Wheat scab grade prediction method and system based on grid search PCA-SVR-RF model
By using grid search to optimize the PCA-SVR-RF model in wheat gibberellosis level prediction, the problems of high-dimensional data processing and model optimization are solved, and efficient monitoring and accurate prediction of wheat gibberellosis level are achieved.
Patent Information
- Application Number
- CN202510039735.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-10
- Publication Date
- 2025-06-10
AI Technical Summary
The prior art has problems such as high-dimensional data processing, feature selection and model optimization in the prediction of wheat gibberellia rating, resulting in limited prediction accuracy and generalization capabilities.
The PCA-SVR-RF model based on grid search is adopted to reduce the dimensionality through principal component analysis, and the worst and best prediction models are selected, and the second-level prediction model is optimized by fusion data to improve prediction accuracy.
Through dimensionality reduction and model optimization, the accuracy and generalization ability of wheat gibberellia grade prediction are improved, and efficient monitoring and accurate prediction of wheat gibberellia grades within the regional range are achieved.
Smart Images
Figure CN120123962A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of agricultural pest and disease prediction, and particularly to a method and system for predicting the grade of wheat scab based on a grid search PCA-SVR-RF model. Background Technique
[0002] Wheat scab (Fusarium head blight, abbreviated as FHB) is a disease that seriously affects the yield and quality of wheat. It mainly damages the ears of wheat, resulting in ear death, decreased grain weight, and reduced yield. Since the occurrence of scab is closely related to meteorological conditions, it is crucial to accurately predict the occurrence and development of scab. In agricultural production, timely and accurate prediction of the grade of scab not only helps in disease control but also provides a basis for crop growth monitoring and scientific management.
[0003] Traditional methods for predicting wheat scab mainly rely on meteorological monitoring and manual judgment. These methods usually require a large amount of time and manpower input, and due to the strong subjectivity of manual judgment, the prediction results may be inaccurate. In addition, traditional statistical regression methods and machine learning models have problems such as difficulty in feature selection and data redundancy when dealing with multi-dimensional meteorological data, resulting in limited prediction accuracy and generalization ability.
[0004] In recent years, with the development of machine learning and big data technologies, disease prediction methods based on multi-source data have made certain progress. However, existing technologies lack theoretical discussions on issues such as high-dimensional data processing, feature selection, and model optimization in predicting the grade of wheat scab.
[0005] Therefore, there is an urgent need to provide a technical solution to address the deficiencies of the above-mentioned existing technologies. Summary of the Invention
[0006] The purpose of this application is to provide a method and system for predicting the grade of wheat scab based on a grid search PCA-SVR-RF model to solve or alleviate the problems existing in the above-mentioned existing technologies.
[0007] To achieve the above purpose, this application provides the following technical solutions:
[0008] This application provides a method for predicting the grade of wheat scab based on a grid search PCA-SVR-RF model, including: Step S101, performing principal component analysis on Q meteorological factors in P regions obtained to obtain N meteorological principal components after dimensionality reduction; where P, Q, and N are all positive integers;
[0009] Step S102: According to the obtained N meteorological principal components and the obtained grade sample data of wheat scab in P regions, screen the pre-constructed wheat scab grade prediction models with M different network architectures to determine the first grade prediction model and the second grade prediction model; where M is a positive integer, the first grade prediction model is the worst model for wheat scab grade prediction, and the second grade prediction model is the best model for wheat scab grade prediction.
[0010] Step S103: Optimize the second grade prediction model according to the fusion data obtained by fusing the prediction result of the first grade prediction model with the grade sample data, and predict the wheat scab grade corresponding to the N meteorological principal components in each region through the optimized second grade prediction model.
[0011] Preferably, in step S101, calculate the covariance matrix of the standardized data of the Q meteorological factors; perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalues of N' principal components; where N'≥N, and N' is a positive integer; determine the N meteorological principal components according to the magnitudes of the eigenvalues of the N' principal components.
[0012] Preferably, in step S102, according to the obtained N meteorological principal components and the obtained grade sample data of wheat scab in P regions, generate M regression relationships between the wheat scab grade and the N meteorological principal components based on the pre-constructed wheat scab grade prediction models with M different network architectures.
[0013] Based on the M regression relationships, predict the wheat scab grade corresponding to the N meteorological principal components in each region to obtain M prediction result data sets; where each prediction result data set contains P wheat scab grade prediction values.
[0014] According to the M prediction result data sets, screen the wheat scab grade prediction models with M different network architectures to determine the first grade prediction model and the second grade prediction model.
[0015] Preferably, in step S103, fuse the prediction result of the first grade prediction model with the N meteorological principal components to obtain the fusion data; based on the fusion data, optimize the number of parameter decision trees and the maximum depth of the tree of the second grade prediction model by using the grid search algorithm.
[0016] Preferably, the first grade prediction model is a grade prediction model based on support vector regression; and / or, the second grade prediction model is a grade prediction model based on random forest.
[0017] Preferably, the meteorological data of the key growth stages of wheat in P regions are obtained as the meteorological factors.
[0018] The embodiment of the present application also provides a wheat scab grade prediction system based on a grid search PCA-SVR-RF model, including: a PCA dimensionality reduction unit configured to perform principal component analysis on Q meteorological factors of P regions obtained to obtain N meteorological principal components after dimensionality reduction; wherein, P, Q, and N are all positive integers;
[0019] A model screening unit configured to screen M pre-constructed wheat scab grade prediction models with different network architectures according to the obtained N meteorological principal components and the obtained grade sample data of wheat scab in P regions to determine a first grade prediction model and a second grade prediction model; wherein, M is a positive integer, the first grade prediction model is the worst model for wheat scab grade prediction, and the second grade prediction model is the best model for wheat scab grade prediction;
[0020] A model optimization unit configured to optimize the second grade prediction model according to the fusion data obtained by fusing the prediction result of the first grade prediction model with the grade sample data, and predict the wheat scab grade corresponding to the N meteorological principal components in each region through the optimized second grade prediction model.
[0021] Beneficial effects:
[0022] In the wheat scab grade prediction method based on the grid search PCA-SVR-RF model provided by the embodiment of the present application, first, principal component analysis is performed on Q meteorological factors of P regions obtained to obtain N meteorological principal components after dimensionality reduction; then, according to the obtained N meteorological principal components and the obtained grade sample data of wheat scab in P regions, M pre-constructed wheat scab grade prediction models with different network architectures are screened to determine the first grade prediction model with the worst wheat scab grade prediction and the second grade prediction model with the best wheat scab pre-grade prediction; finally, according to the fusion data obtained by fusing the prediction result of the first grade prediction model with the grade sample data, the second grade prediction model is optimized, and the wheat scab grade corresponding to the N meteorological principal components in each region is predicted through the optimized second grade prediction model.
[0023] Thus, by performing principal component analysis on meteorological factors to reduce the dimensionality, representative meteorological principal components are obtained, effectively reducing redundant information and improving the efficiency of data processing and the accuracy of model training. By fusing the prediction results of the worst prediction model with the rank sample data, the feature data is expanded, enhancing the diversity of the input data of the second-rank prediction model, effectively improving the accuracy and generalization ability of wheat scab rank prediction, realizing the efficient monitoring and accurate prediction of wheat scab ranks within a regional scope, being highly applicable in large-scale agricultural production, providing technical support for the early warning and scientific prevention of wheat scab, and providing real-time and effective disease warning information for agricultural management. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] The accompanying drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application.
[0025] Wherein:
[0026] Figure 1 is a schematic flow chart of a method for predicting wheat scab ranks based on a grid search PCA-SVR-RF model provided by some embodiments of this application;
[0027] Figure 2 is a schematic framework diagram of a grid search PCA-SVR-RF model provided by some embodiments of this application;
[0028] Figure 3 is a schematic distribution diagram of the predicted values and actual values of a rank prediction model based on support vector regression provided by some embodiments of this application;
[0029] Figure 4 is a schematic distribution diagram of the predicted values and actual values of a rank prediction model based on KNN provided by some embodiments of this application;
[0030] Figure 5 is a schematic distribution diagram of the predicted values and actual values of a rank prediction model based on bayesian provided by some embodiments of this application;
[0031] Figure 6 is a schematic distribution diagram of the predicted values and actual values of a rank prediction model based on GBDT provided by some embodiments of this application;
[0032] Figure 7 is a schematic distribution diagram of the predicted values and actual values of a rank prediction model based on random forest provided by some embodiments of this application;
[0033] Figure 8Schematic diagram of parameter optimization of a random forest-based grade prediction model provided according to some embodiments of the present application;
[0034] Figure 9 Schematic diagram of the structure of a wheat scab grade prediction system based on a grid search PCA-SVR-RF model provided according to some embodiments of the present application. Detailed implementation manners
[0035] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments. Each example is provided by way of explanation of the present application rather than limitation of the present application. In fact, those skilled in the art will clearly understand that modifications and variations can be made to the present application without departing from the scope or spirit of the present application. For example, features shown or described as part of one embodiment can be used in another embodiment to yield yet another embodiment. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the embodiments of the present invention shall fall within the scope protected by the embodiments of the present invention.
[0036] Wheat scab (Fusarium head blight, abbreviated as FHB) is a disease that seriously affects the yield and quality of wheat, mainly damaging the ears of wheat, resulting in ear death, reduced grain weight, and decreased yield. Since the occurrence of scab is closely related to meteorological conditions, it is crucial to accurately predict the occurrence and development of wheat scab. Traditional wheat scab prediction mainly relies on meteorological monitoring and manual determination. These methods usually require a large amount of time and manpower, and due to the strong subjectivity of manual determination, the prediction results may be inaccurate. In addition, traditional statistical regression methods and machine learning models have difficulties in feature selection and data redundancy problems when dealing with multi-dimensional meteorological data, resulting in limited prediction accuracy and generalization ability of the models.
[0037] With the development of machine learning and big data technologies, great progress has been made in disease prediction methods based on multi-source data. However, there is still a lack of theoretical discussion on problems such as high-dimensional data processing, feature selection, and model optimization in wheat scab grade prediction in the prior art.
[0038] Based on this, the embodiments of the present application propose a wheat scab grade prediction method based on a grid search PCA-SVR-RF model, as Figures 1 to 8 shown, the prediction method includes:
[0039] Step S101: Perform principal component analysis on Q meteorological factors in P regions obtained to obtain N meteorological principal components after dimensionality reduction.
[0040] In this application, the meteorological data of the key growth stages of wheat in P (P is a positive integer) regions are used as meteorological factors. Specifically, wheat scab mainly occurs during the flowering and filling stages of wheat. The incidence of wheat scab is closely related to environmental conditions (especially meteorological factors such as temperature and humidity). During the flowering and filling stages of wheat, it is the most sensitive stage to the occurrence of diseases in the growth period of wheat. At this time, the ears of wheat are most likely to be infected by the pathogen of wheat scab. Therefore, the meteorological data of this period are selected as the main data set for data analysis.
[0041] Among them, the meteorological data include but are not limited to temperature, humidity, precipitation, wind speed, etc.; after obtaining the meteorological data, dimensionality reduction processing is performed on the meteorological data based on the principal component analysis (PCA) method to effectively reduce the redundant information in the obtained original data and retain the main features in the meteorological data, providing a simplified and efficient data input for the machine learning model.
[0042] In this application, when performing principal component analysis on Q (Q is a positive integer) meteorological factors in P regions to obtain N dimensionality-reduced meteorological principal components, first, the obtained Q meteorological factors are standardized (with a mean of 0 and a standard deviation of 1) to unify the characteristic dimensions of different meteorological factors; then, the covariance matrix of the standardized data is calculated to measure the linear relationship between different meteorological variables; then, eigenvalue decomposition is performed on the covariance matrix to obtain the eigenvalues of N′ (N′ is a positive integer) principal components among the Q meteorological factors; finally, according to the magnitudes of the eigenvalues of the N′ principal components, the first N (N′≥N) principal components are selected as meteorological principal components to retain the most important meteorological factors.
[0043] Step S102: According to the obtained N meteorological principal components and the grade sample data of wheat scab in P regions, screen the grade prediction models of wheat scab with M different network architectures, and determine the first grade prediction model and the second grade prediction model.
[0044] Among them, M is a positive integer. The first grade prediction model is the worst model for predicting the grade of wheat scab, and the second grade prediction model is the best model for predicting the grade of wheat scab.
[0045] In this application, the N meteorological principal components obtained after dimensionality reduction are combined with the grade sample data of wheat scab in P regions to construct a preliminary prediction model for predicting the grade of wheat scab (grade prediction models of wheat scab with M different network architectures), and the preliminary prediction model is trained, and through the mean square error MSE, coefficient of determination R 2, Accuracy, and composite - score are used as model evaluation indicators to measure the accuracy of the model. Among them, the mean squared error (MSE) is used to represent the average of the squares of the differences between the predicted values and the true values of the preliminary prediction model, which is used to measure the average variance between the predicted values and the true values of the preliminary prediction model; the coefficient of determination (R 2 ) is used to represent the explanatory power of the preliminary prediction model for the target variable, that is, the proportion of the variance between the predicted values and the true values of the preliminary prediction model, with a value range from 0 to 1. The closer it is to 1, the better the explanatory power of the preliminary prediction model for the target variable; the prediction accuracy (Accuracy) is defined as the model prediction being correct when the difference between the predicted value and the true value of the preliminary prediction model is less than 0.2. Since the performance trends of the mean squared error (MSE), the coefficient of determination (R 2 ), and the accuracy (Accuracy) are different, in this application, a composite - score is introduced, and a weighted average method is used to unify the evaluation direction of the indicators, so as to intuitively evaluate the performance of the model. The first - level prediction model with the worst prediction effect on the wheat scab grade and the second - level prediction model with the best prediction effect on the wheat scab grade are selected from the wheat scab grade prediction models with M different network architectures.
[0046] Specifically, according to the obtained N meteorological principal components and the obtained grade sample data of wheat scab in P regions, based on the pre - constructed wheat scab grade prediction models with M different network architectures, M regression relationships between the wheat scab grade and the meteorological principal components are generated. In a specific example, preliminary prediction models with different network architectures are respectively constructed based on support vector machine (SVR), k - nearest neighbor (KNN), Bayesian, gradient boosting decision tree (GBDT), and random forest (RF). The N meteorological principal components are input into the preliminary prediction models, and the grade sample data of wheat scab in P regions are used as labels to train the preliminary prediction models, and the regression relationships between the meteorological factors and the wheat scab grade are established.
[0047] Then, based on the M regression relationships, the wheat scab grades corresponding to the N meteorological principal components in each region are predicted to obtain M predicted result datasets. Among them, each predicted result dataset contains P wheat scab grade prediction values; finally, according to the M predicted result datasets, the wheat scab grade prediction models with M different network architectures are screened to determine the first-grade prediction model and the second-grade prediction model. Specifically, according to the formula:
[0048]
[0049] The prediction effects of the preliminary prediction models constructed based on support vector machine (SVR), k-nearest neighbor (KNN), Bayesian, gradient boosting decision tree (GBDT for short), and random forest (RF) are evaluated to screen out the first-grade prediction model with the worst prediction effect on wheat scab grades and the second-grade prediction model with the best prediction effect on wheat scab grades. In the formula, n is the total number of samples (i.e., grade sample data), y t,i is the actual value of the i-th sample, and y p,i is the predicted value of the i-th sample obtained from the preliminary prediction model, is the average value of the actual values of n samples,
[0050] Ⅱ(·) is an indicator function. When |y t,i -y p,i |<0.2 is true, Ⅱ(·) takes the value of 1, otherwise it is 0.
[0051] In a specific application scenario, the prediction effects of the preliminary prediction models with several different network architectures of support vector machine (SVR), k-nearest neighbor (KNN), Bayesian, gradient boosting decision tree (GBDT), and random forest (RF) are shown in Table 1.
[0052] Table 1 Comparison results of grade prediction models with different network architectures
[0053]
[0054] It can be seen that under the same conditions, the grade prediction model based on random forest (RF) has better fitting ability and generalization ability when dealing with the complex linear relationship between meteorological factors and scab grades; while the prediction effect of the grade prediction model based on support vector regression (SVR) is the worst.
[0055] Step S103: Optimize the second-level prediction model according to the fusion data obtained by fusing the prediction results of the first-level prediction model with the grade sample data, so as to predict the Fusarium head blight grade corresponding to N meteorological principal components in each region through the optimized second-level prediction model.
[0056] In this application, by comparing the prediction effects of preliminary prediction models with several different network architectures such as support vector machine (SVR), k-nearest neighbor (KNN), Bayesian, gradient boosting decision tree (GBDT), and random forest (RF), the prediction results of the grade prediction model of support vector regression (SVR) are used as new features and fused with the N meteorological principal components after dimensionality reduction to form an extended feature dataset (i.e., fusion data). Furthermore, the diversity of the input data of the random forest model is effectively enhanced.
[0057] Specifically, standardize the prediction output of the grade prediction model based on support vector regression (SVR prediction result) to make its dimension consistent with other meteorological factors; then, add the standardized SVR prediction result as a new feature to construct a fusion dataset containing N meteorological principal components and SVR prediction output, and on the basis of the fusion data, retrain and optimize the random forest model. Among them, use the extended feature data (fusion data) as the input, take the actual grade of Fusarium head blight as the label, construct a training dataset, and use the random forest algorithm to retrain multiple decision trees to integrate and learn the influence of different features on the Fusarium head blight grade; finally, through the voting mechanism of the random forest model, predict the extended dataset and output the optimized prediction result of the Fusarium head blight grade.
[0058] In a specific example, according to the fusion data, optimize the parameters of the grade prediction model based on random forest (i.e., the second-level prediction model), the number of decision trees (n-estimators) and the maximum depth of the tree (max-depth), based on the grid search algorithm (Grid Search, abbreviated as GS), so as to improve the prediction accuracy of the grade prediction model based on random forest. Among them, conduct a comprehensive search in the parameter space through grid search to automatically find the optimal parameter configuration. Specifically, set different parameter ranges (the number of decision trees (n-estimators) and the maximum depth of the tree (max-depth)), use each group of parameter combinations to train the grade prediction model based on random forest, and evaluate its performance on the validation set; select the parameter combination that can obtain the best prediction effect on the validation set.
[0059] In this application, by combining the prediction results of a model that integrates support vector regression and random forest, the input features of the random forest-based grade prediction model are enriched, significantly improving the prediction accuracy of the model. In the fused model, the influence of various meteorological factors on the Fusarium head blight grade can be more accurately quantified, effectively making up for the deficiency of a single model in dealing with data complexity, and enabling the model to more precisely evaluate the Fusarium head blight grade under different meteorological factor changes.
[0060] The parameters optimized through grid search further improve the accuracy and stability of the random forest-based grade prediction model, enabling it to accurately predict the Fusarium head blight grade of wheat under different meteorological conditions. In this application, during the optimization process of the random forest-based grade prediction model, the mean squared error MSE, coefficient of determination R 2 , accuracy, and composite score are also used as evaluation metrics to measure the model accuracy, effectively ensuring the reliability of the prediction results.
[0061] By performing principal component analysis on Q meteorological factors in P regions, coupling the prediction results of the support vector regression-based grade prediction model, and optimizing the random forest-based grade prediction model through grid search, the prediction results of the random forest-based grade prediction model are improved to varying degrees within different regional ranges, fully demonstrating that principal component analysis can retain most of the effective information in the original data; as a flexible regression model, the support vector regression-based grade prediction model can more effectively handle non-linear data relationships, and using its results as the input of the random forest-based grade prediction model can provide more valuable information, making up for the deficiency of the random forest-based grade prediction model in capturing complex patterns and further improving the prediction accuracy of the random forest-based grade prediction model. At the same time, by optimizing the n-estimators and max-depth of the random forest-based grade prediction model through grid search, the prediction accuracy of the random forest-based grade prediction model is further significantly improved.
[0062] Compared with the unoptimized random forest-based grade prediction model, the optimized random forest-based grade prediction model has higher accuracy in predicting the Fusarium head blight grade of wheat by combining the dimensionality reduction processing of principal component analysis, coupling the prediction results of the support vector regression-based model, and optimizing the grid search parameters. The combination of multiple models not only enhances the capture and processing of complex data relationships but also improves the robustness of the model.
[0063] As Figure 9 shown, the embodiment of this application also provides a wheat Fusarium head blight grade prediction system based on a grid search PCA-SVR-RF model, including:
[0064] The PCA dimensionality reduction unit 901 is configured to perform principal component analysis on Q meteorological factors of P regions obtained, to obtain N reduced-dimensional meteorological principal components; where P, Q, and N are all positive integers;
[0065] The model screening unit 902 is configured to screen M wheat scab grade prediction models with different network architectures pre-constructed according to the obtained N meteorological principal components and the grade sample data of wheat scab in P regions, to determine a first grade prediction model and a second grade prediction model; where M is a positive integer, the first grade prediction model is the worst model for wheat scab grade prediction, and the second grade prediction model is the best model for wheat scab grade prediction;
[0066] The model optimization unit 903 is configured to optimize the second grade prediction model according to the fusion data obtained by fusing the prediction result of the first grade prediction model and the grade sample data, and predict the wheat scab grade corresponding to the N meteorological principal components in each region through the optimized second grade prediction model.
[0067] The wheat scab grade prediction system based on the grid search PCA-SVR-RF model provided by the embodiments of the present application can implement the steps and processes of the wheat scab grade prediction method based on the grid search PCA-SVR-RF model in any of the above embodiments, and achieve the same technical effects, which will not be elaborated here one by one.
[0068] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of the present invention, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically defined.
[0069] In the present invention, the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.
[0070] The above are only the preferred embodiments of the present application, and are not used to limit the present application. For those skilled in the art, the present application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims
1. A method for predicting wheat scab grade based on grid search PCA-SVR-RF model, characterized in that: Step S101, performing principal component analysis on the Q meteorological factors of the acquired P regions to obtain N meteorological principal components with reduced dimensions; wherein P, Q, and N are all positive integers; Step S102: According to the obtained N meteorological principal components and the grade sample data of wheat scab in P regions, the pre-constructed grade prediction models of wheat scab with different network architectures are screened to determine a first-level prediction model and a second-level prediction model; wherein M is a positive integer, the first-level prediction model is the worst model for wheat scab grade prediction, and the second-level prediction model is the best model for wheat scab grade prediction; Step S103, optimizing the second-level prediction model according to the fusion data obtained by fusing the prediction result of the first-level prediction model with the level sample data, and predicting the wheat fusarium disease level corresponding to the N meteorological main components in each area through the optimized second-level prediction model.
2. The wheat scab grade prediction method based on the grid search PCA-SVR-RF model according to claim 1, characterized in that: In step S101, Calculate the covariance matrix of the data after the Q meteorological factors are standardized; Perform eigenvalue decomposition on the covariance matrix to obtain N ′ The eigenvalues of the principal components; where N ′ ≥N,N ′ is a positive integer; According to N ′ The sizes of the eigenvalues of the principal components are used to determine the N meteorological principal components.
3. The method for predicting wheat scab grade based on grid search PCA-SVR-RF model according to claim 1, characterized in that: In step S102, According to the obtained N meteorological principal components and the obtained wheat scab grade sample data of P regions, based on the pre-constructed wheat scab grade prediction models of M different network architectures, M regression relationships between the wheat scab grade and the N meteorological principal components are generated; Based on the M regression relationships, the wheat scab grade corresponding to the N meteorological principal components in each region is predicted to obtain M prediction result data sets; wherein each prediction result data set contains P predicted values of the wheat scab grade; According to the M prediction result data sets, the graded prediction models of wheat fusarium head blight of M different network architectures are screened to determine the first graded prediction model and the second graded prediction model.
4. The method for predicting wheat scab grade based on grid search PCA-SVR-RF model according to claim 1, characterized in that: In step S103, Fusion the prediction result of the first-level prediction model with the B meteorological principal components to obtain the fused data; According to the fused data, the number of parameter decision trees and the maximum depth of the trees of the second level prediction model are optimized based on a grid search algorithm.
5. The method for predicting wheat scab grade based on grid search PCA-SVR-RF model according to claim 1, characterized in that: The first grade prediction model is a grade prediction model based on support vector regression; and / or, The second level prediction model is a level prediction model based on random forest.
6. The method for predicting wheat scab grade based on the grid search PCA-SVR-RF model according to any one of claims 1 to 5, characterized in that: The meteorological data of the key growth period of wheat in P regions are obtained as the meteorological factors.
7. A wheat scab grade prediction system based on grid search PCA-SVR-RF model, characterized in that: include: The PCA dimension reduction unit is configured to perform principal component analysis on the Q meteorological factors of the acquired P regions to obtain N meteorological principal components of dimension reduction; wherein P, Q, and N are all positive integers; A model screening unit is configured to screen the pre-constructed wheat fusarium grade prediction models of M different network architectures according to the obtained N meteorological principal components and the obtained wheat fusarium grade sample data of P regions, and determine a first-level prediction model and a second-level prediction model; wherein M is a positive integer, the first-level prediction model is the worst model for wheat fusarium grade prediction, and the second-level prediction model is the best model for wheat fusarium grade prediction; The model optimization unit is configured to optimize the second-level prediction model according to the fusion data obtained by fusing the prediction result of the first-level prediction model with the level sample data, and predict the wheat fusarium disease level corresponding to the N meteorological main components in each area through the optimized second-level prediction model.