Maize missing character data filling and yield measuring and calculating method and device

Abstract

The invention provides a corn missing character data filling and yield measuring and calculating method and device, electronic equipment and a storage medium, and relates to the technical field of intelligent agricultural information processing, and the method comprises the steps: obtaining corn planting data, and converting the corn planting data into a bipartite graph data structure; constructing a bipartite graph neural network based on the bipartite graph data structure, constructing a loss function of a corn missing character data filling and yield measuring and calculating model by using a gradient equilibrium mechanism, performing training, and predicting to-be-detected data by using the trained corn missing character data filling and yield measuring and calculating model to obtain a prediction result. The sample weight is dynamically adjusted by calculating the difference between the estimated value and the true value, so that the influence on the model performance caused by the problem of sample imbalance among the planting sites is effectively reduced, the agricultural production efficiency is improved, and the loss caused by overlarge yield estimation deviation of a traditional method is reduced.

Description

Technical Field

[0001] The present invention relates to the field of intelligent agricultural information processing technology, and in particular to a method, device, electronic equipment and storage medium for filling missing trait data and estimating yield of corn. Background Art

[0002] Corn is China's largest food crop, and its yield is of great significance to food security and agricultural policy formulation. The factors affecting corn yield are very complex. Temperature, precipitation, soil, humidity and other meteorological factors, as well as the interaction between these meteorological factors, will significantly affect corn yield. In addition, the performance of corn traits during growth (such as plant height, ear length, ear height, number of kernels per row, etc.) can also largely reflect corn yield.

[0003] In recent years, the application of machine learning and deep learning technologies in the field of crop yield prediction has made significant progress. Specifically, models including multilayer perceptrons, long short-term memory networks, and convolutional neural networks have been developed for accurate yield prediction. Although these models have achieved good prediction results in specific areas, they are usually trained and tested based on data from a fixed area, which limits the generalization ability of the model and its application in different geographical environments. At the same time, there is a spatial correlation between different corn planting locations, and there is an intrinsic connection between different traits of corn. Current prediction models often do not fully consider this spatial correlation and the relationship between traits. At the same time, the collection of corn trait data has problems with missing data and sample imbalance. If these problems are not properly resolved, they will affect the accuracy of the prediction results. Summary of the invention

[0004] The present invention provides a method and device for filling missing trait data and estimating yield of corn, which are used to solve the problems that spatial association and relationship between traits are often not fully considered in related technologies, and the collection of corn trait data has data missing and sample imbalance.

[0005] The present invention provides a method for filling missing trait data and estimating yield of corn, comprising the following steps: Acquire corn planting data, and convert the corn planting data into a bipartite graph data structure; Constructing a bipartite graph neural network based on the bipartite graph data structure to obtain a corn missing trait data filling and yield estimation model; A loss function of the corn missing trait data filling and yield estimation model is constructed by using a gradient balance mechanism, and the corn missing trait data filling and yield estimation model with the loss function is trained by using a corn planting data training set to obtain a trained corn missing trait data filling and yield estimation model; The trained corn missing trait data filling and yield estimation model is used to predict the data to be tested to obtain a prediction result.

[0006] According to a method for filling missing trait data and estimating yield of corn provided by the present invention, corn planting data is obtained and converted into a bipartite graph data structure, including: Meteorological data and corn trait data of corn planting sites are collected, the value of each node is initialized using encoding, and the data in the original table is converted into edge weights to obtain a bipartite graph data structure.

[0007] According to a method for filling missing trait data and estimating yield of corn provided by the present invention, the value of each node is initialized by coding, and the data in the original table is converted into the weight of the edge, including: The meteorological characteristics of the planting area, corn plant height, ear position, growth period, and yield-related characteristics during corn breeding were collected to construct the original data set; Using the feature name of the original data set as the first type of node of the bipartite graph data structure, using the planting data number of the original data set as the second type of node of the bipartite graph data structure, and using the data value in the original data set as the weight of the edge in the bipartite graph data structure; The values ​​in the first type of nodes and the second type of nodes are initialized using a hot encoding method to obtain a bipartite graph data structure.

[0008] According to a method for filling missing trait data and estimating yield of corn provided by the present invention, a bipartite graph neural network is constructed based on the bipartite graph data structure to obtain a corn estimation model, including: Constructing a bipartite graph neural network consisting of a graph update layer and a prediction layer based on the bipartite graph data structure; In the graph update layer, edge embedding is used to update node features, and edge weights are updated based on the node features; New feature vectors are filled for missing data values ​​based on the node features and the edge weights, and the node features obtained from the last layer of the graph update layer are input into the prediction layer to obtain a corn missing trait data filling and yield estimation model.

[0009] A method for filling missing trait data and estimating yield of corn provided by the present invention also includes: The bipartite graph neural network consists of three graph update layers and one prediction layer. It uses edge embedding to update node features, reversely updates edge weights based on node features, and uses a fully connected layer to complete the construction of corn missing trait data filling and yield estimation model.

[0010] According to a method for filling missing trait data and estimating yield of corn provided by the present invention, a loss function of the missing trait data filling and yield estimation model of corn is constructed by using a gradient balance mechanism, including: using a gradient balance mechanism to embed an absolute value error loss function, and the calculation formula thereof is:

[0011] in, represents the number of training samples, represents an estimate of production, represents the true value of output, express Loss, which measures the difference between the predicted value and the true value, is a constant.

[0012] A method for filling missing trait data and estimating yield of corn provided by the present invention also includes: constructing a graphical interface to display missing trait filling results and yield prediction results.

[0013] The present invention also provides a device for filling missing trait data and estimating yield of corn, comprising the following modules: A data acquisition module, which acquires corn planting data and converts the corn planting data into a bipartite graph data structure; A model building module, which builds a bipartite graph neural network based on the bipartite graph data structure to obtain a corn missing trait data filling and yield estimation model; A model training module, which uses a gradient balancing mechanism to construct a loss function of the corn missing trait data filling and yield estimation model, and uses a corn planting data training set to train the corn missing trait data filling and yield estimation model with the loss function constructed, to obtain a trained corn missing trait data filling and yield estimation model; The prediction module uses the trained corn missing trait data filling and yield estimation model to predict the data to be tested to obtain a prediction result.

[0014] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, any of the above methods is implemented.

[0015] The present invention also provides a computer program product, comprising a computer program, and when the computer program is executed by a processor, any of the above methods is implemented.

[0016] The present invention provides a method for filling missing trait data and estimating yield of corn, which respectively uses original data feature names and planting data numbers as two types of nodes of a bipartite graph data structure, converts an original data table containing missing trait data into a bipartite graph data structure, constructs a bipartite graph neural network structure and a loss function based on a gradient balancing mechanism, dynamically adjusts sample weights by calculating the difference between estimated values ​​and true values, effectively reduces the impact of sample imbalance problems between planting locations on model performance, improves agricultural production efficiency, and reduces losses caused by excessive deviations in yield estimation using traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a flow chart of a method for filling missing trait data and estimating yield of corn provided by the present invention; Figure 2 A schematic diagram of bipartite graph data provided by the present invention; Figure 3 A schematic diagram of a bipartite graph neural network provided by the present invention; Figure 4 A schematic diagram of the principle of the neural network model provided by the present invention; Figure 5 A schematic diagram of the structure of a system for filling missing trait data and estimating yield of corn provided by the present invention; Figure 6 It is a structural schematic diagram of a device for filling missing trait data and estimating yield of corn provided by the present invention; Figure 7 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0020] The core of the present invention is to provide a method, device, electronic device and storage medium for filling missing trait data and estimating yield of corn. By constructing a bipartite graph neural network model and a loss function of a gradient balancing mechanism, the impact of sample imbalance between planting locations on model performance is effectively reduced, the prediction efficiency and recognition accuracy are improved, and labor costs are reduced.

[0021] Therefore, a method for filling missing trait data and estimating yield of corn is urgently needed to solve the above problems.

[0022] Combine the following Figure 1-Figure 7 The invention describes a method, device, electronic device and storage medium for filling missing trait data and estimating yield of corn.

[0023] Figure 1 The figure is a flow chart of a method for filling missing trait data and estimating yield of corn provided by the present invention, and the method comprises the following steps: Step 101: Obtain corn planting data, and convert the corn planting data into a bipartite graph data structure.

[0024] In the specific implementation, meteorological data and corn trait data of corn planting sites are collected, the value of each node is initialized using encoding, and the data in the original table is converted into edge weights to obtain a bipartite graph data structure.

[0025] In one example, meteorological characteristics of the planting site, corn plant height, ear position, growth period, and yield-related characteristics during the corn breeding process are collected to construct an original data set; the feature name of the original data set is used as the first type of node of the bipartite graph data structure, the planting data number of the original data set is used as the second type of node of the bipartite graph data structure, and the data value in the original data set is used as the weight of the edge in the bipartite graph data structure; the numerical values ​​in the first type of node and the second type of node are initialized using hot encoding to obtain a bipartite graph data structure.

[0026] According to a method for filling missing trait data and estimating yield of corn based on a bipartite graph neural network provided in this embodiment, according to the description about collecting meteorological data of the planting site and corn planting trait data during the corn breeding process. And the feature names and planting data numbers in the original data table are respectively used as two types of nodes of the bipartite graph data structure. Specifically, all data numbers in the original data table are used as a type of node of the bipartite graph data structure; all feature names and per mu yield names are used as another type of node of the bipartite graph data structure; the data values ​​(feature values) in the table are regarded as the weights of the edges in the bipartite graph data structure. The value of each node is initialized using hot encoding, and each node is encoded as an m+1-dimensional 0,1 vector for distinguishing feature numbers; among which the nodes of the feature names and per mu yield names (row 1) are used as a type of node of the bipartite graph data structure, and are used Indicates that, then F j The j-th dimension feature of is 1, and the features of the other dimensions are all 0; all the data numbers (column 1) in the original data table are used as another type of node in the bipartite graph data structure, and Indicates that at this time the node The encoding of depends on the data missing situation. If the i-th group of data has no data missing and records the actual output, then the node All eigenvalues ​​of are 1; if the j-th dimension feature of the i-th group of data is missing, then the node The j-th dimension feature of is 0. In summary, the original data table containing missing trait data is converted into a bipartite graph data structure, and all key data in the original table are accurately retained through the weight of the edge, and the initial association between corn planting samples and the initial association between sample trait characteristics and meteorological characteristics are established.

[0027] Step 102: construct a bipartite graph neural network based on the bipartite graph data structure to obtain a corn missing trait data filling and yield estimation model.

[0028] In a specific implementation, a bipartite graph neural network consisting of a graph update layer and a prediction layer is constructed based on the bipartite graph data structure; In the graph update layer, edge embedding is used to update node features, and edge weights are updated based on the node features; New feature vectors are filled for missing data values ​​based on the node features and the edge weights, and the node features obtained from the last layer of the graph update layer are input into the prediction layer to obtain a corn missing trait data filling and yield estimation model.

[0029] In one example, the bipartite graph neural network consists of three graph update layers and one prediction layer. It uses edge embedding to update node features, reversely updates edge weights based on node features, and uses a fully connected layer to complete the construction of a corn estimation model.

[0030] According to a method for filling missing trait data and estimating yield of corn based on a bipartite graph neural network provided in this embodiment, according to the bipartite graph neural network composed of a graph update layer and a prediction layer, edge embedding is first used to update node features in the graph update layer, and then the edge weights are updated based on the node features. This process fills new feature vectors for missing data values ​​based on node features and edge weights. The design of the multi-layer graph update layer enables the bipartite graph structure to fully learn the high-order associations between data. Finally, the prediction layer is set, and the node features obtained by the last layer of the graph update layer are used as input, and the missing trait data filling and corn yield prediction are realized through the fully connected layer of the multi-layer structure.

[0031] Step 103: construct a loss function of the corn missing trait data filling and yield estimation model using a gradient balancing mechanism, and train the corn missing trait data filling and yield estimation model with the loss function using a corn planting data training set to obtain a trained corn missing trait data filling and yield estimation model.

[0032] In specific implementation, the absolute value error loss function is embedded using the gradient balancing mechanism, and its calculation formula is:

[0033] in, represents the number of training samples, represents the estimated value of production, represents the true value of output, express Loss, which measures the difference between the predicted value and the true value, is a constant.

[0034] According to a method for filling missing trait data and estimating yield of corn based on a bipartite graph neural network provided in this embodiment, a new loss function based on a gradient balancing mechanism is designed according to the above. Due to the different planting scales and distributions of various corn planting sites, the sample size of some categories is much larger than that of other categories. This imbalance will affect the learning process of the model, causing the model to tend to predict categories with a large number, thereby reducing the prediction performance. The gradient balancing mechanism (GHM), as an effective method for dealing with sample imbalance, is particularly outstanding in classification tasks. By drawing on GHM, an effective method for dealing with similar problems, it is embedded in the L1 loss function commonly used to optimize regression problems, and the loss function L_GHM in the present invention is designed to reduce the impact of sample imbalance. By calculating the difference between the estimated value and the true value, the sample weight is dynamically adjusted to reduce the interference of data imbalance on the model.

[0035] Step 104: Use the trained corn missing trait data filling and yield estimation model to predict the data to be tested to obtain a prediction result.

[0036] Construct a graphical interface to display the missing trait filling results and yield prediction results.

[0037] The present embodiment provides a method for filling missing trait data and calculating yield of corn. According to the collected planting location meteorological data and corn planting trait data, the original data feature name and planting data number are respectively used as two types of nodes of the bipartite graph data structure, the value of each node is initialized by hot encoding, and all key data in the original table are converted into edge weights, so that the original data table containing missing trait data is converted into a bipartite graph data structure, and the initial association between corn planting samples and the initial association between sample trait characteristics and meteorological characteristics are established; the loss function of the bipartite graph neural network structure and the gradient balance mechanism is constructed, and the sample weight is dynamically adjusted by calculating the difference between the estimated value and the true value, so as to effectively reduce the impact of the sample imbalance problem between each planting location on the model performance; and a graphical interface is designed to intuitively display the filling results of the missing trait data and the prediction results of corn yield. This helps to better understand and utilize the yield measurement results, improve agricultural production efficiency and reduce the losses caused by the excessive deviation of yield estimation by traditional methods.

[0038] Based on the above embodiment, this embodiment describes a method for filling missing trait data and estimating yield of corn, as follows: like Figure 2 As shown, the meteorological data of the planting location and the corn planting trait data are converted into a bipartite graph data structure; Data collection: During the corn breeding process, the first step is to collect corn planting data including meteorological characteristics of the planting site, corn trait characteristics, and per-acre yield.

[0039] Use hot encoding to initialize the value of each node, and convert all key data in the original table into edge weights to obtain a bipartite graph data structure: Specifically, the first row (meteorological characteristics, corn trait characteristics, per-acre yield) and the first column (planting data number) of the original data table are used as two types of nodes in the bipartite graph data structure, and the data values ​​in the table are regarded as the weights of the edges of the graph. Among them, meteorological characteristics and corn trait characteristics have m dimensions, and the table records a total of n groups of planting sample data. All data numbers (the first column) in the original data table are used as a type of node in the bipartite graph data structure, and N i , i∈(1,n) represents; all feature names and per-acre yield names (row 1) are regarded as another type of node in the bipartite graph data structure, and F j , j ∈(1,m+1). In summary, the table contains n×(m+1) data values, so the bipartite graph data structure contains n×(m+1) edges. The first Feature available edges e ij Indicates that the data value is e ij The weight of , if the value is empty, then e ij =0.

[0040] The value of each node in the bipartite graph is initialized using hot encoding, and each node is encoded as an m+1-dimensional 0,1 vector. F j No. j The feature of the dimension is 1, and the features of the other dimensions are 0. N i The coding of i The group data has no missing data and also records the actual output, then the node N i All eigenvalues ​​of are 1; if i The first j If the dimension feature is missing, then the node N i No. j The dimension feature is 0. In summary, the initialized bipartite graph data structure accurately retains all the key data in the original table through the edge weights, and can establish the initial association between corn planting samples, as well as the initial association between sample trait characteristics and meteorological characteristics.

[0041] Design a bipartite graph neural network that can simultaneously fill in missing trait data and predict corn yield. Specifically: There are two objective laws in corn planting: first, there is a correlation between meteorological characteristics and corn trait characteristics, and they will jointly affect corn yield; second, there is a spatial correlation between meteorological characteristics, corn trait characteristics, and corn yield among different planting data. In summary, the bipartite graph neural network based filling in missing traits and predicting corn yield has strong applicability. Figure 3 As shown in Figure 1, the constructed bipartite graph neural network consists of three graph update layers and one prediction layer, where the weights of the bipartite graph edges represent all the numerical values ​​in the tabular data, which are crucial for yield prediction.

[0042] The graph update layer solves the problem of using the weights of the initial edges to fill in the missing features and predict the yield. Each graph update layer is completed in two steps.

[0043] First, edge embedding is used to update node features through three graph update layers. The process is shown in the following formula:

[0044] The formula shows that The graph update layer is a process of updating node features based on edge embedding. Hot encoding is used to distinguish the two types of nodes in different bipartite graphs. In the graph update layer, the update strategies of the two types of nodes are exactly the same, so these two types of nodes are collectively called .in and Respectively represented in Node in layer and The characteristic representation of Representation Node With Node The weight of the connected edge (edge). For each adjacent node of node i j (i.e. all j In the collection ), firstly, the feature representation of node j is With edge Embedding for splicing operation . Then the concatenated vector Apply a fully connected layer The purpose is to fuse and transform the features of nodes and the information of edges to extract and learn from nodes. j To Node i The relationship and influence of the aggregated adjacent node features and the target node i In the Layer characteristics Concatenate to retain its own information and combine the information of adjacent nodes. Apply another fully connected layer to the above concatenated vector The information can be further transformed and integrated to obtain the node i In the next feature update layer New feature representation In this way, the characteristics of nodes at each layer not only reflect the information of their adjacent nodes, but also integrate the properties of the edges connecting these nodes, which can provide a more comprehensive understanding of the relationship and influence between nodes.

[0045] Based on the node features, the weights of the edges are updated inversely to learn the high-order associations between the data. The process is shown in the formula:

[0046] First, get two nodes i and j in the The feature representations after layer update are and Then the feature representation of these two nodes and the relationship between these two nodes are Edge weights of layers Perform a concatenation operation (Con). This not only takes into account the latest features of the nodes, but also the existing relationship strength or characteristics between them. Through this concatenation, a new feature vector of the edge is constructed based on the latest state of the node and the existing information of the edge. . The weight of the edge is not static, but will be adjusted dynamically as the network learns more deeply and the node features are updated. It not only captures the direct relationship between nodes and edges, but also learns the high-order associations between data, so that the model can understand the complexity of the graph structure more deeply, and thus better predict the future state of nodes or edges.

[0047] like Figure 4 As shown in the figure, a prediction layer is set up at last, and the node features obtained from the third layer of the graph update layer are used as input. Through two fully connected layers, nonlinear activation functions are introduced at the same time to increase the expressive power of the model. The model finally outputs the prediction of missing trait data and corn yield, and then random dropout is used to prevent the network from overfitting. For each layer of the graph update layer, the hyperparameter of random dropout is set to 0.1, which means that each neuron in the graph update layer is randomly dropped with a probability of 0.1 to prevent the model from overfitting. Through random dropout, the model is forced to learn more robust feature representations because it cannot rely too much on the features of any node or edge, thereby improving the prediction ability of unseen data. The model's excessive dependence on input features is reduced, preventing certain features from having too much influence on the prediction results, making the model more robust.

[0048] Design a new loss function based on the gradient balancing mechanism. Specifically: Due to the different planting scales and distributions of corn planting sites, the sample size of some categories is much larger than that of other categories. This imbalance affects the learning process of the model, causing the model to tend to predict categories with a large number, thereby reducing the prediction performance. The gradient harmonized mechanism (GHM) is an effective method for dealing with sample imbalance, especially in classification tasks.

[0049] By drawing on GHM, an effective method for dealing with similar problems, and embedding it into the L1 loss function commonly used to optimize regression problems, the loss function of this paper is designed To reduce the impact of sample imbalance. As shown in the formula:

[0050] in, represents the number of training samples, represents the estimated value of production, represents the true value of output, express Loss, which measures the difference between the predicted value and the true value, The gradient balancing mechanism is the core of GHM, which aims to adjust the weights of different samples, especially those samples with large differences between predicted values ​​and true values. is a constant used to avoid the situation where the denominator is zero and to smooth the gradient. It is usually a small positive value (such as 0.3) to ensure stability. Generally, the fewer the planting locations in the sample data, the greater the difference between the estimated yield and the true value. By calculating the difference between the estimated value and the true value, the sample weight is dynamically adjusted to reduce the interference of data imbalance on the model.

[0051] Perform missing fill and yield prediction on the input data, and display the final missing fill and yield prediction results. Specifically: Graphical interface design: In order to make the maize missing trait data filling and yield estimation results of this method easier to understand and use, a user-friendly graphical interface is designed to display the maize missing trait data filling and yield estimation results. The interface can display the meteorological characteristics and maize trait characteristic data of different missing conditions to perform maize missing trait data filling and yield estimation analysis on the data, and by mining the high-order associations between different planting data and different characteristics, the optimal yield prediction results can be intuitively reflected.

[0052] According to a method and system for filling missing trait data and estimating yield of corn based on a bipartite graph neural network provided in this application, Figure 5 As shown, the method and system include the following modules: Bipartite graph construction module: collects meteorological data of planting sites and corn planting trait data, uses hot encoding to initialize the value of each node, and converts all key data in the original table into edge weights, thereby obtaining a bipartite graph data structure, and establishing the initial association between corn planting samples, as well as the initial association between sample trait characteristics and meteorological characteristics. (2) Bipartite graph neural network module: constructs a bipartite graph neural network consisting of a graph update layer and a prediction layer, which consists of 3 graph update layers and 1 prediction layer. In the graph update layer, edge embedding is used to update node features, and then the edge weights are updated based on the node features. This process fills new feature vectors for missing data values ​​based on node features and edge weights; using the final prediction layer, the node features obtained from the last layer of the graph update layer are used as input, and through the multi-layer fully connected layer, the problem of filling missing features and predicting yield is finally achieved; at the same time, a loss function based on the gradient balance mechanism is designed to dynamically adjust the sample weights by calculating the difference between the estimated value and the true value, effectively reducing the impact of the imbalance of samples between planting sites on model performance. (3) Graphical yield estimation module: A graphical interface is designed to display the missing trait data and yield estimation results of corn, which helps to better understand and utilize the yield estimation results, improve agricultural production efficiency, and reduce losses caused by excessive deviations in yield estimation using traditional methods.

[0053] The following is a description of the device for filling missing trait data and calculating yield of corn provided by the present invention. The device for filling missing trait data and calculating yield of corn described below and the method for filling missing trait data and calculating yield of corn described above can be referenced to each other.

[0054] like Figure 6 As shown, the present invention provides a device for filling missing trait data and estimating yield of corn, comprising the following modules: The data acquisition module 601 acquires corn planting data and converts the corn planting data into a bipartite graph data structure; A model building module 602 builds a bipartite graph neural network based on the bipartite graph data structure to obtain a corn missing trait data filling and yield estimation model; The model training module 603 uses a gradient balancing mechanism to construct a loss function of the corn missing trait data filling and yield estimation model, and uses a corn planting data training set to train the corn missing trait data filling and yield estimation model with the loss function constructed, to obtain a trained corn missing trait data filling and yield estimation model; The prediction module 604 uses the trained corn missing trait data filling and yield estimation model to predict the data to be tested to obtain a prediction result.

[0055] Figure 7An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, the electronic device may include: a processor 710, a communications interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communications interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the method for filling missing trait data of corn and estimating yield, and the method includes: Acquire corn planting data, and convert the corn planting data into a bipartite graph data structure; Constructing a bipartite graph neural network based on the bipartite graph data structure to obtain a corn missing trait data filling and yield estimation model; A loss function of the corn missing trait data filling and yield estimation model is constructed by using a gradient balance mechanism, and the corn missing trait data filling and yield estimation model with the loss function is trained by using a corn planting data training set to obtain a trained corn missing trait data filling and yield estimation model; The trained corn missing trait data filling and yield estimation model is used to predict the data to be tested to obtain a prediction result.

[0056] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0057] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the corn missing trait data filling and yield estimation method provided by the above methods, the method includes: Acquire corn planting data, and convert the corn planting data into a bipartite graph data structure; Constructing a bipartite graph neural network based on the bipartite graph data structure to obtain a corn missing trait data filling and yield estimation model; A loss function of the corn missing trait data filling and yield estimation model is constructed by using a gradient balance mechanism, and the corn missing trait data filling and yield estimation model with the loss function is trained by using a corn planting data training set to obtain a trained corn missing trait data filling and yield estimation model; The trained corn missing trait data filling and yield estimation model is used to predict the data to be tested to obtain a prediction result.

[0058] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to perform the corn missing trait data filling and yield estimation method provided by the above methods, the method comprising: Acquire corn planting data, and convert the corn planting data into a bipartite graph data structure; Constructing a bipartite graph neural network based on the bipartite graph data structure to obtain a corn missing trait data filling and yield estimation model; A loss function of the corn missing trait data filling and yield estimation model is constructed by using a gradient balance mechanism, and the corn missing trait data filling and yield estimation model with the loss function is trained by using a corn planting data training set to obtain a trained corn missing trait data filling and yield estimation model; The trained corn missing trait data filling and yield estimation model is used to predict the data to be tested to obtain a prediction result.

[0059] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0060] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for filling missing trait data and estimating yield of corn, characterized in that: The method comprises: Acquire corn planting data, and convert the corn planting data into a bipartite graph data structure; Constructing a bipartite graph neural network based on the bipartite graph data structure to obtain a corn missing trait data filling and yield estimation model; A loss function of the corn missing trait data filling and yield estimation model is constructed by using a gradient balance mechanism, and the corn missing trait data filling and yield estimation model with the loss function is trained by using a corn planting data training set to obtain a trained corn missing trait data filling and yield estimation model; The trained corn missing trait data filling and yield estimation model is used to predict the data to be tested to obtain a prediction result.

2. The method for filling missing trait data and estimating yield of corn according to claim 1, characterized in that: The step of obtaining corn planting data and converting the corn planting data into a bipartite graph data structure includes: Meteorological data and corn trait data of corn planting sites are collected, the value of each node is initialized using encoding, and the data in the original table is converted into edge weights to obtain a bipartite graph data structure.

3. The method for filling missing trait data and estimating yield of corn according to claim 2, characterized in that: The encoding method is used to initialize the value of each node and convert the data in the original table into the weight of the edge, including: The meteorological characteristics of the planting area, corn plant height, ear position, growth period, and yield-related characteristics during corn breeding were collected to construct the original data set; Using the feature name of the original data set as the first type of node of the bipartite graph data structure, using the planting data number of the original data set as the second type of node of the bipartite graph data structure, and using the data value in the original data set as the weight of the edge in the bipartite graph data structure; The values ​​in the first type of nodes and the second type of nodes are initialized using a hot encoding method to obtain a bipartite graph data structure.

4. The method for filling missing trait data and estimating yield of corn according to claim 1, characterized in that: The bipartite graph neural network is constructed based on the bipartite graph data structure to obtain a corn missing trait data filling and yield estimation model, including: Constructing a bipartite graph neural network consisting of a graph update layer and a prediction layer based on the bipartite graph data structure; In the graph update layer, edge embedding is used to update node features, and edge weights are updated based on the node features; New feature vectors are filled for missing data values ​​based on the node features and the edge weights, and the node features obtained from the last layer of the graph update layer are input into the prediction layer to obtain a corn missing trait data filling and yield estimation model.

5. The method for filling missing trait data and estimating yield of corn according to claim 4, characterized in that: Also includes: The bipartite graph neural network consists of three graph update layers and one prediction layer. It uses edge embedding to update node features, reversely updates edge weights based on node features, and uses a fully connected layer to complete the construction of corn missing trait data filling and yield estimation model.

6. The method for filling missing trait data and estimating yield of corn according to claim 1, characterized in that: The loss function of the maize missing trait data filling and yield estimation model is constructed by using the gradient balance mechanism, including: using the gradient balance mechanism to embed the absolute value error loss function, and the calculation formula is: in, represents the number of training samples, represents an estimate of production, represents the true value of output, express Loss, which measures the difference between the predicted value and the true value, is a constant.

7. The method for filling missing trait data and estimating yield of corn according to claim 1, characterized in that: Also includes: Construct a graphical interface to display the missing trait filling results and yield prediction results.

8. A device for filling missing trait data and estimating yield of corn, characterized in that: include: A data acquisition module, which acquires corn planting data and converts the corn planting data into a bipartite graph data structure; A model building module, which builds a bipartite graph neural network based on the bipartite graph data structure to obtain a corn missing trait data filling and yield estimation model; A model training module, which uses a gradient balancing mechanism to construct a loss function of the corn missing trait data filling and yield estimation model, and uses a corn planting data training set to train the corn missing trait data filling and yield estimation model with the loss function constructed, to obtain a trained corn missing trait data filling and yield estimation model; The prediction module uses the trained corn missing trait data filling and yield estimation model to predict the data to be tested to obtain a prediction result.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images