Intelligent agriculture big data analysis and decision support platform and implementation method thereof

Abstract

The invention relates to the technical field of big data analysis and decision support, in particular to a smart agriculture big data analysis and decision support platform and an implementation method thereof, and the platform comprises an agricultural data collection unit which is used for collecting farmland data and collective fund, asset and resource information in real time; and the agricultural data analysis unit obtains the occurrence rate of diseases and insect pests through a nonlinear algorithm according to the farmland data and the rural collective fund, asset and resource information collected by the agricultural data acquisition unit, and obtains the early warning judgment of the rural collective fund, asset and resource information data through a linear regression algorithm. When a big data analysis and decision platform is researched, a pest and disease occurrence rate model, a crop yield prediction model and a collective fund, asset and resource data judgment model are respectively constructed, and the pest and disease occurrence rate is introduced into the crop yield prediction model to further obtain a crop yield prediction result.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis and decision support, and more specifically, to a big data analysis and decision support platform for smart agriculture and its implementation method. Background Art

[0002] The big data analysis and decision support platform for smart agriculture is a supervision platform based on big data analysis and decision support technology. By collecting, processing, and analyzing agricultural data from multiple sources, it provides intelligent decision support for agricultural production, aiming to optimize the agricultural production process, improve yield and quality, reduce costs, and enhance environmental sustainability.

[0003] However, traditional big data analysis and decision support platforms cannot integrate data from different sources, usually consuming a huge amount of time, and there are no good agricultural data models for pest and disease incidence, crop yield prediction, and judgment of capital, assets, and resource data. In view of this, a big data analysis and decision support platform for smart agriculture and its implementation method are designed. Summary of the Invention

[0004] The purpose of the present invention is to provide a big data analysis and decision support platform for smart agriculture and its implementation method to solve the problems in the above background art that data from different sources cannot be integrated, usually consuming a huge amount of time, and there are no good agricultural data models for pest and disease incidence, crop yield prediction, and judgment of capital, assets, and resource data.

[0005] To achieve the above object, the present invention aims to provide a big data analysis and decision support platform for smart agriculture, including an agricultural data collection unit for collecting farmland data and capital, assets, and resource information in real time; an agricultural data analysis unit that obtains the probability of pest and disease occurrence from the farmland data and capital, assets, and resource information collected by the agricultural data collection unit through a non - linear algorithm, predicts crop yields through a PSO algorithm, and obtains a judgment of capital, assets, and resource data through a linear regression algorithm; a risk warning unit that performs threshold judgment on the pest and disease incidence predicted by the agricultural data analysis unit through a cross - validation algorithm, obtains a crop yield warning for the crop yields predicted by the agricultural data analysis unit through a threshold algorithm, and outputs the decision data of capital, assets, and resources judged by the agricultural data analysis unit.

[0006] As a further improvement of this technical solution, the farmland data collected by the agricultural data collection unit includes remote sensing data; wherein the remote sensing data includes temperature, humidity, geographical location, crop variety, and land area; The capital, asset, and resource information includes the collective funds, assets, and resources invested; The collective funds invested include agricultural production investment, agricultural machinery purchase, and water conservancy facility construction; The assets include fixed land assets and movable unsold agricultural product assets; The resources include land available for agricultural production, water resources for irrigation and aquaculture, and labor resources.

[0007] As a further improvement of this technical solution, the agricultural data analysis unit includes a data processing module, a pest and disease prediction module, a crop yield prediction module, and a capital, asset, and resource data judgment module; Among them, the data processing module constructs a pest and disease incidence model based on a non-linear algorithm, constructs a crop yield prediction model based on the PSO algorithm, and constructs a capital, asset, and resource data judgment model based on a linear regression algorithm; The data processing module integrates an agricultural data model, where the agricultural data model consists of a pest and disease incidence model, a crop yield prediction model, and a capital, asset, and resource data judgment model; The data processing module evaluates and predicts the probability of pest and disease occurrence through the pest and disease incidence model, and the pest and disease prediction module outputs the evaluation and prediction results; The data processing module predicts the crop yield through the crop yield prediction model, and the crop yield prediction module outputs the prediction results; The data processing module realizes the prediction and judgment of capital, assets, and resources through the capital, asset, and resource data judgment model, and the capital, asset, and resource data judgment module outputs the judgment results, where the capital, asset, and resource data judgment module includes a rural collective funds judgment module, a rural collective assets judgment module, and a rural collective resources judgment module; The rural collective funds judgment module obtains the fund health value through the current ratio algorithm for the collective investment fund information data in the capital, asset, and resource information; The rural collective assets judgment module obtains the asset estimated value through the market comparison algorithm for the asset information data in the capital, asset, and resource information; The rural collective resources judgment module obtains the resource utilization rate through remote sensing analysis for the resource information data in the capital, asset, and resource information.

[0008] As a further improvement of this technical solution, the steps for constructing the pest and disease incidence model based on the non-linear algorithm are as follows: S1.1. Collect data from the agricultural data collection unit, including temperature, humidity, crop variety, and geographical location, and successively perform data denoising, filling in missing values, and standardizing the numerical range; ; Among them, represents the training set; represents the th sample input feature vector; represents the th sample's pest and disease label; represents the number of samples S1.2. Construct the kernel function: ; Among them, represents the feature vector of the th sample in the training set; represents a pre-set parameter; S1.3. Maximize the following dual function to obtain the coefficients; ; Among them, ; ; represents the th sample's corresponding Lagrange multiplier; represents the Lagrange multiplier; represents the bias term; represents the regularization parameter; represents the set of Lagrange multipliers; represents the th sample's corresponding Lagrange multiplier; represents the th sample's corresponding pest and disease label; represents the th sample's corresponding pest and disease label; S1.4. Construct the SVM model: ; Among them, represents the number of support vectors; represents the bias term; S1.5. Calculate the prediction probability: ; Among them, represents the input feature of the th sample; represents the probability of the th sample having pests and diseases; represents whether there are pests and diseases.

[0009] As a further improvement of this technical solution, the specific steps for constructing a crop yield prediction model based on the PSO algorithm are as follows: S2.1. Collect data on various factors affecting crop yields from the agricultural data collection unit, including temperature, humidity, and precipitation, and perform necessary data cleaning and preprocessing, filling in missing values, removing outliers, and normalizing the value range; S2.2. Select 3 neurons for the input layer, corresponding to temperature, humidity, and precipitation, select 5 neurons for the hidden layer and use ReLU as the activation function, and select 1 neuron for the output layer to construct an MLP model: S2.3.1. Construct the input layer: ; S2.3.2. Construct the hidden layer, using ReLU as the activation function: ; Among them, ; ; represents the transpose of the weight matrix connecting the input layer and the first hidden layer; represents the bias vector of the first hidden layer; represents the input of the first hidden layer; represents the output vector of the first hidden layer after being processed by the activation function; S2.3.3. Construct the output layer: ; ; Among them, represents the net input vector of the output layer; represents the transpose of the weight matrix connecting the first hidden layer and the output layer; represents the bias term of the output layer; represents the predicted output value of the agricultural product; S2.4. Initialize the particle swarm size: ; S2.5. Define the loss function: ; Among them, represents the actual yield; represents the predicted yield; represents the number of samples; S2.6. Introduce the PSO algorithm to find the parameter combination that minimizes the loss function: ; ; ; ; ; ; wherein, represents the inertia weight; represents the acceleration coefficient; represents the acceleration coefficient; represents a random number; represents a random number; represents the th particle's best position in dimension ; represents the th particle's best position in dimension ; represents the th particle's best position in dimension ; represents the entire population's best position in dimension ; represents the entire population's best position in dimension ; represents the entire population's best position in dimension ; represents the th particle's velocity in dimension at time ; represents the th particle's velocity in dimension at time ; represents the th particle's velocity in dimension at time ; represents the th particle's temperature at time ; represents the th particle's humidity at time ; represents the th particle's precipitation at time ; S2.7. Consider the incidence rate of pests and diseases and further optimize the model: ; wherein, ; represents the original expected output; represents the difference between the actual output and the original expected output; represents the impact degree factor of pests and diseases on the output; represents the error term; S2.8. Obtain the best estimated value by using the least squares method: ; ; wherein, represents the best estimated value of; S2.9. Calculate the adjusted output: ; wherein, ; represents the adjustment coefficient.

[0010] As a further improvement of this technical solution, the method for constructing a capital asset resource data judgment model based on a linear regression algorithm specifically includes the following steps: S3.1. Collect and organize a data set from the agricultural data collection unit, clean the data, process missing values and outliers, perform feature scaling, and select the invested capital asset resources, land area, and labor cost as the data to be input; S3.2. Construct a linear regression model; ; wherein, represents the expected revenue of the predicted agricultural project; represents the invested capital asset resources; represents the land area; represents the labor cost; represents the initial model parameters; represents the model parameters of the invested capital asset resources; represents the model parameters of the land area; represents the model parameters of the labor cost; represents the error term; S3.3. Train the linear regression model and define the loss function; ; wherein, represents the number of samples; represents the th sample's actual revenue; represents the estimated value of the parameter ; Represents the estimated value of the parameter ; Represents the estimated value of the parameter ; Represents the estimated value of the parameter ; Represents the investment capital assets resources corresponding to the th sample; Represents the land area corresponding to the th sample; Represents the labor cost corresponding to the th sample; S3.4. Obtain the optimal parameter estimation by gradient descent: ; Among them, Represents the th parameter; Represents the learning rate; Represents the parameter estimation value; S3.5. Calculate the predicted value: ; S3.6. Make a decision based on the predicted value and the threshold: ; Among them, Represents the set threshold.

[0011] As a further improvement of this technical solution, the risk warning unit includes a pest and disease risk warning module, a crop yield warning module, and a capital assets resources data warning module; Among them, the pest and disease risk warning module performs threshold judgment on the pest and disease incidence rate obtained by the data analysis unit through a cross-validation algorithm; The crop yield warning module realizes crop yield warning through a threshold algorithm; The capital data warning module outputs the decision-making data obtained by the rural collective capital judgment module; The asset data warning module outputs the decision-making data obtained by the rural collective asset judgment module; The resource data warning module outputs the decision-making data obtained by the rural collective resource judgment module.

[0012] As a further improvement of this technical solution, the cross-validation algorithm performs threshold judgment, and the specific steps are as follows: S4.1. Introduce a cost function to optimize the pest and disease occurrence threshold: ; Among them, Represents the false negative cost; Denote the false positive cost; Denote the prediction result based on ; Denote the official label; Denote the candidate set of pest and disease incidence thresholds; S4.2. Evaluate the model performance under different pest and disease incidence thresholds using the cross - validation algorithm: ; Among them, Denote the number of folds of cross - validation; Denote the th cross - validation; Denote the th cross - validation and the th sample; Denote the th cross - validation and the prediction result of the th sample; Denote the th cross - validation and the true label of the th sample; S4.3. Define the parameter space; ; ; Among them, Denote the candidate set of model hyperparameters; S4.4. Calculate the average score of cross - validation using the traversal algorithm; ; Among them, Denote the average score under the combination of the th group of model production parameters and the th threshold ; Denote the currently considered threshold; Denote the set of model production parameters; Denote the number of cross - validations; Denote the th cross - validation iteration.

[0013] As a further improvement of this technical solution, the threshold algorithm realizes crop yield early warning, and its specific steps are as follows: S5.1. Collect the pest and disease incidence and yield change rate of each sample point from the data analysis unit: S5.2. Define the early warning index: ; Among them, Indicates the warning index of the th sample point; S5.3. Calculate the warning threshold: ; ; ; Among them, ; represents the number of samples; represents the warning mean; represents the warning standard deviation; represents the warning threshold; represents whether to issue a warning signal; represents the adjustment coefficient.

[0014] On the other hand, the present invention also provides a method for implementing a smart agriculture big data analysis and decision support platform, which is characterized in that it is applied to the smart agriculture big data analysis and decision support platform as described in any one of the above, and includes the following steps: S6.1. The data collected by the agricultural data collection unit includes farmland data and capital asset resource information: S6.2. The agricultural data analysis unit includes a data processing module, a pest and disease prediction module, a crop yield prediction module, and a capital asset resource data judgment module: The data processing module evaluates and predicts the probability of pest and disease occurrence through a pest and disease incidence model, and the pest and disease prediction module outputs the evaluation and prediction results; The data processing module predicts the crop yield through a crop yield prediction model, and the crop yield prediction module outputs the prediction results; The data processing module realizes the prediction and judgment of capital asset resources through a capital asset resource data judgment model, and the capital asset resource data judgment module outputs the judgment results; S6.3. The risk warning unit includes a pest and disease risk warning module, a crop yield warning module, a capital data warning module, an asset data warning module, and a resource data warning module; Among them, the pest and disease risk warning module uses a cross-validation algorithm to perform threshold judgment on the pest and disease incidence rate obtained by the data analysis unit; The crop yield warning module performs crop yield warning on the crop yield obtained by the crop yield prediction module through a threshold algorithm; The capital data warning module, the asset data warning module, and the resource data warning module output the decision data obtained by the capital asset resource data judgment module.

[0015] Compared with the prior art, the beneficial effects of the present invention: The intelligent agriculture big data analysis and decision support platform and its implementation method, when studying the intelligent agriculture big data analysis and decision platform, respectively construct a pest incidence model, a crop yield prediction model, and a capital asset resource information data judgment model. The probability of pest occurrence is obtained through a non-linear algorithm, the crop yield is predicted through the PSO algorithm, and the pest incidence is introduced into the calculation of crop yield. The capital asset resource information data judgment is obtained through a linear regression algorithm. The pest incidence is judged by a threshold through a cross-validation algorithm, and the crop yield early warning is obtained through a threshold algorithm for the crop yield. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is the overall flow block diagram of the present invention; The meanings of each label in the figure are as follows: 1. Agricultural data collection unit; 2. Agricultural data analysis unit; 21. Data processing module; 22. Pest prediction module; 23. Crop yield prediction module; 24. Capital asset resource data judgment module; 3. Risk early warning unit; 31. Pest risk early warning module; 32. Crop yield early warning module; 33. Capital data early warning module; 34. Asset data early warning module; 35. Resource data early warning module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0018] Embodiment 1 Please refer to Figure 1 As shown, an intelligent agriculture big data analysis and decision support platform is provided, including an agricultural data collection unit 1 for collecting farmland data and capital asset resource information in real time; In this embodiment, an intelligent agriculture big data analysis and decision support platform further includes an agricultural data analysis unit 2. The agricultural data analysis unit 2 obtains the probability of pest occurrence from the farmland data and capital asset resource information collected by the agricultural data collection unit 1 through a non-linear algorithm, predicts the crop yield through the PSO algorithm, and obtains the capital asset resource data judgment through a linear regression algorithm; In this embodiment, a big data analysis and decision support platform for smart agriculture further includes a risk warning unit 3. The risk warning unit 3 performs threshold judgment on the pest and disease incidence rate predicted by the agricultural data analysis unit 2 through a cross-validation algorithm, obtains a crop yield warning for the crop yield predicted by the agricultural data analysis unit 2 through a threshold algorithm, and outputs the decision data of the capital asset resources judged by the agricultural data analysis unit 2.

[0019] In this embodiment, the farmland data collected by the agricultural data collection unit 1 includes remote sensing data; Among them, the remote sensing data includes temperature, humidity, geographical location, crop types, and land area; the capital asset resource information includes invested collective funds, assets, and resources; The collective investment fund information data includes agricultural production investment, agricultural machinery purchase, and water conservancy facility construction; The asset information data includes fixed land assets and mobile unsold agricultural product assets; The resource information data includes land available for agricultural production, water resources for irrigation and aquaculture, and labor resources.

[0020] In this embodiment, the agricultural data analysis unit 2 includes a data processing module 21, a pest and disease prediction module 22, a crop yield prediction module 23, and a capital asset resource data judgment module 24; Among them, the data processing module 21 constructs a pest and disease incidence rate model based on a non-linear algorithm, constructs a crop yield prediction model based on a PSO algorithm, and constructs a capital asset resource data judgment model based on a linear regression algorithm; The data processing module 21 integrates an agricultural data model, where the agricultural data model is composed of a pest and disease incidence rate model, a crop yield prediction model, and a capital asset resource data judgment model; The data processing module 21 evaluates and predicts the probability of pest and disease occurrence through the pest and disease incidence rate model, and outputs the evaluation and prediction results by the pest and disease prediction module 22; The data processing module 21 predicts the crop yield through the crop yield prediction model, and outputs the prediction results by the crop yield prediction module 23; The data processing module 21 realizes the prediction and judgment of capital asset resources through the capital asset resource data judgment model, and outputs the judgment results by the capital asset resource data judgment module 24, where the capital asset resource data judgment module 24 includes a rural collective fund judgment module 241, a rural collective asset judgment module 242, and a rural collective resource judgment module 243; The rural collective fund judgment module 241 obtains a fund health value for the collective investment fund information data in the capital asset resource information through a current ratio algorithm; The rural collective asset judgment module 242 obtains the asset estimated value from the asset information data in the capital, asset, and resource information through the market comparison algorithm; The rural collective resource judgment module 243 obtains the resource utilization rate from the resource information data in the capital, asset, and resource information through remote sensing analysis.

[0021] In this embodiment, a pest incidence model is constructed based on a non-linear algorithm, and the specific steps are as follows: S1.1. Collect data from the agricultural data collection unit 1, including temperature, humidity, crop type, and geographical location, and perform data denoising, filling in missing values, and normalizing the value range in sequence; ; Among them, represents the training set; represents the th sample input feature vector; represents the th sample pest infestation label; represents the number of samples S1.2. Construct a kernel function: ; Among them, represents the feature vector of the th sample in the training set; represents a preset parameter; S1.3. Maximize the following dual function to obtain the coefficients; ; Among them, ; ; represents the th Lagrange multiplier corresponding to the sample; represents the Lagrange multiplier; represents the bias term; represents the regularization parameter; represents the set of Lagrange multipliers; represents the th Lagrange multiplier corresponding to the sample; represents the th sample pest infestation label; represents the th sample pest infestation label; S1.4. Construct an SVM model: ; Among them, represents the number of support vectors; represents the bias term; S1.5. Calculate the prediction probability: ; Among them, represents the input feature of the th sample; represents the probability of the th sample having pests and diseases; represents whether there are pests and diseases.

[0022] In this embodiment, a crop yield prediction model is constructed based on the PSO algorithm, and the specific steps are as follows: S2.1. Collect data on various factors affecting crop yield from the agricultural data collection unit 1, including temperature, humidity, and precipitation, and perform necessary data cleaning and preprocessing, filling in missing values, removing outliers, and normalizing the value range; Select 3 neurons for the input layer, corresponding to temperature, humidity, and precipitation, select 5 neurons for the hidden layer and use ReLU as the activation function, and select 1 neuron for the output layer to construct an MLP model: S2.3.1. Construct the input layer: ; Among them, represents the temperature parameter set; represents the humidity parameter set; represents the precipitation parameter set; S2.3.2. Construct the hidden layer, using ReLU as the activation function: ; Among them, ; ; represents the transpose of the weight matrix connecting the input layer and the first hidden layer; represents the bias vector of the first hidden layer; represents the input of the first hidden layer; represents the output vector of the first hidden layer after being processed by the activation function; S2.3.3. Construct the output layer: ; ; Among them, represents the net input vector of the output layer; represents the transpose of the weight matrix connecting the first hidden layer and the output layer; represents the bias term of the output layer; represents the predicted output value of agricultural products; S2.4. Initialize the particle swarm size: ; S2.5. Define the loss function: ; where, represents the actual output; represents the predicted output; represents the number of samples; S2.6. Introduce the PSO algorithm to find the parameter combination that minimizes the loss function: ; ; ; ; ; ; where, represents the inertia weight; represents the acceleration coefficient; represents the acceleration coefficient; represents a random number; represents a random number; represents the best position of the th particle in dimension ; represents the best position of the th particle in dimension ; represents the best position of the th particle in dimension ; represents the best position of the entire population in dimension ; represents the best position of the entire population in dimension ; represents the best position of the entire population in dimension ; represents the velocity of the th particle in dimension at time ; represents the velocity of the th particle in dimension at time ; Indicates the velocity of the th particle at time in dimension Indicates the temperature of the th particle at time Indicates the humidity of the th particle at time Indicates the precipitation of the th particle at time S2.7. Consider the incidence of pests and diseases and further optimize the model: ; Among them, ; represents the original expected yield; represents the difference between the actual yield and the original expected yield; represents the impact degree factor of pests and diseases on the yield; represents the error term; S2.8. Obtain the best estimate value by the least squares method: ; Among them, represents the best estimate value; S2.9. Calculate the adjusted yield: ; Among them, ; represents the adjustment coefficient.

[0023] Since pests and diseases are one of the important factors affecting crop yields, especially in the case of severe outbreaks, they may lead to a significant reduction in yields. Therefore, incorporating the incidence of pests and diseases as an influencing factor into the prediction model of crop yields can help farmers plan agricultural activities more effectively. Through prediction, high-risk areas can be identified in advance and preventive measures can be taken in a timely manner.

[0024] In this embodiment, a judgment model for financial asset resource data is constructed based on a linear regression algorithm, and the specific steps are as follows: S3.1. Collect and organize the data set from the agricultural data collection unit 1, clean the data, handle missing values and outliers, perform feature scaling, and select the invested financial asset resources, land area, and labor cost as the data to be input; S3.2. Construct a linear regression model; ; Among them, represents the expected return of the predicted agricultural project; represents the invested capital, assets, and resources; represents the land area; represents the labor cost; represents the initial model parameters; represents the model parameters of the invested capital, assets, and resources; represents the model parameters of the land area; represents the model parameters of the labor cost; represents the error term; S3.3. Train the linear regression model and define the loss function; ; Among them, represents the number of samples; represents the actual return of the th sample; represents the estimated value of the parameter ; represents the estimated value of the parameter ; represents the estimated value of the parameter ; represents the estimated value of the parameter ; represents the invested capital, assets, and resources corresponding to the th sample; represents the land area corresponding to the th sample; represents the labor cost corresponding to the th sample; ; Among them, represents the th parameter; represents the learning rate; represents the estimated value of the parameter; S3.5. Calculate the predicted value: ; S3.6. Make a decision based on the predicted value and the threshold: ; Among them, represents the set threshold.

[0025] In this embodiment, the risk warning unit 3 includes a pest and disease risk warning module 31, a crop yield warning module 32, and a fund data warning module 33; Among them, the pest and disease risk warning module 31 uses a cross-validation algorithm to perform threshold judgment on the pest and disease incidence rate obtained by the data analysis unit 2; The crop yield warning module 32 realizes crop yield warning through a threshold algorithm; The fund data warning module 33 outputs the decision data obtained by the fund asset resource data judgment module 24.

[0026] In this embodiment, the cross-validation algorithm performs threshold judgment, and the specific steps are as follows: S4.1. Introduce a cost function to optimize the pest and disease occurrence threshold: ; Among them, represents the false negative cost; represents the false positive cost; represents based on the prediction result of; represents the official label; represents the candidate set of pest and disease incidence rate thresholds; S4.2. Use the cross-validation algorithm to evaluate the model performance under different pest and disease incidence rate thresholds: ; Among them, represents the number of folds of cross-validation; represents the th cross-validation; represents the th sample in the th cross-validation; represents the th prediction result of the th sample in the th cross-validation; represents the th true label of the th sample in the th cross-validation; S4.3. Define the parameter space; ; ; Among them, represents the candidate set of model hyperparameters; S4.4. Use the traversal algorithm to calculate the average score of cross-validation; ; Among them, represents the Group model production parameters and the th threshold The average score under the combination; Indicates the currently considered threshold; Indicates the set of model production parameters; Indicates the number of cross-validations; Indicates the th cross-validation iteration.

[0027] In this embodiment, the threshold algorithm realizes crop yield early warning, and its specific steps are as follows: S5.1. Collect the incidence of pests and diseases and the yield change rate of each sample point from the data analysis unit 2: S5.2. Define the early warning index: ; Among them, Indicates the th early warning index of the sample point; S5.3. Calculate the early warning threshold: ; ; ; Among them, ; Indicates the number of samples; Indicates the early warning mean; Indicates the early warning standard deviation; Indicates the early warning threshold; Indicates whether to issue an early warning signal; Indicates the adjustment coefficient.

[0028] Embodiment 2 The present invention provides a method for implementing a smart agriculture big data analysis and decision support platform, which is applied to the smart agriculture big data analysis and decision support platform described in any one of the above, and includes the following steps: S6.1. The data collected by the agricultural data collection unit 1 includes farmland data and financial asset resource information: S6.2. The agricultural data analysis unit 2 includes a data processing module 21, a pest and disease prediction module 22, a crop yield prediction module 23, and a capital, asset, and resource data judgment module 24. Among them, the data processing module 21 evaluates and predicts the probability of pest and disease occurrence through a pest and disease incidence model, and the pest and disease prediction module 22 outputs the evaluation and prediction results; the data processing module 21 predicts the crop yield through a crop yield prediction model, and the crop yield prediction module 22 outputs the prediction results. The data processing module 21 realizes the prediction and judgment of capital, assets, and resources through a capital, asset, and resource data judgment model, and the capital, asset, and resource data judgment module 24 outputs the judgment results. S6.3. The risk warning unit 3 includes a pest and disease risk warning module 31, a crop yield warning module 32, a capital data warning module 33, an asset data warning module 34, and a resource data warning module 35. Among them, the pest and disease risk warning module 31 performs threshold judgment on the pest and disease incidence rate obtained by the data analysis unit 2 through a cross-validation algorithm; the crop yield warning module 32 performs crop yield warning on the crop yield obtained by the crop yield prediction module 23 through a threshold algorithm; the capital data warning module 33, the asset data warning module 34, and the resource data warning module 35 output the decision-making data obtained by the capital, asset, and resource data judgment module 24.

[0029] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. A smart agriculture big data analysis and decision support platform, characterized by: include An agricultural data collection unit (1), the agricultural data collection unit (1) being used to collect farmland data and financial asset resource information in real time; An agricultural data analysis unit (2), wherein the agricultural data analysis unit (2) uses a nonlinear algorithm to obtain the probability of occurrence of pests and diseases from the farmland data and the capital asset resource information collected by the agricultural data collection unit (1), predicts crop yields using a PSO algorithm, and obtains capital asset resource data judgment using a linear regression algorithm; A risk warning unit (3) is provided, wherein the risk warning unit (3) performs threshold judgment on the incidence of pests and diseases predicted by the agricultural data analysis unit (2) through a cross-validation algorithm, obtains crop yield warnings through a threshold algorithm based on the crop yield predicted by the agricultural data analysis unit (2), and outputs the capital asset resource decision data determined by the agricultural data analysis unit (2).

2. The smart agriculture big data analysis and decision support platform according to claim 1, characterized in that: The farmland data collected by the agricultural data collection unit (1) includes remote sensing data; Among them, remote sensing data include temperature, humidity, geographic location, crop types and land area; Information on funds, assets and resources includes invested collective funds, assets and resources; The collective funds invested include investment in agricultural production, purchase of agricultural machinery and construction of water conservancy facilities; The assets mentioned include fixed land assets and current unsold agricultural products assets; The resources include land that can be used for agricultural production, water resources for irrigation and breeding, and labor resources.

3. The smart agriculture big data analysis and decision support platform according to claim 1, characterized in that: The agricultural data analysis unit (2) comprises a data processing module (21), a pest and disease prediction module (22), a crop yield prediction module (23) and a capital asset resource data judgment module (24); The data processing module (21) constructs a pest incidence model based on a nonlinear algorithm, a crop yield prediction model based on a PSO algorithm, and a capital asset resource data judgment model based on a linear regression algorithm; The data processing module (21) is integrated with an agricultural data model, wherein the agricultural data model is composed of a pest incidence model, a crop yield prediction model, and a capital asset resource data judgment model; The data processing module (21) evaluates and predicts the probability of occurrence of pests and diseases through a pest and disease incidence model, and the pest and disease prediction module (22) outputs the evaluation and prediction results; The data processing module (21) predicts the crop yield using a crop yield prediction model, and the crop yield prediction module (23) outputs the prediction result; The data processing module (21) implements forecasting and judging of funds, assets and resources through a funds, assets and resources data judging model, and outputs the judgment result through a funds, assets and resources data judging module (24), wherein the funds, assets and resources data judging module (24) includes a rural collective funds judging module (241), a rural collective assets judging module (242) and a rural collective resources judging module (243); The rural collective capital judgment module (241) obtains the capital health value by using the current ratio algorithm to obtain the collective investment capital information data in the capital asset resource information; The rural collective asset judgment module (242) obtains an estimated asset value by using a market comparison algorithm to obtain asset information data in the fund asset resource information; The rural collective resource judgment module (243) obtains resource utilization rate by remote sensing analysis of resource information data in the capital asset resource information.

4. The smart agriculture big data analysis and decision support platform according to claim 3 is characterized by: The specific steps of constructing the pest incidence model based on the nonlinear algorithm are as follows: S1.

1. Collect data from the agricultural data collection unit (1), including temperature, humidity, crop type and geographic location, and then perform data denoising, fill in missing values ​​and standardize the value range; ； in, represents the training set; Indicates Sample input feature vector; Indicates The samples have pest and disease labels; represents the number of samples; S1.

2. Construct kernel function: ； in, Indicates the training set The feature vector of samples; Indicates pre-set parameters; S1.3, maximize the following dual function to obtain the coefficients; ； in, ； ； Indicates The Lagrange multiplier corresponding to the samples; represents the Lagrange multiplier; represents the bias term; represents the regularization parameter; represents the set of Lagrange multipliers; Indicates The Lagrange multiplier corresponding to the samples; Indicates The presence or absence of pests and diseases labels corresponding to each sample; Indicates The presence or absence of pests and diseases labels corresponding to each sample; S1.

4. Constructing SVM model: ； in, represents the number of support vectors; represents the bias term; S1.

5. Calculate the predicted probability: ； in, Indicates Input features of samples; Indicates The probability that a sample has pests and diseases; Indicates whether there are pests and diseases.

5. The smart agriculture big data analysis and decision support platform according to claim 3 is characterized by: The crop yield prediction model is constructed based on the PSO algorithm, and the specific steps are as follows: S2.

1. Collect data on various factors affecting crop yields, including temperature, humidity, and precipitation, from the agricultural data collection unit (1), and perform necessary data cleaning and preprocessing to fill in missing values, remove outliers, and standardize the value range; S2.2, select 3 neurons in the input layer, corresponding to temperature, humidity, and precipitation, select 5 neurons in the hidden layer and use ReLU as the activation function, select 1 neuron in the output layer, and build the MLP model: S2.3.

1. Construct the input layer: ； in, Represents a set of temperature parameters; Represents a set of humidity parameters; represents a set of precipitation parameters; S2.3.

2. Construct the hidden layer and use ReLU as the activation function: ； in, ； ； represents the transpose of the weight matrix connecting the input layer to the first hidden layer; represents the bias vector of the first hidden layer; represents the input of the first hidden layer; Represents the output vector of the first hidden layer after being processed by the activation function; S2.3.

3. Construct the output layer: ； ； in, represents the net input vector of the output layer; represents the transpose of the weight matrix connecting the first hidden layer to the output layer; Represents the bias term of the output layer; represents the predicted output value of agricultural products; S2.4, Initialize particle swarm size: ； S2.

5. Define the loss function: ； in, Indicates actual output; represents the predicted output; represents the number of samples; S2.

6. Introduce the PSO algorithm to find the parameter combination that minimizes the loss function: ； ； ； ； ； ； in, represents the inertia weight; represents the acceleration factor; represents the acceleration factor; Represents a random number; Represents a random number; Indicates Particles in dimension The best position on Indicates Particles in dimension The best position on Indicates Particles in dimension The best position on Indicates that the entire group is in dimension The best position on Indicates that the entire group is in dimension The best position on Indicates that the entire group is in dimension The best position on Indicates Particles in time Time in Dimension Speed ​​on; Indicates Particles in time Time in Dimension Speed ​​on; Indicates Particles in time Time in Dimension Speed ​​on; Indicates Particles in time The temperature from time to time; Indicates Particles in time humidity at the time; Indicates Particles in time The amount of precipitation at that time; S2.

7. Consider the incidence of pests and diseases and further optimize the model: ； in, ； represents the original expected output; It represents the difference between actual output and original expected output; A factor indicating the degree of impact of pests and diseases on yield; represents the error term; S2.8, using the least squares method to obtain Best estimate of: ； in, express The best estimate of ; S2.

9. Calculate adjusted production: ； in, ； Represents the adjustment factor.

6. The smart agriculture big data analysis and decision support platform according to claim 3, characterized in that: The specific steps of constructing the fund asset resource data judgment model based on the linear regression algorithm are as follows: S3.

1. Collect and organize data sets from the agricultural data collection unit (1), clean the data, handle missing values ​​and outliers, scale features, and select input capital asset resources, land area, and labor costs as input data; S3.2, build a linear regression model; ； in, represents the expected benefits of the predicted agricultural project; It represents the investment of capital, assets and resources; Indicates land area; represents labor cost; represents the initial model parameters; Model parameters representing invested capital asset resources; Model parameter representing land area; Model parameters representing labor costs; represents the error term; S3.3, train the linear regression model and define the loss function; ； in, represents the number of samples; Indicates The actual income of the samples; Representation parameters An estimated value of Representation parameters An estimated value of Representation parameters An estimated value of Representation parameters An estimated value of Indicates The investment capital and asset resources corresponding to each sample; Indicates The land area corresponding to each sample; Indicates The labor cost corresponding to each sample; S3.

4. Use gradient descent to obtain the best parameter estimate: ； in, Indicates parameters; represents the learning rate; represents parameter estimates; S3.

5. Calculate the predicted value: ； S3.

6. Make decisions based on predicted values ​​and thresholds: ； in, Indicates the set threshold.

7. The smart agriculture big data analysis and decision support platform according to claim 1, characterized in that: The risk warning unit (3) comprises a pest risk warning module (31), a crop yield warning module (32), a capital data warning module (33), an asset data warning module (34) and a resource data warning module (35); The pest risk warning module (31) uses a cross-validation algorithm to perform threshold judgment on the pest incidence rate obtained by the data analysis unit (2); The crop yield early warning module (32) realizes crop yield early warning through a threshold algorithm; The fund data early warning module (33) outputs the decision data obtained by the rural collective fund judgment module (241); The asset data early warning module (34) outputs the decision data obtained by the rural collective asset judgment module (242); The resource data early warning module (35) outputs the decision data obtained by the rural collective resource judgment module (243).

8. The smart agriculture big data analysis and decision support platform according to claim 7, characterized in that: The cross-validation algorithm performs threshold judgment, and its specific steps are as follows: S4.

1. Introduce cost function to optimize the threshold of pest occurrence: ； in, represents the false negative cost; represents the false positive cost; Indicates based on The prediction results; Indicates the official label; represents the candidate set of thresholds for the incidence of pests and diseases; S4.

2. Use cross-validation algorithm to evaluate model performance under different pest and disease incidence thresholds: ； in, represents the number of folds for cross validation; Indicates Cross validation; Indicates In the cross validation samples; Indicates In the cross validation The prediction results of samples; Indicates In the cross validation The true labels of samples; S4.3, define parameter space; ； ； in, Represents the candidate set of model hyperparameters; S4.4, use the traversal algorithm to calculate the average score of cross validation; ； in, Indicates Group model parameters and Threshold The average score under the combination; Indicates the threshold currently under consideration; Represents the set of model parameters; represents the number of cross validations; Indicates cross validation iterations.

9. The smart agriculture big data analysis and decision support platform according to claim 7, characterized in that: The threshold algorithm realizes crop yield early warning, and its specific steps are as follows: S5.

1. Collect the pest and disease incidence and yield change rate of each sample point from the data analysis unit (2): S5.

2. Define early warning indicators: ； in, Indicates Early warning indicators for each sample point; S5.

3. Calculate the warning threshold: ； ； ； in, ； represents the number of samples; represents the warning mean; represents the warning standard deviation; Indicates the warning threshold; Indicates whether a warning signal is issued; Represents the adjustment factor.

10. A method for implementing a smart agriculture big data analysis and decision support platform, characterized in that: The method is applied to the smart agriculture big data analysis and decision support platform as claimed in any one of claims 1 to 9, comprising the following steps: S6.

1. Agricultural data collection unit (1) The data collected include farmland data and financial asset resource information: S6.2, the agricultural data analysis unit (2) comprises a data processing module (21), a pest and disease prediction module (22), a crop yield prediction module (23) and a capital asset resource data judgment module (24); wherein the data processing module (21) evaluates and predicts the probability of pest and disease occurrence through a pest and disease incidence model, and the pest and disease prediction module (22) outputs the evaluation and prediction results; the data processing module (21) predicts the crop yield through a crop yield prediction model, and the crop yield prediction module (22) outputs the prediction results; The data processing module (21) realizes the prediction and judgment of capital assets and resources through the capital assets and resources data judgment model, and the capital assets and resources data judgment module (24) outputs the judgment result; S6.3, the risk warning unit (3) includes a pest risk warning module (31), a crop yield warning module (32), a capital data warning module (33), an asset data warning module (34) and a resource data warning module (35); wherein the pest risk warning module (31) uses a cross-validation algorithm to perform a threshold judgment on the pest incidence rate obtained by the data analysis unit (2); the crop yield warning module (32) uses a threshold algorithm to perform a crop yield warning on the crop yield obtained by the crop yield prediction module (23); and the capital data warning module (33), the asset data warning module (34) and the resource data warning module (35) output the decision data obtained by the capital asset resource data judgment module (24).

Images