Cultivated land yield prediction method and device, electronic equipment and storage medium

Abstract

The invention provides a cultivated land yield prediction method and device, electronic equipment and a storage medium, belongs to the technical field of agricultural data processing, realizes dynamic prediction of cultivated land quality based on deep learning by obtaining quality feature information with a long-term dependency relationship, and improves the accuracy of determining a cultivated land quality prediction result. Meanwhile, the cultivated land yield prediction result is obtained according to the key yield index data and the cultivated land quality prediction result, the cultivated land quality and the grain yield are effectively combined, and the cultivated land yield prediction result can be comprehensively determined. The correlation between the key yield index data and the cultivated land quality prediction result and the cultivated land yield is determined through geographically weighted regression analysis, and the accuracy of the cultivated land yield prediction result is improved. According to the method, the knowledge graph is used for integrating multi-source data, providing background knowledge, assisting feature selection and the like, so that the accuracy and reliability of the cultivated land quality prediction model and the cultivated land yield prediction model are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural data processing, and particularly relates to a cultivated land yield prediction method and device, electronic equipment and storage medium. BACKGROUND

[0002] Predicting the cultivated land quality and grain yield of soil is conducive to improving the cultivated land quality and increasing the grain yield.

[0003] Existing large-scale cultivated land quality evaluation and grain yield prediction mainly rely on meteorological and remote sensing data, and are relatively single in considering influencing factors, which leads to inaccurate yield prediction of cultivated land. SUMMARY

[0004] The present application provides a cultivated land yield prediction method and device, electronic equipment and storage medium to solve the problem of inaccurate yield prediction of cultivated land in the prior art, and to improve the accuracy of yield prediction of cultivated land.

[0005] The present application provides a cultivated land yield prediction method, comprising the following steps: obtaining quality index data related to cultivated land quality and quality historical research data related to cultivated land quality; constructing a quality estimation knowledge graph based on the quality index data and the quality historical research data to obtain key quality index data; inputting the key quality index data into a cultivated land quality prediction model to obtain a cultivated land quality prediction result output by the cultivated land quality prediction model; wherein the cultivated land quality prediction model performs feature extraction and feature operation on the key quality index data to obtain quality feature information with long-term dependence relationship, and converts the quality feature information into the cultivated land quality prediction result; obtaining yield index data related to cultivated land yield and yield historical research data related to cultivated land yield; constructing a yield estimation knowledge graph based on the yield index data and the yield historical research data to obtain key yield index data; inputting the key yield index data and the cultivated land quality prediction result into a cultivated land yield prediction model to obtain a cultivated land yield prediction result output by the cultivated land yield prediction model; wherein the cultivated land yield prediction model performs geographical weighted regression analysis on the key yield index data and the cultivated land quality prediction result to obtain the cultivated land yield prediction result.

[0006] According to the method for predicting the cultivated land yield provided by the application, the quality estimation knowledge graph is constructed based on the quality index data and the quality historical research data to obtain the key quality index data, which comprises the following steps: data cleaning is performed on the quality index data and the quality historical research data, and the quality standardization data is obtained by performing standardization processing on the cleaned quality index data and the quality historical research data; the quality estimation knowledge graph is obtained by performing knowledge extraction, knowledge fusion and knowledge processing on the quality standardization data, the quality estimation knowledge graph comprises the initial quality index data related to the cultivated land quality and the quality correlation degree, the quality correlation degree represents the correlation degree between the initial quality index data and the cultivated land quality, and the data amount of the initial quality index data is greater than that of the quality index data; and when the quality correlation degree is greater than a set value, the initial quality index data corresponding to the quality correlation degree is taken as the key quality index data.

[0007] According to the method for predicting the cultivated land yield provided by the application, the key yield index data comprises a plurality of key yield indexes, and the cultivated land yield prediction model is used to determine the cultivated land yield prediction result: the weight coefficient of each key yield index and the cultivated land quality prediction result is obtained according to the regression coefficient vector of the cultivated land yield prediction model; the plurality of key yield indexes and the cultivated land quality prediction result are weighted and summed according to the weight coefficient, and the cultivated land yield prediction result is obtained.

[0008] According to the method for predicting the cultivated land yield provided by the application, the cultivated land yield prediction model is obtained based on the following steps: the sample yield data is obtained according to the sample key yield index data and the sample cultivated land quality prediction result of the sample cultivated land; the sample yield data carrying labels is obtained by marking the sample yield data according to the actual cultivated land yield data of the sample cultivated land; the preset cultivated land yield model is trained according to the sample yield data carrying labels until the error of the prediction data output by the preset cultivated land yield model reaches a minimum value, and the preset cultivated land yield model comprises a preset spatial weight vector; the regression coefficient vector of the cultivated land yield prediction model is determined based on the sample yield data carrying labels and the preset spatial weight vector, so as to obtain the cultivated land yield prediction model.

[0009] The application provides a cultivated land yield prediction method, which is based on yield index data and yield historical research data to construct a yield estimation knowledge graph to obtain key yield index data, and comprises the following steps: performing data cleaning on the yield index data and the yield historical research data, and performing standardization processing on the cleaned yield index data and yield historical research data to obtain yield standardization data; performing knowledge extraction, knowledge fusion and knowledge processing on the yield standardization data to obtain a yield estimation knowledge graph, wherein the yield estimation knowledge graph comprises initial yield index data related to cultivated land yield and yield correlation degrees, the yield correlation degrees represent the correlation degrees between the initial yield index data and the cultivated land yield, and the data amount of the initial yield index data is greater than that of the yield index data; and when the yield correlation degree is greater than a set value, the initial yield index data corresponding to the yield correlation degree is taken as the key yield index data.

[0010] The application provides a cultivated land yield prediction method, wherein the key quality index data comprises key quality index data of multiple continuous time periods, the cultivated land quality prediction model is a long short-term memory model, and the cultivated land quality prediction model is used to obtain quality feature information with long-term dependence; feature extraction is performed on the key quality index data of the multiple continuous time periods to obtain quality input features of the multiple continuous time periods; the memory cell state of the current time period of the long short-term memory model is updated based on the operation result of the input gate of the long short-term memory model on the quality input features of the current time period, the output data of the previous time period and the hidden information of the previous time period; the forgetting data of the current time period and the reserved data of the current time period are determined based on the operation result of the forgetting gate of the long short-term memory model on the hidden information of the previous time period and the quality input features of the current time period, and the reserved data of the current time period is taken as the memory cell state of the current time period which is updated again; the output data of the current time period is determined based on the output gate of the long short-term memory model on the memory cell state of the current time period which is updated again, the quality input features of the current time period and the hidden information of the previous time period; the hidden information of the current time period is determined based on the output gate on the quality input features of the current time period, the hidden information of the previous time period and the memory cell state of the current time period which is updated again; and the current time period is taken as the previous time period until the output data of the last time period is obtained, and the quality feature information with long-term dependence is obtained according to the output data of the multiple continuous time periods.

[0011] The application provides a cultivated land yield prediction method, wherein the cultivated land quality prediction model is used to obtain a cultivated land quality prediction result; the weight of each quality feature in the quality feature information is obtained; each quality feature is weighted and summed based on the weight of the quality feature to obtain a cultivated land quality prediction value; and the cultivated land quality prediction result is obtained by processing the cultivated land quality prediction value according to an activation function.

[0012] According to the cultivated land yield prediction method provided by the application, the quality index data related to the quality of the cultivated land is obtained, including: at least based on the quality meteorological data, quality soil data and quality human index data of the cultivated land, the quality index data is obtained, the quality meteorological data includes precipitation and temperature, the quality soil data includes soil nutrient condition, soil structure, soil conductivity, soil PH value and water and soil conservation coefficient, and the quality human index data includes cultivated land investment, irrigation amount and contiguous degree.

[0013] According to the cultivated land yield prediction method provided by the application, the quality index data related to the quality of the cultivated land is obtained, including: at least based on the quality meteorological data, quality soil data and quality human index data of the cultivated land, the quality index data is obtained, the quality meteorological data includes precipitation and temperature, the quality soil data includes soil nutrient condition, soil structure, soil conductivity, soil PH value and water and soil conservation coefficient, and the quality human index data includes cultivated land investment, irrigation amount and contiguous degree.

[0014] The application further provides a cultivated land yield prediction device, comprising: a quality index data acquisition module, configured to acquire quality index data related to the quality of the cultivated land and quality historical research data related to the quality of the cultivated land; a quality estimation knowledge graph construction module, configured to construct a quality estimation knowledge graph based on the quality index data and the quality historical research data, so as to obtain key quality index data; a cultivated land quality prediction module, configured to input the key quality index data into a cultivated land quality prediction model, and acquire a cultivated land quality prediction result output by the cultivated land quality prediction model; wherein the cultivated land quality prediction model performs feature extraction and feature operation on the key quality index data, so as to obtain quality feature information with long-term dependence relationship, and converts the quality feature information into the cultivated land quality prediction result; a yield index data acquisition module, configured to acquire yield index data related to the yield of the cultivated land and yield historical research data related to the yield of the cultivated land; a yield estimation knowledge graph construction module, configured to construct a yield estimation knowledge graph based on the yield index data and the yield historical research data, so as to obtain key yield index data; a cultivated land yield prediction module, configured to input the key yield index data and the cultivated land quality prediction result into a cultivated land yield prediction model, and acquire a cultivated land yield prediction result output by the cultivated land yield prediction model; wherein the cultivated land yield prediction model performs geographical weighted regression analysis on the key yield index data and the cultivated land quality prediction result, so as to obtain the cultivated land yield prediction result.

[0015] The application further provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor implements any of the above cultivated land yield prediction methods when executing the program.

[0016] The application further provides a non-transitory computer-readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement any one of the above-mentioned cultivated land yield prediction methods.

[0017] The cultivated land yield prediction method, device, electronic equipment and storage medium provided by the application realize dynamic prediction of cultivated land quality based on deep learning by acquiring quality feature information with long-term dependence, improve the accuracy of determining the cultivated land quality prediction result. At the same time, the cultivated land yield prediction result is obtained according to the key yield index data and the cultivated land quality prediction result, effectively combining the cultivated land quality and the grain yield, which is conducive to comprehensively determining the cultivated land yield prediction result. The correlation between the key yield index data, the cultivated land quality prediction result and the cultivated land yield is determined by the geographic weighted regression analysis, which improves the accuracy of the cultivated land yield prediction result. The application uses a knowledge graph to significantly improve the accuracy and reliability of the cultivated land quality prediction model and the cultivated land yield prediction model by integrating multi-source data, providing background knowledge and assisting feature selection and the like. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0019] Figure 1 is one of the flowcharts of the cultivated land yield prediction method provided by the application.

[0020] Figure 2 is the second flowchart of the cultivated land yield prediction method provided by the application.

[0021] Figure 3 is the structural diagram of the long short-term memory neural network provided by the application.

[0022] Figure 4 is the structural diagram of the quality estimation knowledge graph provided by the application.

[0023] Figure 5 is the structural diagram of the cultivated land yield prediction device provided by the application.

[0024] Figure 6 is the structural diagram of the electronic equipment provided by the application. DETAILED DESCRIPTION

[0025] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below in conjunction with the accompanying drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0026] The technical solutions of the present application will be described below in conjunction with the accompanying drawings. Figures 1-6 The ploughland yield prediction method, device and electronic equipment of the present application are described.

[0027] The existing ploughland yield prediction method rarely uses new methods such as deep learning, lacks dynamic prediction, and does not model the ploughland quality and yield jointly, so the relationship between ploughland quality and yield improvement is not clear.

[0028] Figure 1 is one of the flowcharts of the ploughland yield prediction method provided by the present application, as Figure 1 shown, the method comprises the following steps.

[0029] S100: obtaining quality index data related to ploughland quality and quality historical research data related to ploughland quality.

[0030] The quality index data related to the quality of the ploughland is obtained, specifically, at least based on the quality meteorological data of the ploughland, the quality soil data of the ploughland and the quality human index data of the ploughland, the quality index data is obtained, the quality meteorological data includes precipitation and temperature, the quality soil data includes soil nutrient condition, soil structure, soil conductivity, soil PH value and water and soil conservation coefficient, and the quality human index data includes ploughland input, irrigation amount and contiguity.

[0031] The quality historical research data related to the quality of the ploughland is obtained, for example, the scientific research literature related to the quality of the ploughland is obtained.

[0032] The quality meteorological data of the ploughland to be evaluated is collected, specifically including precipitation data and temperature data. The digital elevation data of the ploughland to be evaluated is obtained. The quality soil data is obtained by field sampling, specifically including soil organic matter content, soil nutrient condition (for example, nitrogen content, available phosphorus content, available potassium content, etc.), soil texture, obstacle factor, soil bulk density, soil structure (for example, effective soil layer thickness, plough layer thickness, black soil layer thickness, soil configuration), soil PH value, conductivity, water content, water and soil conservation coefficient. The quality human index data is collected, specifically including ploughland input, irrigation amount and contiguity. Through the above-mentioned multi-element big data input, the problem of single consideration of influencing factors in the ploughland quality prediction process is solved.

[0033] S200: Construct a quality estimation knowledge graph based on the quality index data and the quality historical research data to obtain key quality index data.

[0034] The quality estimation knowledge graph is constructed based on the quality index data and the quality historical research data to obtain the key quality index data. Specifically, the quality index data and the quality historical research data are cleaned, and the cleaned quality index data and the quality historical research data are standardized to obtain quality standardized data. The quality standardized data is subjected to knowledge extraction, knowledge fusion and knowledge processing to obtain the quality estimation knowledge graph. The quality estimation knowledge graph includes initial quality index data related to the cultivated land quality and a quality correlation degree. The quality correlation degree represents the correlation degree between the initial quality index data and the cultivated land quality. The data volume of the initial quality index data is greater than that of the quality index data. When the quality correlation degree is greater than a set value, the initial quality index data corresponding to the quality correlation degree is taken as the key quality index data.

[0035] The quality index data and the quality historical research data are preprocessed, including missing value processing, abnormal value detection and processing, etc., to ensure the quality and integrity of the quality index data. Then the preprocessed quality index data and the quality historical research data are subjected to knowledge extraction, knowledge fusion and knowledge processing to obtain initial quality index data and a quality correlation degree. As shown in Figure 4 The initial quality index data and the quality correlation degree are graphically displayed to obtain the quality estimation knowledge graph. For example, the initial quality index data is taken as an entity point, the cultivated land quality is taken as a center point, and the distance between the entity point and the center point is drawn according to the size of the quality correlation degree. The greater the quality correlation degree of the entity point, the smaller the distance between the entity point and the center point. When the quality correlation degree is greater than a set value (the distance between the entity point and the center point is less than a set distance), the initial quality index data corresponding to the quality correlation degree is taken as the key quality index data.

[0036] S300: Input the key quality index data into the cultivated land quality prediction model to obtain a cultivated land quality prediction result output by the cultivated land quality prediction model.

[0037] The cultivated land quality prediction model extracts and operates features of the key quality index data to obtain quality feature information with long-term dependence relationship, and converts the quality feature information into the cultivated land quality prediction result.

[0038] The key quality indicator data includes key quality indicator data of multiple continuous time periods. The key quality indicator data of multiple continuous time periods is input into the cultivated land quality prediction model. The cultivated land quality prediction model is a long short-term memory model, which includes a feature extraction layer, a long short-term memory neural network (LSTM), and a fully connected layer. The key quality indicator data is input into the feature extraction layer (e.g., a convolutional layer) to convert the key quality indicator data into a more representative and high-level abstract feature representation, outputting quality input features corresponding to the key quality indicator data. The quality input features are input into the LSTM, outputting quality feature information with long-term dependencies. The quality feature information is input into the fully connected layer to obtain the cultivated land quality prediction result.

[0039] The cultivated land quality prediction model is trained by the following steps. Sample quality indicator data and corresponding sample cultivated land quality prediction results of multiple continuous time periods are collected. The sample key quality indicator data is labeled according to the sample cultivated land quality prediction results to obtain sample data. The sample data is divided into a training set (e.g., 60%), a test set (e.g., 20%), and a validation set (e.g., 20%) according to a prediction ratio. The cultivated land quality preset model is constructed according to the initial feature extraction layer, the initial long short-term memory neural network, and the initial fully connected layer. The cultivated land quality preset model is trained according to the training set, verified according to the validation set, and tested according to the test set. Various evaluation indicators, such as mean squared error (MSE), root mean squared error (RMSE), etc., are used to evaluate the cultivated land quality prediction performance of the cultivated land quality preset model. When the error of the cultivated land quality prediction result output by the cultivated land quality preset model meets the set condition, it is determined that the cultivated land quality preset model training is completed, and the cultivated land quality prediction model is obtained.

[0040] Further, the cultivated land quality prediction result of the cultivated land quality prediction model is explained. The change trend and influencing factors of the cultivated land quality are analyzed, and the cultivated land quality prediction result is presented in the form of statistical charts and maps using visualization tools, which more intuitively displays the spatial and temporal distribution of the cultivated land quality.

[0041] S400: Obtain yield indicator data related to cultivated land yield and yield historical research data related to cultivated land yield.

[0042] The yield index data related to the yield of the cultivated land is acquired. Specifically, the yield index data is acquired based on at least yield meteorological data, yield soil data and yield human index data of the cultivated land. The yield meteorological data includes precipitation, temperature, wind speed, annual sunshine duration and water evaporation amount. The yield soil data includes soil nutrient condition, soil structure, soil texture and soil pH value. The yield human index data includes grain sowing area, effective irrigation rate, agricultural labor force, agricultural machinery input, pesticide usage and film usage.

[0043] The yield meteorological data of the cultivated land to be evaluated is collected, specifically including precipitation, temperature, wind speed, water evaporation amount and annual sunshine duration data. The yield soil data is acquired by field sampling, specifically including soil organic matter content, soil nutrient condition, soil texture, soil structure (including soil bulk density, plough layer thickness, black soil layer thickness) and soil pH value. The yield human index data is acquired, specifically including grain sowing area, effective irrigation rate, agricultural labor force, agricultural machinery input, pesticide usage and film usage. Through the above multi-element data input, the problem of single consideration of the cultivated land yield influencing factor is solved.

[0044] The yield historical research data related to the yield of the cultivated land is acquired, for example, scientific research literature related to the yield of the cultivated land is acquired.

[0045] S500: Constructing a yield estimation knowledge graph based on the yield index data and the yield historical research data to obtain key yield index data.

[0046] The yield estimation knowledge graph is constructed based on the yield index data and the yield historical research data to obtain the key yield index data. Specifically, the yield index data and the yield historical research data are data cleaned, and the cleaned yield index data and yield historical research data are standardized to obtain yield standardized data. The yield standardized data is subjected to knowledge extraction, knowledge fusion and knowledge processing to obtain the yield estimation knowledge graph. The yield estimation knowledge graph includes initial yield index data related to the yield of the cultivated land and yield correlation degree. The yield correlation degree represents the correlation degree between the initial yield index data and the yield of the cultivated land. The data amount of the initial yield index data is greater than that of the yield index data. When the yield correlation degree is greater than a set value, the initial yield index data corresponding to the yield correlation degree is taken as the key yield index data.

[0047] The yield index data and yield historical research data are preprocessed, including missing value processing, abnormal value detection and processing, etc., to ensure the quality and integrity of the yield index data. Then the preprocessed yield index data and yield historical research data are knowledge extraction, knowledge fusion and knowledge processing, to obtain initial yield index data and yield correlation degree. The initial yield index data and yield correlation degree are graphically displayed to obtain the yield estimation knowledge graph. For example, the initial yield index data is taken as an entity point, the arable land yield is taken as a center point, and the distance between the entity point and the center point is drawn according to the size of the yield correlation degree. The greater the yield correlation degree of the entity point, the smaller the distance between the entity point and the center point. When the yield correlation degree is greater than a set value (the distance between the entity point and the center point is less than a set distance), the initial yield index data corresponding to the yield correlation degree is taken as the key yield index data.

[0048] S600: input the key yield index data and the arable land quality prediction result into the arable land yield prediction model to obtain the arable land yield prediction result output by the arable land yield prediction model.

[0049] The arable land yield prediction model performs geographical weighted regression analysis on the key yield index data and the arable land quality prediction result to obtain the arable land yield prediction result.

[0050] The arable land quality prediction result and the key yield index data are coupled to obtain yield coupling data. The yield coupling data is input into the arable land yield prediction model.

[0051] The application couples the arable land quality prediction result output by the arable land quality prediction model with the key yield index data, solves the problem that the arable land quality and yield estimation cannot be effectively combined for modeling, and the correlation between the arable land quality and the grain yield improvement is not clear enough.

[0052] As shown in Figure 2 , the key yield index data includes multiple key yield indicators, and the arable land yield prediction model is used to determine the arable land yield prediction result: according to the regression coefficient vector of the arable land yield prediction model, the weight coefficient of each key yield indicator and the arable land quality prediction result is obtained; the multiple key yield indicators and the arable land quality prediction result are weighted and summed according to the weight coefficient to obtain the arable land yield prediction result.

[0053] The calculation formula of the arable land yield prediction result is as follows.

[0054] ; Wherein, is the arable land yield prediction result, is a vector composed of the key yield indicators and the arable land quality prediction result, is a regression coefficient vector composed of multiple weight coefficients.

[0055] The cultivated land yield prediction model of the present application is a yield estimation model constructed using geographically weighted regression (GWR). The cultivated land yield prediction model analyzes through geographically weighted regression, and calculates a regression coefficient vector to assign weights to each key yield indicator and quality prediction result input by this method. The key yield indicators and quality prediction results after being assigned weights are summed to obtain the cultivated land yield prediction result.

[0056] Further, the cultivated land yield prediction result of the cultivated land yield prediction model is explained. The change trend and influencing factors of cultivated land yield are analyzed, and the cultivated land yield prediction result is presented in the form of statistical charts and maps using visualization tools, which more intuitively displays the spatial and temporal distribution of cultivated land yield.

[0057] The cultivated land yield prediction method provided by the embodiments of the present application realizes dynamic prediction of cultivated land quality based on deep learning by obtaining quality characteristic information with long-term dependence, improves the accuracy of determining the cultivated land quality prediction result. At the same time, the cultivated land yield prediction result is obtained according to the key yield indicator data and the cultivated land quality prediction result, effectively combining the cultivated land quality and the grain yield, which is conducive to comprehensively determining the cultivated land yield prediction result. The correlation between the key yield indicator data, the cultivated land quality prediction result and the cultivated land yield is determined through geographically weighted regression analysis, which improves the accuracy of the cultivated land yield prediction result. The present application uses a knowledge graph to integrate multi-source data, provide background knowledge and assist feature selection, etc., which significantly improves the accuracy and reliability of the cultivated land quality prediction model and the cultivated land yield prediction model.

[0058] At present, large-scale cultivated land quality evaluation and grain yield prediction mainly rely on meteorological and remote sensing data, and the influencing factors are relatively single. However, the quality indicator data and yield indicator data selected by the present application involve various aspects, and the use of multi-element networking observation to obtain big data and multi-factor modeling makes the model more reliable. Using the knowledge graph to determine the key quality indicator data is conducive to improving the accuracy of the cultivated land quality prediction model. Using the knowledge graph to determine the key yield indicator data is conducive to improving the accuracy of the cultivated land yield prediction model. The cultivated land quality prediction model uses LSTM deep learning network modeling to predict future results, which solves the problems of poor effect of single model, less application of deep learning and lack of dynamic prediction. Coupling the cultivated land quality prediction result in the cultivated land yield prediction model helps to clarify the internal relationship between cultivated land quality and grain yield improvement, so that the yield can be more accurately estimated based on the dynamic evolution of cultivated land quality.

[0059] Based on the above embodiments, the cultivated land yield prediction model is obtained based on the following steps.

[0060] S410: Sample yield data are obtained based on the sample key yield index data and the sample arable land quality prediction results.

[0061] S420: Label the sample yield training data based on the actual cultivated land yield data of the sample cultivated land to obtain labeled sample yield data.

[0062] S430: Train the preset farmland yield model based on the labeled sample yield data until the error of the predicted data output by the preset farmland yield model reaches the minimum value. The preset farmland yield model includes a preset spatial weight vector.

[0063] S440: Based on the labeled sample yield data and the preset spatial weight vector, determine the regression coefficient vector of the cultivated land yield prediction model to obtain the cultivated land yield prediction model.

[0064] Obtain sample yield data from multiple sample farmlands. Label the sample yield data according to the actual farmland quality data of each sample farmland to obtain multiple labeled sample yield data (multiple sample points). Divide the sample yield data into a training set (e.g., 60%), a test set (e.g., 20%), and a validation set (e.g., 20%) according to the prediction ratio.

[0065] The preset farmland yield model is trained, tested, and validated using training, testing, and validation sets until the error of the farmland yield prediction result output by the preset farmland yield model is less than the set error.

[0066] The presupposed model for arable land yield is the GWR model. The GWR model incorporates spatial relationship changes caused by spatial location differences into the calculation of regression coefficients. The formula for calculating the arable land yield prediction results using the GWR model is as follows.

[0067] ; in, In order to be in Actual cultivated land yield data at the location, In order to be in Key yield data and farmland quality prediction results for the sample at the location. For the intercept term, For the first The regression coefficient vector of each sample point, which is related to the spatial location (Location (related to coordinates) The number of regression coefficients Is in Error term at position, ,and .

[0068] The regression coefficient vector is solved using the weighted least squares method. For each sample point... Establish the objective function.

[0069] ; in, In the first The sample point and the first Each sample point is located at The spatial weight at position is the first The sample point and the first A monotonically decreasing function of the spatial distance between sample points. In order to be in Actual cultivated land yield data at the location, In order to be in Key yield data and farmland quality prediction results for the sample at the location. For the intercept term, For the first The regression coefficient vector of each sample point For bandwidth, a parameter describing the non-negative decreasing relationship between spatial weights and distance. For the first The sample point and the first Spatial distance between sample points.

[0070] When the error between the predicted data and the actual cultivated land yield data from the preset cultivated land yield model reaches its minimum, the objective function can be solved. The formula for calculating the regression coefficient vector is as follows.

[0071] ; in, For the regression coefficient vector, In order to be in Spatial weight at location, for Preset spatial weight vector at the location, In order to be in Location (the first) Actual cultivated land yield data (from 1 sample point) For the first A vector consisting of key yield indicators for each sample point and the predicted farmland quality for each sample point.

[0072] The embodiments of the present invention determine the regression coefficient vector through model training, and comprehensively consider the correlation between the sample arable land quality prediction results and the sample key yield indicator data, which is conducive to improving the accuracy of arable land yield prediction results.

[0073] Based on the above embodiment, the key quality indicator data includes key quality indicator data of a plurality of continuous time periods, the cultivated land quality prediction model is a long short-term memory model, the cultivated land quality prediction model is used to obtain quality feature information with long-term dependence, and includes steps S210 to S250, and each step is specifically as follows.

[0074] S210: performing feature extraction on the key quality indicator data of the plurality of continuous time periods to obtain quality input features of the plurality of continuous time periods.

[0075] S220: updating a memory cell state of a current time period of the long short-term memory model based on an operation result of the quality input features of the current time period, output data of a previous time period and hidden information of the previous time period by an input gate of the long short-term memory model.

[0076] S230: determining forgetting data of the current time period and reserved data of the current time period based on an operation result of the hidden information of the previous time period and the quality input features of the current time period by a forgetting gate of the long short-term memory model, and taking the reserved data of the current time period as the memory cell state of the current time period which is updated again.

[0077] S240: determining the output data of the current time period based on the memory cell state of the current time period which is updated again, the quality input features of the current time period and the hidden information of the previous time period by an output gate of the long short-term memory model, and determining the hidden information of the current time period based on the quality input features of the current time period, the hidden information of the previous time period and the memory cell state of the current time period which is updated again by the output gate.

[0078] S250: taking the current time period as the previous time period until output data of a last time period is obtained, and obtaining the quality feature information with long-term dependence based on the output data of the plurality of continuous time periods.

[0079] The cultivated land quality prediction model is used to obtain a cultivated land quality prediction result, weights of each quality feature in the quality feature information are obtained, each quality feature is weighted and summed based on the weight of the quality feature to obtain a cultivated land quality prediction value, and the cultivated land quality prediction value is processed according to an activation function to obtain the cultivated land quality prediction result.

[0080] The key quality indicator data of the plurality of continuous time periods is feature-extracted by the feature extraction layer to obtain the quality input features of the plurality of continuous time periods. Figure 2 、 3 As shown in FIGS. 1 to 3, the long short-term memory (LSTM) model of the present application includes an input gate, a forgetting gate and an output gate.

[0081] The input gate is used to determine which data needs to be input data, which consists of an activation function (sigmoid function) and a point multiplication operation. The quality input features (X t ) of the current time period, the output data (c t-1 ) of the previous time period, and the hidden information (h t-1 ) of the previous time period are input into the input gate to update the memory cell state of the current time period of the long short-term memory model. For example, X t of 2001 (previous time period), c t-1 of 2001, and h t-1 of 2001 are input into the input gate to obtain the memory cell state of 2002 (current time period).

[0082] The forget gate is used to determine which data needs to be retained or forgotten, which consists of an activation function and a point multiplication operation. The hidden information (h t-1 ) of the previous time period and the quality input features (X t ) of the current time period are input into the forget gate to obtain the retained data of the current time period and the forgotten data of the current time period, and the retained data of the current time period is taken as the updated memory cell state of the current time period.

[0083] The output gate is used to determine which data needs to be the final output feature information, which consists of an activation function and a point multiplication operation. The updated memory cell state of the current time period, the quality input features of the current time period, and the hidden information of the previous time period are input into the output gate to obtain the output data of the current time period. The quality input features (X t ) of the current time period, the hidden information (h t-1 ) of the previous time period, and the updated memory cell state of the current time period are input into the output gate to obtain the hidden information (h t ) of the current time period.

[0084] Similarly, until the output data of the last time period is obtained, the final quality feature information with long-term dependence is obtained according to the output data of multiple consecutive time periods.

[0085] The final quality feature information with long-term dependence is input into the fully connected layer of the cultivated land quality prediction model. The fully connected layer performs weighted summation on each quality feature in the quality feature information to obtain the cultivated land quality prediction value. The cultivated land quality prediction value is activated by an activation function to obtain the cultivated land quality prediction result.

[0086] The embodiment of the application processes key quality indicator data of multiple continuous time periods through a long short-term memory neural network to obtain quality characteristic information with long-term dependence, considers the relationship of key quality indicator data of different time periods, and is beneficial to improving the accuracy of cultivated land quality prediction results.

[0087] The cultivated land yield prediction device provided by the application is described below, and the cultivated land yield prediction device described below can be correspondingly referred to the cultivated land yield prediction method described above.

[0088] As shown in Figure 5 A cultivated land yield prediction device includes: a quality indicator data acquisition module 501 configured to acquire quality indicator data related to cultivated land quality and quality historical research data related to cultivated land quality.

[0089] A quality estimation knowledge graph construction module 502 is configured to construct a quality estimation knowledge graph based on the quality indicator data and the quality historical research data to obtain key quality indicator data.

[0090] A cultivated land quality prediction module 503 is configured to input the key quality indicator data into a cultivated land quality prediction model to obtain a cultivated land quality prediction result output by the cultivated land quality prediction model; wherein the cultivated land quality prediction model performs feature extraction and feature operation on the key quality indicator data to obtain quality characteristic information with long-term dependence, and converts the quality characteristic information into the cultivated land quality prediction result.

[0091] A yield indicator data acquisition module 504 is configured to acquire yield indicator data related to cultivated land yield and yield historical research data related to cultivated land yield.

[0092] A yield estimation knowledge graph construction module 505 is configured to construct a yield estimation knowledge graph based on the yield indicator data and the yield historical research data to obtain key yield indicator data.

[0093] A cultivated land yield prediction module 506 is configured to input the key yield indicator data and the cultivated land quality prediction result into a cultivated land yield prediction model to obtain a cultivated land yield prediction result output by the cultivated land yield prediction model; wherein the cultivated land yield prediction model performs geographical weighted regression analysis on the key yield indicator data and the cultivated land quality prediction result to obtain the cultivated land yield prediction result.

[0094] The cultivated land yield prediction device provided by the embodiment of the present application realizes dynamic prediction of cultivated land quality based on deep learning by acquiring quality characteristic information with long-term dependence, and improves the accuracy of the determined cultivated land quality prediction result. Meanwhile, the cultivated land yield prediction result is acquired according to the key yield index data and the cultivated land quality prediction result, effectively combining the cultivated land quality and the grain yield, which is conducive to comprehensively determining the cultivated land yield prediction result. The correlation between the key yield index data, the cultivated land quality prediction result and the cultivated land yield is determined through the geographic weighted regression analysis, improving the accuracy of the cultivated land yield prediction result. The present application uses the knowledge graph to significantly improve the accuracy and reliability of the cultivated land quality prediction model and the cultivated land yield prediction model by integrating multi-source data, providing background knowledge and assisting feature selection and the like.

[0095] In one embodiment, the quality estimation knowledge graph construction module 502 is configured to: perform data cleaning on the quality index data and the quality historical research data, perform standardization processing on the cleaned quality index data and the quality historical research data to obtain quality standardized data; perform knowledge extraction, knowledge fusion and knowledge processing on the quality standardized data to obtain a quality estimation knowledge graph, the quality estimation knowledge graph including initial quality index data related to cultivated land quality and a quality correlation degree, the quality correlation degree representing the correlation degree between the initial quality index data and the cultivated land quality, and the data amount of the initial quality index data being greater than that of the quality index data; and when the quality correlation degree is greater than a set value, taking the initial quality index data corresponding to the quality correlation degree as key quality index data.

[0096] In one embodiment, the key yield index data includes a plurality of key yield indexes, and the cultivated land yield prediction model is configured to determine the cultivated land yield prediction result by: acquiring a weight coefficient of each key yield index and the cultivated land quality prediction result according to a regression coefficient vector of the cultivated land yield prediction model; and performing weighted summation on the plurality of key yield indexes and the cultivated land quality prediction result according to the weight coefficient to obtain the cultivated land yield prediction result.

[0097] In one embodiment, the cultivated land yield prediction module 506 is configured to: obtain sample yield data according to sample key yield index data and sample cultivated land quality prediction result of a sample cultivated land; label the sample yield data according to actual cultivated land yield data of the sample cultivated land to obtain sample yield data carrying labels; train a preset cultivated land yield model according to the sample yield data carrying labels until the error of prediction data output by the preset cultivated land yield model reaches a minimum value, the preset cultivated land yield model including a preset spatial weight vector; and determine a regression coefficient vector of the cultivated land yield prediction model based on the sample yield data carrying labels and the preset spatial weight vector to obtain the cultivated land yield prediction model.

[0098] In an embodiment, the yield estimation knowledge graph construction module 505 is configured to: perform data cleaning on the yield indicator data and the yield historical research data, perform standardization processing on the cleaned yield indicator data and the yield historical research data to obtain yield standardized data; perform knowledge extraction, knowledge fusion and knowledge processing on the yield standardized data to obtain a yield estimation knowledge graph, the yield estimation knowledge graph including initial yield indicator data related to the cultivated land yield and a yield correlation degree, the yield correlation degree representing a correlation degree between the initial yield indicator data and the cultivated land yield, and a data volume of the initial yield indicator data being greater than a data volume of the yield indicator data; and when the yield correlation degree is greater than a set value, taking the initial yield indicator data corresponding to the yield correlation degree as key yield indicator data.

[0099] In an embodiment, the key quality indicator data includes key quality indicator data of a plurality of continuous time periods, the cultivated land quality prediction model is a long short-term memory model, and the cultivated land quality prediction model is configured to obtain quality feature information with long-term dependence: performing feature extraction on the key quality indicator data of the plurality of continuous time periods to obtain quality input features of the plurality of continuous time periods; updating a memory cell state of a current time period of the long short-term memory model based on an operation result of a current time period quality input feature, a previous time period output data and a previous time period hidden information of an input gate of the long short-term memory model; determining a forgetting data of the current time period and a reserved data of the current time period based on an operation result of the previous time period hidden information and the current time period quality input feature of a forgetting gate of the long short-term memory model, and taking the reserved data of the current time period as the memory cell state of the current time period which is updated again; determining the output data of the current time period based on the memory cell state of the current time period which is updated again, the current time period quality input feature and the previous time period hidden information of an output gate of the long short-term memory model; determining the hidden information of the current time period based on the current time period quality input feature, the previous time period hidden information and the memory cell state of the current time period which is updated again of the output gate; and taking the current time period as the previous time period until the output data of the last time period is obtained, and obtaining the quality feature information with long-term dependence based on the output data of the plurality of continuous time periods.

[0100] In an embodiment, the cultivated land quality prediction model is configured to obtain a cultivated land quality prediction result: obtaining a weight of each quality feature in the quality feature information; performing weighted summation on each quality feature based on the weight of the quality feature to obtain a cultivated land quality prediction value; and processing the cultivated land quality prediction value according to an activation function to obtain the cultivated land quality prediction result.

[0101] In one embodiment, the quality index data acquisition module 501 is configured to acquire quality index data based on at least quality meteorological data of the cultivated land, quality soil data of the cultivated land, and quality human index data of the cultivated land, wherein the quality meteorological data includes precipitation and temperature, the quality soil data includes soil nutrient condition, soil structure, soil conductivity, soil PH value, and water and soil conservation coefficient, and the quality human index data includes cultivated land investment, irrigation amount, and contiguity.

[0102] In one embodiment, the yield index data acquisition module 504 is configured to acquire yield index data based on at least yield meteorological data of the cultivated land, yield soil data of the cultivated land, and yield human index data of the cultivated land, wherein the yield meteorological data includes precipitation, temperature, wind speed, annual sunshine duration, and water evaporation amount, the yield soil data includes soil nutrient condition, soil structure, soil texture, and soil PH value, and the yield human index data includes grain sowing area, effective irrigation rate, agricultural labor force, agricultural machinery investment, pesticide usage amount, and film usage amount.

[0103] Figure 6 An example of a schematic diagram of a physical structure of an electronic device is shown in FIG. 6. Figure 6 As shown in FIG. 6, the electronic device can include a processor 610, a communications interface 620, a memory 630, and a communications bus 640, wherein the processor 610, the communications interface 620, and the memory 630 can communicate with each other through the communications bus 640. The processor 610 can invoke a logical instruction in the memory 630 to execute a cultivated land yield prediction method, which includes: acquiring quality index data related to cultivated land quality and quality historical research data related to cultivated land quality; constructing a quality estimation knowledge graph based on the quality index data and the quality historical research data to obtain key quality index data; inputting the key quality index data into a cultivated land quality prediction model to obtain a cultivated land quality prediction result output by the cultivated land quality prediction model; wherein the cultivated land quality prediction model performs feature extraction and feature operation on the key quality index data to obtain quality feature information with long-term dependence, and converts the quality feature information into the cultivated land quality prediction result; acquiring yield index data related to cultivated land yield and yield historical research data related to cultivated land yield; constructing a yield estimation knowledge graph based on the yield index data and the yield historical research data to obtain key yield index data; inputting the key yield index data and the cultivated land quality prediction result into a cultivated land yield prediction model to obtain a cultivated land yield prediction result output by the cultivated land yield prediction model; wherein the cultivated land yield prediction model performs geographical weighted regression analysis on the key yield index data and the cultivated land quality prediction result to obtain the cultivated land yield prediction result.

[0104] In addition, the logic instructions in the memory 630 described above can be implemented in the form of a software function unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0105] In another aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program, which, when executed by a processor, implements a cultivated land yield prediction method provided by the above method, the method comprising: obtaining quality index data related to cultivated land quality and quality historical research data related to cultivated land quality; constructing a quality estimation knowledge graph based on the quality index data and the quality historical research data to obtain key quality index data; inputting the key quality index data into a cultivated land quality prediction model to obtain a cultivated land quality prediction result output by the cultivated land quality prediction model; wherein the cultivated land quality prediction model performs feature extraction and feature operation on the key quality index data to obtain quality feature information with long-term dependence, and converts the quality feature information into the cultivated land quality prediction result; obtaining yield index data related to cultivated land yield and yield historical research data related to cultivated land yield; constructing a yield estimation knowledge graph based on the yield index data and the yield historical research data to obtain key yield index data; inputting the key yield index data and the cultivated land quality prediction result into a cultivated land yield prediction model to obtain a cultivated land yield prediction result output by the cultivated land yield prediction model; wherein the cultivated land yield prediction model performs geographical weighted regression analysis on the key yield index data and the cultivated land quality prediction result to obtain the cultivated land yield prediction result.

[0106] The device embodiments described above are only schematic, wherein the units illustrated as separate components can or can not be physically separate, and the components illustrated as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment. Those skilled in the art can understand and implement without creative labor.

[0107] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0108] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method of predicting the yield of a cultivated property, characterized by, The method comprises the following steps: obtaining quality index data related to arable land quality and quality historical research data related to arable land quality; constructing a quality estimation knowledge graph based on the quality index data and the quality historical research data to obtain key quality index data; inputting the key quality index data into an arable land quality prediction model to obtain an arable land quality prediction result output by the arable land quality prediction model; wherein the arable land quality prediction model performs feature extraction and feature operation on the key quality index data to obtain quality feature information with long-term dependence, and converts the quality feature information into the arable land quality prediction result; obtaining yield index data related to arable land yield and yield historical research data related to arable land yield; constructing a yield estimation knowledge graph based on the yield index data and the yield historical research data to obtain key yield index data; inputting the key yield index data and the arable land quality prediction result into an arable land yield prediction model to obtain an arable land yield prediction result output by the arable land yield prediction model; wherein the arable land yield prediction model performs geographical weighted regression analysis on the key yield index data and the arable land quality prediction result to obtain the arable land yield prediction result.

2. The cultivated land yield prediction method according to claim 1, characterized by, The method comprises the following steps: performing data cleaning on the quality index data and the quality historical research data, and performing standardization processing on the cleaned quality index data and the quality historical research data to obtain quality standardized data; performing knowledge extraction, knowledge fusion and knowledge processing on the quality standardized data to obtain the quality estimation knowledge graph, wherein the quality estimation knowledge graph comprises initial quality index data related to arable land quality and quality correlation degree, the quality correlation degree represents the correlation degree between the initial quality index data and the arable land quality, and the data amount of the initial quality index data is greater than that of the quality index data; when the quality correlation degree is greater than a set value, the initial quality index data corresponding to the quality correlation degree is taken as the key quality index data.

3. The cultivated yield prediction method according to claim 1, characterized by, The key yield index data comprises a plurality of key yield indicators, and the arable land yield prediction model is used to determine the arable land yield prediction result: obtaining a weight coefficient of each key yield indicator and the arable land quality prediction result according to a regression coefficient vector of the arable land yield prediction model; performing weighted summation on a plurality of key yield indicators and the arable land quality prediction result according to the weight coefficient to obtain the arable land yield prediction result.

4. The cultivated yield prediction method according to claim 1, characterized by, The arable land yield prediction model is obtained based on the following steps: obtaining sample yield data according to sample key yield index data and sample arable land quality prediction results of sample arable land; labeling the sample yield data according to actual arable land yield data of the sample arable land to obtain sample yield data carrying labels; The preset cultivated land yield model is trained according to the sample yield data carrying labels until an error of predicted data output by the preset cultivated land yield model reaches a minimum value, and the preset cultivated land yield model comprises a preset spatial weight vector; Based on the sample yield data carrying labels and the preset spatial weight vector, a regression coefficient vector of the cultivated land yield prediction model is determined to obtain the cultivated land yield prediction model.

5. The cultivated yield prediction method according to claim 1, characterized by, The yield estimation knowledge graph is constructed based on the yield index data and the yield historical research data to obtain key yield index data, comprising: The yield index data and the yield historical research data are subjected to data cleaning, and the cleaned yield index data and yield historical research data are subjected to standardization processing to obtain yield standardized data; The yield standardized data is subjected to knowledge extraction, knowledge fusion and knowledge processing to obtain the yield estimation knowledge graph, the yield estimation knowledge graph comprising initial yield index data related to cultivated land yield and yield correlation degree, the yield correlation degree representing a correlation degree between the initial yield index data and the cultivated land yield, and a data volume of the initial yield index data being greater than a data volume of the yield index data; When the yield correlation degree is greater than a set value, the initial yield index data corresponding to the yield correlation degree is taken as the key yield index data.

6. The cultivated yield prediction method according to claim 1, characterized by, The key quality index data comprises key quality index data of multiple continuous time periods, the cultivated land quality prediction model is a long short-term memory model, and the cultivated land quality prediction model is used to obtain the quality feature information with long-term dependence: Feature extraction is performed on the key quality index data of multiple continuous time periods to obtain quality input features of multiple continuous time periods; Based on an operation result of the quality input features of a current time period, output data of a previous time period and hidden information of the previous time period by an input gate of the long short-term memory model, a memory cell state of the current time period of the long short-term memory model is updated; Based on an operation result of the hidden information of the previous time period and the quality input features of the current time period by a forget gate of the long short-term memory model, forgetting data of the current time period and reserved data of the current time period are determined, and the reserved data of the current time period is taken as the memory cell state of the current time period which is updated again; Based on the memory cell state of the current time period which is updated again, the quality input features of the current time period and the hidden information of the previous time period by an output gate of the long short-term memory model, output data of the current time period is determined; and based on the quality input features of the current time period, the hidden information of the previous time period and the memory cell state of the current time period which is updated again by the output gate, hidden information of the current time period is determined; The current time period is taken as the previous time period until output data of a last time period is obtained, and the quality feature information with long-term dependence is obtained according to the output data of the multiple continuous time periods.

7. The cultivated yield prediction method according to claim 1, characterized by, The cultivated land quality prediction model is used to obtain the cultivated land quality prediction result: obtaining a weight of each quality feature in the quality feature information; performing weighted summation on each of the quality features based on the weight of the quality feature to obtain a cultivated land quality prediction value; processing the cultivated land quality prediction value according to an activation function to obtain the cultivated land quality prediction result.

8. The cultivated yield prediction method according to claim 1, characterized by, The quality index data related to the quality of cultivated land comprises: The quality index data is obtained based on at least quality meteorological data, quality soil data and quality human index data of the cultivated land, the quality meteorological data comprises precipitation and temperature, the quality soil data comprises soil nutrient condition, soil structure, soil conductivity, soil PH value and water and soil conservation coefficient, and the quality human index data comprises cultivated land input, irrigation amount and contiguity.

9. The cultivated yield prediction method according to claim 1, characterized by, The yield index data related to the yield of cultivated land comprises: The yield index data is obtained based on at least yield meteorological data, yield soil data and yield human index data of the cultivated land, the yield meteorological data comprises precipitation, temperature, wind speed, annual sunshine duration and water evaporation amount, the yield soil data comprises soil nutrient condition, soil structure, soil texture and soil PH value, and the yield human index data comprises grain sowing area, effective irrigation rate, agricultural labor force, agricultural machinery input, pesticide usage and film usage.

10. A farmland yield prediction device characterized by comprising: The method comprises: a quality index data obtaining module configured to obtain quality index data related to the quality of cultivated land and quality historical research data related to the quality of cultivated land; a quality estimation knowledge graph construction module configured to construct a quality estimation knowledge graph based on the quality index data and the quality historical research data to obtain key quality index data; a cultivated land quality prediction module configured to input the key quality index data into a cultivated land quality prediction model to obtain a cultivated land quality prediction result output by the cultivated land quality prediction model, wherein the cultivated land quality prediction model performs feature extraction and feature operation on the key quality index data to obtain quality feature information with long-term dependence, and converts the quality feature information into the cultivated land quality prediction result; a yield index data obtaining module configured to obtain yield index data related to the yield of cultivated land and yield historical research data related to the yield of cultivated land; a yield estimation knowledge graph construction module configured to construct a yield estimation knowledge graph based on the yield index data and the yield historical research data to obtain key yield index data; a cultivated land yield prediction module configured to input the key yield index data and the cultivated land quality prediction result into a cultivated land yield prediction model to obtain a cultivated land yield prediction result output by the cultivated land yield prediction model, wherein the cultivated land yield prediction model performs geographical weighted regression analysis on the key yield index data and the cultivated land quality prediction result to obtain the cultivated land yield prediction result.

11. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the cultivated land yield prediction method according to any one of claims 1 to 9.

12. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the cultivated land yield prediction method according to any one of claims 1 to 9.

Images