient;

[0091] S308, constructing a spatial graph of the ecological efficiency of food production in regional blocks, wherein the regional blocks are used as nodes in the spatial graph, and the influencing factor characteristics, the ecological efficiency sequence of food production and the geographical information of each regional block are used as node attributes, the edge structure between the nodes is determined according to the association between the regional blocks, and an adjacency matrix is ​​constructed to characterize the spatial graph;

[0092] S310, using a graph attention network to perform representation learning on the spatial graph, importing the regional block nodes and the adjacency matrix into the graph attention network, calculating the attention coefficient of each edge in the spatial graph through the graph attention mechanism, achieving weighting of the regional block nodes, and extracting the spatial characteristics of the ecological efficiency of food production in different regional blocks;

[0093] S312, obtain the spatial features of the graph attention network output to perform calculations in the second time convolution stage, splice the outputs of the two time convolution stages, and obtain the time characteristics of the ecological efficiency of food production in different regional blocks in the target area. The first time convolution stage and the second time convolution stage are both composed of a two-dimensional convolution module and a GLU module.

[0094] It should be noted that the driving factors of the ecological efficiency of food production include mechanization density, level of fiscal support for agriculture, farmers' income level, planting structure, industrialization level, urbanization rate, location variables, etc. The mRMR algorithm is used for feature selection to extract the combination of influencing factors with the greatest correlation with the ecological efficiency of food production and the least redundancy between influencing factors.

[0095] A first time convolution stage and a second time convolution stage are respectively set before and after the graph attention network, and the time convolution stage includes a two-dimensional convolution and a GLU module. The input of the first time convolution stage is the food production ecological efficiency sequence of the regional block at a preset time step, and the input of the second time convolution stage is the spatial features between the regional blocks obtained through the graph attention network. The input data is convolved and feature mapped by the two-dimensional convolution and GLU modules. The two-dimensional convolution can extract the time features of multiple time steps near each timestamp, and then the time series features are mapped. The outputs of the first time convolution stage and the second time convolution stage are integrated using a splicing operation to obtain the output time features.

[0096] The spatial graph is constructed using the influencing factor characteristics of regional blocks, the sequence of ecological efficiency of food production and geographic information. The regional blocks are taken as nodes, and the influencing factor characteristics, the sequence of ecological efficiency of food production and the geographic information of each regional block node are taken as the attributes of the node. The Pearson correlation coefficient is introduced to determine the edge structure between the nodes. In the calculation of the geographic information, the shortest distance between the nodes of two regional blocks is calculated according to the latitude and longitude data combined with the Haversine formula. The adjacency matrix is ​​constructed according to the correlation between the regional blocks, and the spatial graph is represented by the graph attention network. The weights between the nodes are dynamically adjusted using the graph attention mechanism, which does not rely on a fixed weight matrix, and reflects the flexibility of the ecological efficiency of food production between regional blocks to a certain extent. The node characteristics and adjacency matrix of the regional block nodes are imported into the graph attention network as input. The node characteristics include the influencing factor characteristics, the sequence of ecological efficiency of food production and geographic information. The strength of the spatial relationship is quantified by calculating the attention coefficient of each edge.

[0097] The acquired temporal features and spatial features are concatenated as the input of the Transformer architecture, and the temporal position information of the food production eco-efficiency sequence is used to embed the input. A gated multi-head attention mechanism is introduced into the Transformer architecture. The gated multi-head attention mechanism consists of a self-attention mechanism and an update gate, which can reduce dependence on external information and is better at learning the internal correlation of data. Scaled dot product attention is used here to calculate the self-attention score. By adding an update gate, the model can better handle long-term dependencies in time series data and selective updates of information.

[0098] And replace the linear regression of each attention head in the gated multi-head attention mechanism with graph convolution to improve the spatial correlation of the mapped features. The formula for the graph convolution calculation is: ,in are the output and input features of graph convolution, respectively. represents the adjacency matrix of the spatial graph, Represents trainable parameters, uses graph convolution to replace linear regression in the self-attention calculation process, and obtains the attention parameter matrix , , ,in represents the input sequence, Respectively represent the query vector, key vector and value vector of the input sequence. The output of the gated multi-head attention mechanism is obtained and spliced ​​with the original input processed by the activation function, and the time attention representation of each attention head is determined by using the update gate. All attention heads are aggregated to obtain the output of the Transformer architecture, and the spatiotemporal characteristics of the ecological efficiency of food production in different regional blocks in the target area are obtained; the temporal characteristics, spatial characteristics and spatiotemporal characteristics are imported into the feedforward neural network training prediction network, and a flattening layer and a linear head are used in the prediction network to obtain the ecological efficiency of food production in different regional blocks at a preset future time.

[0099] It should be noted that the data analysis and visualization module clusters the regional blocks according to the ecological efficiency of food production in different regional blocks of the target area at the current timestamp, and obtains the horizontal distribution of different levels of ecological efficiency of food production, namely, low level, medium-low level, medium-high level and high level;

[0100] The prediction model and observation model are constructed by iterative Kalman filter algorithm, and the target state is estimated by linear difference equation. The iterative Kalman filter algorithm is mainly divided into two stages: prediction and measurement update. In the update step, if the detection is successful, the target state is updated with the actual measurement value; if the detection fails, the target state is updated with the best prediction value of the previous frame. The state vector is constructed according to the ecological efficiency of food production in the corresponding regional blocks of the same horizontal distribution, and the prediction model is used to track the estimated value corresponding to the level of ecological efficiency of food production in different horizontal distributions, and the predicted value of ecological efficiency of food production in each regional block is used for correction in the observation model; the horizontal distribution is updated according to the estimated value corresponding to the level of ecological efficiency of food production, and the liquidity analysis results are obtained through the difference in horizontal distribution, the transfer of the ecological efficiency level type of food production in the target area is obtained, and the stability and spillover of the ecological efficiency of food production in different regional blocks are evaluated.

[0101] Use visualization management software to visualize the food production eco-efficiency forecast data, and visualize the stability assessment results and spillover trends using different visual effects based on liquidity analysis to reflect the level of change in the eco-efficiency of food production in the target area, making it easier to formulate a series of targeted policy recommendations through comparative analysis of time series evolution, average eco-efficiency, etc.

[0102] Figure 4 A flow chart showing the method for calculating the eco-efficiency of food production.

[0103] Another embodiment of the present invention provides a method for calculating the ecological efficiency of food production, comprising the following steps:

[0104] S402, constructing an input-output indicator system, collecting basic data on grain production in different regions of the target area, and calculating indicator parameters based on the input-output system;

[0105] S404, using the SBM-Undesirable model to establish a food production ecological efficiency calculation model, importing the indicator parameters into the food production ecological efficiency calculation model, and calculating and outputting the food production ecological efficiency of different regional blocks;

[0106] S406, obtaining the time series of food production ecological efficiency and the distribution of food production ecological efficiency in different regional blocks of the target area, and extracting the time series characteristics, spatial characteristics and spatiotemporal characteristics of food production ecological efficiency;

[0107] S408, constructing a prediction model based on the extracted temporal features, spatial features and spatiotemporal features, using the prediction model to predict the ecological efficiency of food production in different regional blocks, and performing liquidity analysis through the predicted data of ecological efficiency of food production, and visualizing the predicted data of ecological efficiency of food production and liquidity analysis.

[0108] According to an embodiment of the present invention, a database of the ecological efficiency of grain production in the target area is constructed, basic data on grain production in different regional blocks, results of grain production ecological efficiency calculation, and predicted values ​​of grain production ecological efficiency in future time periods are stored in the database, and portraits of grain production ecological efficiency in different regional blocks are generated in the database. The ecological efficiency portraits of grain production are used to generate an optimization plan for the configuration of grain production materials in the regional block, and actual grain production ecological efficiency change values ​​are extracted according to historical configuration optimization instances. Historical configuration optimization instances that meet preset requirements are selected and combined with a data-driven model to train a benchmark model. The grain production material configuration optimization plan for the regional block is imported into the benchmark model to obtain an ideal change value of the grain production ecological efficiency of the regional block after a preset time. The predicted value of the grain production ecological efficiency after the preset time is used to calculate the residual with the current measured value, and the absolute value of the residual is compared with the absolute value of the ideal change value of the grain production ecological efficiency. When the comparison deviation is less than the preset deviation threshold, it means that the grain production material configuration optimization plan in the regional block is not applicable to the regional block.

[0109] In the several embodiments provided in the present application, it should be understood that the disclosed method can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as: multiple modules can be combined, or can be integrated into another system, or some features can be ignored, or not executed. All functional modules can be integrated into one processing module, or each module can be a separate module, or two or more modules can be integrated into one module; the above-mentioned integrated modules can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0110] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0111] Alternatively, if the above-mentioned integrated module of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0112] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A food production ecological efficiency calculation system, characterized in that: The system includes: basic data collection module for grain production, input-output indicator system module, grain production ecological efficiency calculation module, trend prediction module and data analysis and visualization module; The basic data collection module for grain production collects basic data on grain production in different regions of the target area and performs preprocessing on the basic data on grain production; The input-output indicator system module selects output variables and input variables, constructs an input-output indicator system, and determines indicator parameters from basic grain production data according to the input-output indicator system; The food production ecological efficiency calculation module uses the SBM-Undesirable model to establish a food production ecological efficiency calculation model, imports the indicator parameters into the food production ecological efficiency calculation model, and calculates and outputs the food production ecological efficiency of different regional blocks; The trend prediction module extracts the temporal characteristics, spatial characteristics and spatiotemporal characteristics of the ecological efficiency of food production in different regional blocks of the target area, and constructs a prediction model to predict the ecological efficiency of food production in different regional blocks; The data analysis and visualization module performs liquidity analysis on the food production eco-efficiency prediction data, and visualizes the food production eco-efficiency prediction data and the liquidity analysis results; The input-output indicator system module is specifically: Obtain the input variables, expected output variables and unexpected output variables of different regional blocks in the target area, and count the input variables, expected output variables and unexpected output variables shared by different regional blocks; Obtain food output variables from the common expected output variables to construct expected output indicators, select relevant variables from the common non-expected output variables based on the perspective of non-point source pollution and carbon emissions, and use the selected relevant variables and specific variables in the common input variables to retrieve and obtain the participating eco-efficiency evaluation examples; Based on SHAP, the eco-efficiency evaluation example is explained, the characteristic variables involved in the eco-efficiency example are extracted, and the Shapley value of each characteristic variable is generated by SHAP local attribution analysis to characterize the contribution of the characteristic variable, obtain the local explanation of each eco-efficiency example, and construct a local explanation matrix; Each row of the local explanation matrix represents an instance, and each column represents the Shapley value of each feature variable. The average value of the absolute value of the Shapley value of each feature variable is calculated based on the local explanation matrix to generate the global importance of the feature variable; The characteristic variables are grouped according to the non-expected output variables and input variables, and the characteristic variables are sorted in descending order according to the global importance of the characteristic variables in the input variable grouping, a preset number of characteristic variables are obtained to construct input indicators, and non-point source pollution factors and carbon emission factors are constructed according to the global importance of the characteristic variables in the non-expected output variable grouping, and non-expected output indicators are constructed based on the constructed non-point source pollution factors and carbon emission factors; Obtain the specific variables and variable descriptions corresponding to the expected output indicators, non-expected output indicators, and input indicators, build an input-output indicator system, and use the constructed input-output indicator system to determine the indicator parameters in the basic data of grain production.

2. A food production ecological efficiency calculation system according to claim 1, characterized in that: The basic data collection module for grain production is specifically: Obtaining basic data on grain production in the target area, dividing the target area into regional blocks according to preset regional division standards, matching the regional blocks with the basic data on grain production in the region, standardizing the basic data on grain production, using PCA dimensionality reduction processing to reduce data dimensionality, and clustering data according to grain production input conditions and grain production output conditions; The food production input status data set and the food production output status data set are obtained through clustering, and sliding window processing is performed on different data sets to increase the dimensional correlation and time correlation of the basic data of food production in the data set; Variable keyword extraction is used to perform variable extraction in the food production input status data set and the food production output status data set to extract input variables, expected output variables and unexpected output variables.

3. A food production ecological efficiency calculation system according to claim 1, characterized in that: The food production eco-efficiency calculation module is specifically: A food production ecological efficiency calculation model is constructed based on the super efficiency SBM-Undesirable model, and non-point source pollution and carbon emissions are taken as undesirable outputs. The food production ecological efficiency calculation model includes n decision units, each of which is constructed by an expected output index, an undesirable output index, and an input index. The index parameters are imported into the food production ecological efficiency calculation model to calculate and output the food production ecological efficiency of different regional blocks; The food production eco-efficiency calculation model is expressed as: ， in, represents the ecological efficiency of food production, Respectively represent the total number of expected output indicators, undesired output indicators, and input indicators, The number of items representing input indicators, expected output indicators, and undesired output indicators, They represent the input index, expected output and unexpected output index parameters respectively. represents the weight vector, represents the number of decision-making units, represents the input redundant variable, represents the expected output deficit variable, Represents the variable of undesired excess output.

4. A food production ecological efficiency calculation system according to claim 1, characterized in that: The trend prediction module is specifically: Obtaining the food production eco-efficiency sequences of different regional blocks in the target area at a preset time step, spatially annotating the food production eco-efficiency sequences based on the spatial distribution of the different regional blocks, and performing time convolution on the food production eco-efficiency sequences of the different regional blocks through the first time convolution stage; Obtaining examples of food production eco-efficiency driving factor analysis through big data retrieval methods, extracting influencing factors of food production eco-efficiency driving factor analysis examples to construct data samples, evaluating the importance scores of influencing factors based on the mRMR algorithm, and determining the influencing factors of food production eco-efficiency in the target area based on the importance scores; Determine the influencing factor characteristics of different regional blocks according to the influencing factors, calculate the correlation between regional blocks using the influencing factor characteristics, food production ecological efficiency sequence and geographic information, and determine the correlation between regional blocks based on the Pearson correlation coefficient; Constructing a spatial graph of the ecological efficiency of food production in regional blocks, wherein the regional blocks are used as nodes in the spatial graph, the influencing factor characteristics, the ecological efficiency sequence of food production and the geographical information of each regional block are used as node attributes, the edge structure between the nodes is determined according to the association between the regional blocks, and an adjacency matrix is ​​constructed to characterize the spatial graph; The spatial graph is represented and learned by using a graph attention network, and the regional block nodes and the adjacency matrix are introduced into the graph attention network. The attention coefficient of each edge in the spatial graph is calculated by the graph attention mechanism to weight the regional block nodes and extract the spatial characteristics of the ecological efficiency of food production in different regional blocks. The spatial features of the graph attention network output are obtained for calculation in the second time convolution stage, and the outputs of the two time convolution stages are spliced ​​to obtain the time characteristics of the ecological efficiency of food production in different regional blocks in the target area. The first time convolution stage and the second time convolution stage are both composed of a two-dimensional convolution module and a GLU module.

5. A food production ecological efficiency calculation system according to claim 4, characterized in that: The temporal features and spatial features are fused to obtain the spatiotemporal features, specifically: The acquired temporal features and spatial features are concatenated as the input of the Transformer architecture, and the input is embedded. A gated multi-head attention mechanism is introduced into the Transformer architecture, and the linear regression of each attention head in the gated multi-head attention mechanism is replaced by a graph convolution to improve the spatial correlation of the mapped features. The output of the gated multi-head attention mechanism is obtained and concatenated with the original input processed by the activation function, and the time attention representation of each attention head is determined by using an update gate. All attention heads are aggregated to obtain the output of the Transformer architecture, and the spatiotemporal characteristics of the ecological efficiency of food production in different regional blocks in the target area are obtained; The temporal features, spatial features and spatiotemporal features are introduced into a feedforward neural network, and a flattening layer and a linear head are used to obtain a prediction of the ecological efficiency of food production in different regional blocks at a preset future time.

6. A food production ecological efficiency calculation system according to claim 1, characterized in that: The data analysis and visualization module is specifically: According to the food production ecological efficiency of different regional blocks in the target area at the current time stamp, the regional blocks are clustered using the food production ecological efficiency to obtain the horizontal distribution of different food production ecological efficiency levels; The prediction model and observation model are constructed by iterative Kalman filtering algorithm, and the state vector is constructed according to the ecological efficiency of food production in the corresponding regional blocks with the same level distribution. The prediction model is used to track the estimated value corresponding to the level of ecological efficiency of food production at different levels of distribution, and the predicted value of ecological efficiency of food production in each regional block is used for correction in the observation model. The level distribution is updated according to the estimated values ​​corresponding to the levels of ecological efficiency of food production, and the liquidity analysis results are obtained through the differences in level distribution.

7. A method for calculating the ecological efficiency of food production, characterized in that: The food production ecological efficiency calculation system as claimed in any one of claims 1 to 6 comprises the following steps: Construct an input-output indicator system, collect basic data on grain production in different regions of the target area, and calculate indicator parameters based on the input-output system; Using the SBM-Undesirable model to establish a food production ecological efficiency calculation model, importing the indicator parameters into the food production ecological efficiency calculation model, and calculating and outputting the food production ecological efficiency of different regional blocks; Obtain the time series and distribution of food production ecological efficiency in different blocks of the target area, and extract the time series, spatial characteristics and spatiotemporal characteristics of food production ecological efficiency; A prediction model is constructed based on the extracted temporal features, spatial features and spatiotemporal features, and the prediction model is used to predict the ecological efficiency of food production in different regional blocks. The liquidity analysis is performed through the food production ecological efficiency prediction data, and the food production ecological efficiency prediction data and liquidity analysis are visualized.

Images