A method for constructing optimal corn planting density based on multi-source data

Abstract

The invention discloses a method for constructing an optimal planting density of corn based on multi-source data, and belongs to the technical field of agricultural production. The invention discloses a method for constructing an optimal planting density of corn based on multi-source data, and obtains corn density test data from Chinese and English literature databases by keyword retrieval, performs quadratic fitting on density gradient and yield data, identifies and processes abnormal values ​​in the data, and finally screens an optimal planting density OPD data set. According to the size of a research area, a corresponding resolution data grid and its coordinate system are selected, and sample points of soil and meteorological data are collected, and synergistic variables related to the optimal planting density OPD are screened. A machine learning model is constructed based on the synergistic variables, and spatial prediction and mapping of regional optimal planting density OPD are performed. The invention solves the problems that data in traditional interpolation methods do not conform to normal distribution and have strong spatial heterogeneity, and simultaneously realizes high-precision regional OPD prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural production, and in particular to a method for constructing optimal corn planting density based on multi-source data. Background Art

[0002] Food security is an important issue of global concern. As an important economic crop, corn plays an important role in food supply and industrial production, and directly affects the development of agriculture in my country. Against the background of increasingly tight arable land resources, how to increase corn yields and build high-yield corn groups on limited arable land is an urgent problem to be solved. The increase in crop yield is the result of the combined effects of varieties, climate environment, soil and agricultural management. Compared with the improvement of crop varieties, increasing planting density is one of the most significant means of increasing production. Lower planting density often causes waste of land resources, while too high planting density will affect the growth and development of crops. Therefore, determining the optimal planting density OPD of regional corn and building high-yield corn groups in different regions are the key to increasing corn yields.

[0003] At present, most of the research on the optimal planting density (OPD) of regional corn is based on farm experiments and is still in the field experiment stage. Due to the differences in environmental conditions in various regions, these research results are difficult to directly apply to agricultural production in large areas. At the same time, existing research on corn OPD pays less attention to the differences in corn varieties in different regions. Different varieties have different plant types and growth periods, which often affect OPD during the planting process. The existing optimal planting density construction method is still imperfect, the model accuracy is low, and it is difficult to meet the needs of actual agricultural management. Summary of the invention

[0004] The object of the present invention is to provide a method for constructing optimal corn planting density based on multi-source data to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for constructing optimal corn planting density based on multi-source data, comprising the following steps:

[0006] S1: Obtain corn density test data from Chinese and English literature databases by keyword search;

[0007] S2: Perform a quadratic fit between the density gradient and the yield data, and define the density at the highest point of the fitting curve as the optimal planting density OPD;

[0008] S3: Process the outliers in the data and select the optimal planting density OPD data set;

[0009] S4: According to the size of the study area, select the corresponding resolution data grid and its coordinate system, and collect samples of soil and meteorological data;

[0010] S5: Screening of synergistic variables related to the optimal planting density (OPD), including soil, meteorological and crop factors;

[0011] S6: Construct a machine learning model based on collaborative variables to perform spatial prediction and mapping of regional optimal planting density OPD.

[0012] Furthermore, in step S1: based on domestic and foreign literature databases, the corresponding journals and master's and doctoral theses on corn planting density experiments were searched by using the keywords "yield", "density", "corn" and "China", ensuring that the data sources of relevant domestic and foreign research were covered. The selected paper data included independent test sites, crop varieties, planting densities and their yield information, ensuring the comprehensiveness and diversity of the data. The selection rules mainly included the following five points: field experiments conducted in specific areas, excluding greenhouses and potted plants, ensuring the environmental consistency of the data; the plant density evaluated in a given field experiment exceeded three levels, ensuring the representativeness and reliability of the data; providing specific yield information, excluding model-simulated yield, ensuring the authenticity and accuracy of the data; having test site information; stress and intercropping experiments were not taken into account. Data extraction was mainly obtained through article tables. For data displayed only in graphics, GetData and Graph Digitizer software were used to extract the corresponding values. Strict selection rules and standardized data extraction methods ensured the internal consistency of the data set. Special software was used to extract graphic data, which improved the efficiency and accuracy of data extraction and reduced the errors and workload of manual data extraction.

[0013] Furthermore, in step S2: the density gradient of each independent test is subjected to quadratic fitting, and the density at the highest point of the fitting curve is defined as the optimal planting density OPD, and the corresponding yield is the optimal density yield. The yield at this time is the maximum yield that can be achieved in the area, and the corresponding optimal planting density OPD value of each sample is obtained; the data that cannot reflect the true optimal planting density OPD is eliminated, that is, the data in which the OPD of the fitting quadratic function with an opening upward a>0 and an opening downward a<0 is the maximum or minimum density in the original treatment, which effectively removes outliers that deviate from the normal range, ensures the authenticity and reliability of the data, and statistically calculates the optimal planting density OPD data of the remaining fitting results to construct an optimal planting density OPD data set as the prediction result data set.

[0014] Furthermore, in step S3: the obtained optimal planting density OPD data is further cleaned, and the Z-score method and IQR method are used to remove outliers and remove noise. The cleaned data conforms to the normal distribution and is used as the dependent variable for machine learning.

[0015] Further, in step S4: according to the size of the study area, select the corresponding resolution data grid and its coordinate system, and collect soil and meteorological data samples. The specific steps are as follows:

[0016] S4-1: Crop growth period information, soil information, meteorological information and crop genotype data are supplemented as collaborative variables, and crop growth period information is obtained through the published crop data platform: soil information is obtained through the basic attribute data set of the National Soil Information Grid, including soil organic carbon, pH, total nitrogen, total phosphorus, total potassium, cation exchange capacity, soil texture type, soil bulk density and soil thickness, to ensure the comprehensiveness and detail of soil information; meteorological data are obtained through the spatiotemporal environment big data platform, including rainfall during the growth period, accumulated temperature, daily average minimum temperature, average high temperature and potential evaporation, which fully reflects the impact of climate conditions on corn growth; crop genotype data are obtained through the existing attribute maps of the data platform, including variety plant type and growth period information, and soil information, meteorological information and crop genotype data are obtained through multiple data platforms, so that the data set covers a variety of key factors affecting corn growth;

[0017] S4-2: Unify the coordinate system and define the grid, including: unify the point information of different experimental sites and the soil and meteorological data into the same coordinate system, ensure the alignment of all data in the same spatial reference system, extract the soil and meteorological data of the corresponding points, and extract the soil and meteorological data of each experimental point by unifying the coordinate system and defining the grid.

[0018] Further, in step S5: LASSO regression and principal component analysis PCA are used to screen the co-variables related to the optimal planting density OPD, including selecting soil bulk density, total nitrogen content and soil pH as soil factors, selecting rainfall during the growth period, accumulated temperature, daily average minimum temperature, average high temperature and potential evapotranspiration as meteorological factors, selecting corn plant type and growth cycle time as crop factors, and selecting a total of 10 co-variables for random forest model training, among which corn plant type is divided into compact type, intermediate type and flat type, and feature engineering is applied to convert them into digital labels, the compact type is represented by the digital label '1', and the intermediate type is represented by the digital label '2' The flat type is represented by the digital label '3'. By using LASSO regression to screen variables, the multicollinearity problem was successfully eliminated and unnecessary noise variables were reduced, thereby improving the robustness and prediction accuracy of the model. At the same time, PCA was used to reduce the data dimension and extract the main features, further improving the calculation efficiency and prediction accuracy of the model. The automatic screening of key synergistic variables reduced the complexity of manual variable selection and improved the efficiency of data processing and analysis. The screened synergistic variables covered three major factors: soil, meteorology, and crops, making the model prediction results more interpretable and credible, which helps to understand the impact of various factors on the optimal planting density of corn.

[0019] Further, in step S6: a machine learning model is constructed based on the synergistic variables to perform spatial prediction mapping of the optimal planting density OPD in the region. The specific steps are as follows:

[0020] S6-1: For the acquired data set, the optimal planting density OPD of the sample point coordinates is used as the dependent variable, and the co-variable corresponding to the sample point is used as the independent variable to construct a machine learning model for predicting the optimal planting density OPD. Multi-source data is used to cover various factors affecting corn planting density, which enhances the model's comprehensive prediction ability for complex environmental variables.

[0021] S6-2: Improving prediction accuracy by using the random forest model as a prediction model;

[0022] S6-3: In the defined grid, the synergistic variable value of the unknown point is obtained and substituted into the trained random forest model to obtain the optimal planting density OPD prediction value of the grid. By traversing each grid, the OPD spatial distribution prediction of the region can be realized.

[0023] Further, in step S6: calculate the root mean square error RMSE, the relative root mean square error RRSE and the determination coefficient R 2 , the calculation formulas for each indicator are as follows: Calculate the root mean square error RMSE:

[0024]

[0025] Calculate the relative root mean square error RRSE:

[0026]

[0027] Calculate the coefficient of determination R 2 :

[0028]

[0029] Among them, y i represents the i-th observation value, represents the i-th predicted value, represents the average value of the observations, n represents the number of samples, and the root mean square error RMSE, relative root mean square error RRSE and determination coefficient R are calculated. 2 Evaluate the accuracy of the random forest model.

[0030] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: the present invention retrieves relevant journals and master's and doctoral theses on corn planting density experiments in Chinese and English literature through the keywords of "yield", "density", "corn" and "China", obtains corn density test data in a specific area, ensures the comprehensiveness and representativeness of the data, performs quadratic fitting on the gradient data to obtain the optimal planting density OPD at the test point, can improve the data accuracy of OPD, and is suitable for large-scale research, uses LASSO regression and principal component analysis PCA to screen synergistic variables related to the optimal planting density OPD, uses variety data, soil data and environmental data related to corn growth and development as synergistic variables, planting density as the dependent variable, and related synergistic variables as independent variables, and constructs a random forest model to realize spatial prediction of regional corn OPD, overcomes the problems of data not conforming to normal distribution and strong spatial heterogeneity in traditional interpolation methods, and realizes high-precision regional OPD prediction at the same time, providing an important reference for corn planting and yield improvement in actual agricultural production. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0032] Figure 1 It is a flow chart of a method for constructing optimal corn planting density based on multi-source data of the present invention;

[0033] Figure 2 It is a relationship diagram between optimal density and yield of a method for constructing optimal corn planting density based on multi-source data of the present invention;

[0034] Figure 3 It is a comparison chart of predicted values ​​and measured values ​​of a method for constructing optimal corn planting density based on multi-source data of the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] See also Figure 1-Figure 3 The present invention provides a technical solution: a method for constructing optimal corn planting density based on multi-source data, comprising the following steps:

[0037] S1: Obtain corn density test data from Chinese and English literature databases by keyword search;

[0038] S2: Perform a quadratic fit between the density gradient and the yield data, and define the density at the highest point of the fitting curve as the optimal planting density OPD;

[0039] S3: Process the outliers in the data and select the optimal planting density OPD data set;

[0040] S4: According to the size of the study area, select the corresponding resolution data grid and its coordinate system, and collect samples of soil and meteorological data;

[0041] S5: Screening of synergistic variables related to the optimal planting density (OPD), including soil, meteorological and crop factors;

[0042] S6: Construct a machine learning model based on collaborative variables to perform spatial prediction and mapping of regional optimal planting density OPD.

[0043] In step S1: Based on domestic and foreign literature databases, the corresponding journals and master's and doctoral theses on corn planting density experiments were searched by using the keywords "yield", "density", "corn" and "China", ensuring that the data sources of relevant domestic and foreign research were covered. The selected paper data included independent test sites, crop varieties, planting density and its yield information, ensuring the comprehensiveness and diversity of the data. The selection rules mainly included the following five points: field experiments conducted in China, excluding greenhouses and potted plants, ensuring the environmental consistency of the data; the plant density evaluated in a given field experiment exceeded three levels, ensuring the representativeness and reliability of the data; providing specific yield information, excluding model-simulated yield, ensuring the authenticity and accuracy of the data; having test site information; stress and intercropping experiments were not taken into account. Data extraction was mainly obtained through article tables. For data displayed only in graphics, GetData and Graph Digitizer software were used to extract the corresponding values. Strict selection rules and standardized data extraction methods ensured the internal consistency of the data set. The use of dedicated software to extract graphic data improved the efficiency and accuracy of data extraction and reduced the errors and workload of manual data extraction.

[0044] In step S2: perform quadratic fitting on the density gradient of each independent test, define the density at the highest point of the fitting curve as the optimal planting density OPD, and the corresponding yield as the optimal density yield. The yield at this time is the maximum yield that can be achieved in the area, and obtain the corresponding optimal planting density OPD value for each sample; eliminate data that cannot reflect the true optimal planting density OPD, that is, eliminate data where the OPD in the fitting quadratic function with an upward opening a>0 and a downward opening a<0 is the maximum or minimum density in the original treatment, effectively eliminate outliers that deviate from the normal range, and ensure the authenticity and reliability of the data; statistically analyze the optimal planting density OPD data of the remaining fitting results, and construct an optimal planting density OPD data set as the prediction result data set.

[0045] In step S3: the obtained optimal planting density OPD data is further cleaned, and the Z-score method and IQR method are used to remove outliers and remove noise. The cleaned data conforms to the normal distribution and is used as the dependent variable for machine learning.

[0046] In step S4: according to the size of the study area, select the corresponding resolution data grid and its coordinate system, and collect soil and meteorological data samples. The specific steps are as follows:

[0047] S4-1: Crop growth period information, soil information, meteorological information and crop genotype data are supplemented as synergistic variables, and crop growth period information is obtained through the published crop data platform: soil information is obtained through the basic attribute data set of China's high-resolution national soil information grid, including soil organic carbon, pH, total nitrogen, total phosphorus, total potassium, cation exchange capacity, soil texture type, soil bulk density and soil thickness, to ensure the comprehensiveness and detail of soil information; meteorological data are obtained through the spatiotemporal environment big data platform, including rainfall during the growth period, accumulated temperature, daily average minimum temperature, average high temperature and potential evaporation, which fully reflects the impact of climate conditions on corn growth; crop genotype data are obtained through the existing attribute maps of the data platform, including variety plant type and growth period information, and soil information, meteorological information and crop genotype data are obtained through multiple data platforms, so that the data set covers a variety of key factors affecting corn growth;

[0048] S4-2: Unify the coordinate system and define the grid, including: unify the point information of different experimental sites and the soil and meteorological data into the same coordinate system, ensure the alignment of all data in the same spatial reference system, extract the soil and meteorological data of the corresponding points, and extract the soil and meteorological data of each experimental point by unifying the coordinate system and defining the grid.

[0049] In step S5: LASSO regression and principal component analysis PCA are used to screen the synergistic variables related to the optimal planting density OPD, including selecting soil bulk density, total nitrogen content and soil pH as soil factors, selecting rainfall during the growing period, accumulated temperature, daily average minimum temperature, average high temperature and potential evapotranspiration as meteorological factors, selecting corn plant type and growing cycle time as crop factors, and selecting a total of 10 synergistic variables for random forest model training. Among them, corn plant type is divided into compact type, intermediate type and flat type, and feature engineering is used to convert them into digital labels. The compact type is represented by the digital label '1', the intermediate type is represented by the digital label '2', and the flat type is represented by the digital label '3'. The expansion type is represented by the digital label '3'. By using LASSO regression to screen variables, the multicollinearity problem was successfully eliminated and unnecessary noise variables were reduced, thereby improving the robustness and prediction accuracy of the model. At the same time, PCA was used to reduce the data dimension and extract the main features, further improving the computational efficiency and prediction accuracy of the model. Automated screening of key synergistic variables reduced the complexity of manual variable selection and improved the efficiency of data processing and analysis. The screened synergistic variables covered three major factors: soil, meteorology, and crops, making the model prediction results more interpretable and credible, which helps to understand the impact of various factors on the optimal planting density of corn.

[0050] In step S6: a machine learning model is constructed based on the synergistic variables to perform spatial prediction mapping of the optimal planting density OPD in the region. The specific steps are as follows:

[0051] S6-1: For the acquired data set, the optimal planting density OPD of the sample point coordinates is used as the dependent variable, and the co-variable corresponding to the sample point is used as the independent variable to construct a machine learning model for predicting the optimal planting density OPD. Multi-source data is used to cover various factors affecting corn planting density, which enhances the model's comprehensive prediction ability for complex environmental variables.

[0052] S6-2: Improving prediction accuracy by using the random forest model as a prediction model;

[0053] S6-3: In the defined grid, the synergistic variable value of the unknown point is obtained and substituted into the trained random forest model to obtain the optimal planting density OPD prediction value of the grid. By traversing each grid, the OPD spatial distribution prediction of the region can be realized.

[0054] In step S6: calculate the root mean square error RMSE, relative root mean square error RRSE and determination coefficient R 2 , the calculation formulas for each indicator are as follows: Calculate the root mean square error RMSE:

[0055]

[0056] Calculate the relative root mean square error RRSE:

[0057]

[0058] Calculate the coefficient of determination R 2 :

[0059]

[0060] Among them, y i represents the i-th observation value, represents the i-th predicted value, represents the average value of the observations, n represents the number of samples, and the root mean square error RMSE, relative root mean square error RRSE and determination coefficient R are calculated. 2 Evaluate the accuracy of the random forest model.

[0061] In the embodiment: based on domestic and foreign literature databases, the corresponding journals and master's and doctoral theses on corn planting density experiments published on CNKI and Web of Science between 2000 and 2021 were retrieved by searching the keywords "yield", "density", "corn" and "China", and a total of 2590 available independent experiments were obtained from 838 documents, including 11071 pairs of observation data; the density gradient was quadratically fitted with the yield data, and the density at the highest point of the fitting curve was defined as the optimal planting density OPD; the outliers in the data were removed and denoised using the Z-score method and the IQR method, and the optimal planting density OPD data set was screened. The final data set used contained 998 available independent experiment OPDs obtained from 421 documents; the corresponding Crop growth period information, soil information, and meteorological information are supplemented as collaborative variables. The soil data uses the basic attribute data set of China's high-resolution national soil information grid with a resolution of 1 km from 2010 to 2018, including soil organic carbon, pH, total nitrogen, total phosphorus, total potassium, cation exchange capacity, soil texture type, soil bulk density, and soil thickness. The meteorological data uses the growth period rainfall, accumulated temperature, daily average minimum temperature, average high temperature, and potential evapotranspiration of the spatiotemporal three-pole environment big data platform with a resolution of 1 km from 1901 to 2021. The genotype data of crops are obtained through existing attribute maps. :Variety plant type and growth cycle time; unify the experimental site coordinates with soil and meteorological data to the same coordinate grid, extract soil and meteorological data of the corresponding points; use LASSO regression and principal component analysis PCA to screen the co-variables related to the optimal planting density OPD, including selecting soil bulk density, total nitrogen content and soil pH as soil factors, selecting rainfall during the growth period, accumulated temperature, daily average minimum temperature, average high temperature and potential evapotranspiration as meteorological factors, selecting corn plant type and growth cycle time as crop factors, and selecting a total of 10 co-variables for random forest model training, climate data; for the obtained The data set uses OPD as the dependent variable and the adjacent co-variables corresponding to the sample points as independent variables to build a machine learning model for predicting OPD. Specifically, the random forest model is selected as the prediction model. The data set includes 998 data points in total, and the training set and validation set of the model are divided into 8:2. The parameters of the random forest model are optimized through ten-fold cross validation. The final parameter settings are as follows: the number of decision trees in the random forest is set to 150, the number of variables is set to 10, and the maximum depth is set to 20; the coefficient of determination R 2 , root mean square error RMSE and relative root mean square error RRSE evaluate the accuracy of the random forest model, and the final model prediction accuracy R 2The coefficient of variation is 0.64, the RMSE is 0.81, and the RRMSE is 0.11. In the defined grid, the values ​​of the covariates of the unknown points are obtained, including soil bulk density, total nitrogen content, soil pH, rainfall during the growing period, accumulated temperature, average daily minimum temperature, average high temperature, potential evapotranspiration, corn plant type and growing cycle covariates. The specific data are as follows: [1.35, 0.2, 6.7, 550, 3100, 12, 29, 1250, 1, 111]. Substitute it into the trained random forest model and obtain the optimal planting density OPD prediction value of the grid as 70,000 plants / hectare. Traverse each grid to realize the prediction of the OPD spatial distribution of the region.

[0062] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0063] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for constructing optimal corn planting density based on multi-source data, comprising the following steps: S1: Obtain corn density test data from Chinese and English literature databases by keyword search; S2: Perform a quadratic fit between the density gradient and the yield data, and define the density at the highest point of the fitting curve as the optimal planting density OPD; S3: Process the outliers in the data and select the optimal planting density OPD data set; S4: According to the size of the study area, select the corresponding resolution data grid and its coordinate system, and collect samples of soil and meteorological data; S5: Screening of synergistic variables related to the optimal planting density (OPD), including soil, meteorological and crop factors; S6: Construct a machine learning model based on collaborative variables to perform spatial prediction and mapping of regional optimal planting density OPD.

2. The method for constructing optimal corn planting density based on multi-source data according to claim 1, characterized in that: In step S1: the corn density test data includes independent test sites, crop varieties, planting density and yield information.

3. The method for constructing optimal corn planting density based on multi-source data according to claim 1, characterized in that: In step S2: perform quadratic fitting on the density gradient of each independent experiment, define the density at the highest point of the fitting curve as the optimal planting density OPD, and the corresponding yield as the optimal density yield. The yield at this time is the maximum yield that can be achieved in the area, and obtain the corresponding optimal planting density OPD value for each sample; eliminate the data that cannot reflect the true optimal planting density OPD, that is, the data where the OPD in the fitting quadratic function with an upward opening a>0 and a downward opening a<0 is the maximum or minimum density in the original treatment, and count the optimal planting density OPD data of the remaining fitting results to construct an optimal planting density OPD data set.

4. The method for constructing optimal corn planting density based on multi-source data according to claim 1, characterized in that: In step S3: the obtained optimal planting density OPD data is further cleaned, and the Z-score method and IQR method are used to remove outliers and remove noise.

5. The method for constructing optimal corn planting density based on multi-source data according to claim 1, characterized in that: In step S4: according to the size of the study area, select the corresponding resolution data grid and its coordinate system, and collect soil and meteorological data samples. The specific steps are as follows: S4-1: Crop growth period information, soil information, meteorological information and crop genotype data are supplemented as collaborative variables, and crop growth period information is obtained through the published crop data platform; soil information is obtained through the basic attribute data set of the National Soil Information Grid, including soil organic carbon, pH, total nitrogen, total phosphorus, total potassium, cation exchange capacity, soil texture type, soil bulk density and soil thickness; meteorological data is obtained through the spatiotemporal environment big data platform, including rainfall during the growth period, accumulated temperature, daily average minimum temperature, average high temperature and potential evaporation; crop genotype data is obtained through the existing attribute map of the data platform, including variety plant type and growth period information; S4-2: Unify the coordinate system and define the grid, including: unify the point information of different experimental sites and the soil and meteorological data into the same coordinate system, and extract the soil and meteorological data of the corresponding points.

6. The method for constructing optimal corn planting density based on multi-source data according to claim 1, characterized in that: In step S5: LASSO regression and principal component analysis PCA are used to screen the co-variables related to the optimal planting density OPD, including selecting soil bulk density, total nitrogen content and soil pH as soil factors, selecting rainfall during the growing period, accumulated temperature, average daily minimum temperature, average high temperature and potential evaporation as meteorological factors, selecting corn plant type and growing cycle time as crop factors, and selecting a total of 10 co-variables for random forest model training, among which corn plant type is divided into compact, intermediate and flat types.

7. The method for constructing optimal corn planting density based on multi-source data according to claim 1, characterized in that: In step S6: a machine learning model is constructed based on the synergistic variables to perform spatial prediction mapping of the optimal planting density OPD in the region. The specific steps are as follows: S6-1: For the acquired data set, the optimal planting density OPD of the sample point coordinates is used as the dependent variable, and the co-variable corresponding to the sample point is used as the independent variable to construct a machine learning model for predicting the optimal planting density OPD; S6-2: Improving prediction accuracy by using the random forest model as a prediction model; S6-3: In the defined grid, obtain the synergistic variable value of the unknown point, substitute it into the trained random forest model, obtain the optimal planting density OPD prediction value of the grid, traverse each grid, and realize the OPD spatial distribution prediction of the region.

8. The method for constructing optimal corn planting density based on multi-source data according to claim 7, characterized in that: In step S6: calculate the root mean square error RMSE, relative root mean square error RRSE and determination coefficient R 2 , the calculation formulas for each indicator are as follows: Calculate the root mean square error RMSE: Calculate the relative root mean square error RRSE: Calculate the coefficient of determination R 2 : Among them, y i represents the i-th observation value, represents the i-th predicted value, represents the average value of the observations, n represents the number of samples, and the root mean square error RMSE, relative root mean square error RRSE and determination coefficient R are calculated. 2 Evaluate the accuracy of the random forest model.

Images