Comprehensive evaluation method and evaluation system for heat resistance of rice

Abstract

The invention discloses a rice heat resistance comprehensive evaluation method and evaluation system, and relates to the technical field of rice heat resistance evaluation, and the method comprises the steps: extracting heat resistance evaluation indexes in rice heat resistance evaluation instance data, building an interaction diagram according to the interaction relationship between each heat resistance evaluation index and a rice heat resistance evaluation task, and obtaining a rice heat resistance evaluation result; screening key heat resistance evaluation indexes by using a graph neural network; performing knowledge enhancement on the key heat-resistant evaluation indexes by using a domain knowledge graph, generating expanded heat-resistant evaluation indexes to supplement the key heat-resistant evaluation indexes, and selecting an optimal key heat-resistant evaluation index combination through feature selection; and constructing a rice heat resistance comprehensive evaluation model based on the data reconstruction network, and obtaining a rice heat resistance comprehensive evaluation result of the to-be-tested rice. The rice heat resistance evaluation model is constructed by combining the key heat resistance evaluation indexes and the data reconstruction model, automation and intelligence of rice heat resistance evaluation are achieved, and rice heat resistance evaluation precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of rice heat tolerance evaluation, and more specifically, to a comprehensive evaluation method and evaluation system for rice heat tolerance. Background Art

[0002] Rice is one of the world's most important food crops. With global warming, heat stress has become a major factor affecting rice yield and quality. High temperatures can stunt rice growth, cause pollen sterility, and impaired grain filling, severely reducing yield. Therefore, studying heat tolerance in rice and developing heat-tolerant varieties are crucial for ensuring food security. High temperature stress can inhibit root development and tillering, disrupt chloroplast structure, reduce photosynthetic efficiency, lead to dwarf plants, and impair carbohydrate accumulation. Furthermore, high temperatures can cause pollen sterility and reduce seed set, severely impacting yield and reducing rice quality.

[0003] Traditional methods for evaluating rice heat tolerance rely primarily on field trials and manual observation. These involve growing rice under naturally high temperatures to observe its growth, development, and yield performance, as well as simulating high temperatures in artificial climate chambers, controlling temperature and time, and evaluating rice heat tolerance. These evaluation indicators typically include morphological indicators such as plant height, tiller number, and leaf area, as well as physiological and biochemical indicators such as chlorophyll content, photosynthetic rate, and antioxidant enzyme activity. However, the lack of a comprehensive and scientific evaluation system for rice heat tolerance leads to inconsistent evaluation results and small sample sizes. Therefore, leveraging artificial intelligence and big data technologies to develop an intelligent heat tolerance evaluation system for comprehensive evaluation of rice heat tolerance is an urgent issue that needs to be addressed. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention proposes a comprehensive evaluation method and evaluation system for rice heat tolerance, which combines key heat tolerance evaluation indicators and data reconstruction models to construct a rice heat tolerance evaluation model, thereby realizing the automation and intelligence of rice heat tolerance evaluation.

[0005] A first aspect of the present invention provides a method for comprehensive evaluation of heat tolerance of rice, comprising the following steps: Retrieving and obtaining rice heat tolerance evaluation instance data, extracting heat tolerance evaluation indicators from the rice heat tolerance evaluation instance data, and constructing a heat tolerance evaluation indicator set; constructing an interaction graph based on the interactive relationship between each heat tolerance evaluation indicator in the heat tolerance evaluation indicator set and the rice heat tolerance evaluation task, using a graph neural network to perform representation learning on the interaction graph, and screening key heat tolerance evaluation indicators; Use domain knowledge graphs to enhance the knowledge of key heat resistance evaluation indicators, generate extended heat resistance evaluation indicators to supplement key heat resistance evaluation indicators, and select the best key heat resistance evaluation indicator combination; A comprehensive evaluation model of rice heat resistance is constructed based on the optimal combination of key heat resistance evaluation indicators. The indicator parameter values ​​of the rice to be tested are collected as model input to obtain the comprehensive evaluation results of the rice heat resistance of the rice to be tested.

[0006] In this solution, we retrieved and obtained rice heat tolerance evaluation instance data, extracted heat tolerance evaluation indicators from the rice heat tolerance evaluation instance data, and constructed a heat tolerance evaluation indicator set, specifically: Use "rice heat tolerance evaluation" as a keyword to obtain relevant descriptive text about rice heat tolerance evaluation. Use the pre-trained BERT model for word embedding. Use the attention mechanism to obtain the global context of different words in the relevant descriptive text and obtain the corresponding global context embedding. Obtaining semantic relative distances in the relevant description text based on position information of different word vectors and keyword vectors, selecting words with a semantic relative distance less than a preset threshold as local contexts, and dynamically weighting the local contexts to obtain local context embedding; The global context embedding and the local context embedding are spliced ​​and imported into a fully connected layer to obtain a corresponding semantic feature sequence, the semantic feature sequence is used in combination with the IVF algorithm and the HNSW algorithm to construct a retrieval index, and the HNSW structure is applied to the inverted index to generate a rice heat tolerance evaluation index; Using the rice heat tolerance evaluation index combined with the cosine distance measurement function to obtain similarity between data, obtaining rice heat tolerance evaluation instance data that meets the similarity standard, and performing structured processing on the rice heat tolerance evaluation instance data; The heat tolerance evaluation indicators involved are extracted from the structured rice heat tolerance evaluation instance data to construct a heat tolerance evaluation indicator set.

[0007] In this scheme, the interactive relationship between heat tolerance evaluation indicators and rice heat tolerance evaluation tasks is obtained to construct an interaction graph. A graph neural network is used to learn the representation of the interaction graph and screen key heat tolerance evaluation indicators. Specifically, Performing usage frequency statistics on the heat resistance evaluation indicators in the heat resistance evaluation indicator set, performing graph processing on the heat resistance evaluation indicator set, using the heat resistance evaluation indicators and rice heat resistance evaluation tasks as nodes, setting connection relationships between the nodes according to the interactive relationships between different rice heat resistance evaluation tasks and heat resistance evaluation indicators, generating an interaction graph, and setting edge weights in the interaction graph using the usage frequency; Using a graph neural network to perform representation learning on the interaction graph, obtaining feature representations of heat tolerance evaluation index nodes and rice heat tolerance evaluation task nodes, calculating an inner product value using a feature vector of the heat tolerance evaluation index node and a feature vector of the rice heat tolerance evaluation task node, and characterizing the degree of influence of the heat tolerance evaluation index node and the rice heat tolerance evaluation task through the inner product value; And in the graph neural network layer, the inner product value and edge weight of each heat-resistant evaluation index node are used for aggregation to obtain the single-layer neighborhood update representation of the heat-resistant evaluation index node, and the single-layer neighborhood representation of the heat-resistant evaluation index node and the single-layer neighborhood update representation of the heat-resistant evaluation index node are used; By stacking several layers of graph neural networks, the heat resistance evaluation index node and the final feature vector of the heat resistance evaluation index node are obtained, and the CTR estimation model is used to calculate the matching degree of the final feature vector of the heat resistance evaluation index node and the final feature vector of the heat resistance evaluation index node, and the key heat resistance evaluation indicators are screened based on the matching degree.

[0008] In this solution, domain knowledge graph is used to enhance the knowledge of key heat resistance evaluation indicators, and extended heat resistance evaluation indicators are generated to supplement the key heat resistance evaluation indicators. Specifically: Access the knowledge graph related to rice heat tolerance evaluation, and use key heat tolerance evaluation indicators to obtain the corresponding knowledge graph subgraph in the domain knowledge graph; Taking "obtaining extended indicators of key heat resistance evaluation indicators" as the problem input, the problem is vectorized and features are extracted. The multi-head attention mechanism is used to obtain fine-grained information of the problem vectorization representation to obtain intermediate reasoning information; Generate an intermediate indicator vector based on the intermediate reasoning information, combine attention with the intermediate indicator vector to obtain a final indicator vector, use the final indicator vector to perform local path matching on the knowledge graph subgraph, and import the entity relationship in the local path into the BERT model to obtain the local path embedding; The local path embedding is scored for similarity using the indicator vector, and the entity score is updated using the similarity score. The entity scores of each hop are stacked and multiplied by the attention and summed to obtain the final entity score. According to the final entity score, entities that meet the standards are screened to generate an expanded heat resistance evaluation index to supplement the key heat resistance evaluation index.

[0009] In this scheme, the best combination of key heat resistance evaluation indicators is selected, specifically: Obtaining the supplemented key heat-resistance evaluation indicators for encoding, randomly generating an initial heat-resistance evaluation indicator combination to construct an initial population, optimizing the heat-resistance evaluation indicator combination through a genetic algorithm, constructing a classification model using an SVM model, importing the feature vector corresponding to the initial heat-resistance evaluation indicator combination into the classification model, obtaining the classification accuracy, and constructing a fitness function using the classification accuracy; Calculating the fitness of the heat-resistance evaluation index combination according to the fitness function, selecting a preset number of heat-resistance evaluation index combinations for replication based on the fitness to achieve population expansion, and cross-recombining the selected heat-resistance evaluation index combinations to change the coding of individual heat-resistance evaluation indicators therein; The optimal solution is obtained by iteratively optimizing the population and outputting the best combination of key heat resistance evaluation indicators.

[0010] In this scheme, a comprehensive evaluation model for heat tolerance of rice is constructed based on the optimal combination of key heat tolerance evaluation indicators, specifically: Extracting single-indicator evaluation data from rice heat tolerance evaluation instance data based on each key heat tolerance evaluation indicator in the optimal key heat tolerance evaluation indicator combination, and generating data labels using the rice heat tolerance evaluation results; A comprehensive evaluation model for rice heat tolerance was constructed using a conditional variational autoencoder. Labeled single-indicator evaluation data was used for model training. The encoder obtained the latent variable distribution of the training samples, and the decoder used this latent variable distribution to reconstruct the data. The rice heat tolerance evaluation result label is combined with the latent variable distribution to train the discriminator, obtain the estimated single indicator evaluation data under the specific rice heat tolerance evaluation result label generated by data reconstruction, and use the discriminator to determine the data label. Through alternating training, a trained conditional variational autoencoder is obtained; The estimated single-indicator evaluation data generated by the conditional variational autoencoder is imported into a multi-layer perceptron to obtain the deviation from the test data. The rice heat tolerance evaluation result label classification result is obtained based on the deviation. When the classification accuracy meets the preset requirements, the comprehensive evaluation model of rice heat tolerance is output.

[0011] In this scheme, the index parameter values ​​of the rice to be tested are collected as model input to obtain the comprehensive evaluation results of the heat tolerance of the rice to be tested, specifically: Using the best key heat tolerance evaluation index combination to collect index parameter values ​​of the tested rice, the index parameter values ​​are imported into the comprehensive evaluation model of rice heat tolerance to obtain the estimated index parameter values ​​under each rice heat tolerance evaluation result label; The residual between the estimated index parameter value and the index parameter value is obtained through a multi-layer perceptron, the rice heat tolerance evaluation result label corresponding to the minimum residual is obtained, and the heat tolerance evaluation result of the tested rice is output.

[0012] The second aspect of the present invention provides a comprehensive evaluation system for heat tolerance of rice, comprising: an evaluation index screening module, an evaluation index combination optimization module, a data acquisition module, a rice heat tolerance evaluation module and an evaluation result output module; The evaluation index screening module is responsible for extracting heat tolerance evaluation indicators from rice heat tolerance evaluation instance data, constructing an interaction graph based on the interactive relationship between each heat tolerance evaluation indicator and the rice heat tolerance evaluation task, using a graph neural network to perform representation learning on the interaction graph, and screening key heat tolerance evaluation indicators; The evaluation index combination optimization module uses the domain knowledge graph to enhance the knowledge of key heat-resistant evaluation indicators, constructs extended heat-resistant evaluation indicators to supplement the key heat-resistant evaluation indicators, and selects the best key heat-resistant evaluation indicator combination through feature selection; The data acquisition module is responsible for collecting the index parameter values ​​of the rice to be tested according to the best key heat resistance evaluation index combination; The rice heat tolerance evaluation module constructs a comprehensive evaluation model for rice heat tolerance based on the data reconstruction model combined with the best key heat tolerance evaluation index combination, and imports the index parameter values ​​to obtain the comprehensive evaluation result of rice heat tolerance of the tested rice; The evaluation result output module is responsible for outputting and visually displaying the comprehensive evaluation result of the heat resistance of the rice to be tested in a predicted manner.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The present invention uses a graph neural network to capture the complex relationship between heat resistance evaluation indicators, screen out the key indicators that have the greatest impact on rice heat resistance, and improve the pertinence of the evaluation. Combining key heat resistance evaluation indicators and data reconstruction models, the constructed rice heat resistance evaluation model has high accuracy and robustness, while ensuring the consistency of rice heat resistance evaluation results, reducing dependence on manual experience, and realizing the automation and intelligence of rice heat resistance evaluation. The rice heat resistance evaluation model constructed by the present invention has the characteristics of high efficiency, accuracy and intelligence. Obtaining accurate rice heat resistance evaluation data provides a data foundation for multiple fields such as rice breeding, agricultural research and agricultural production. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments or exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or exemplary descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained according to these drawings without paying any creative work.

[0015] Figure 1 A flow chart of a comprehensive evaluation method for heat tolerance of rice is shown; Figure 2 A flow chart showing the embodiment of screening key heat resistance evaluation indicators is shown; Figure 3 A flowchart of an embodiment for constructing a comprehensive evaluation model for heat tolerance of rice is shown; Figure 4 A block diagram of a comprehensive evaluation system for heat tolerance of rice is shown. DETAILED DESCRIPTION

[0016] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0017] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0018] Figure 1 A flow chart of a comprehensive evaluation method for heat tolerance of rice is shown.

[0019] like Figure 1 As shown, this embodiment provides a comprehensive evaluation method for heat tolerance of rice, comprising: S102, retrieving and obtaining rice heat tolerance evaluation instance data, extracting heat tolerance evaluation indicators from the rice heat tolerance evaluation instance data, and constructing a heat tolerance evaluation indicator set; S104, constructing an interaction graph based on the interactive relationship between each heat tolerance evaluation indicator in the heat tolerance evaluation indicator set and the rice heat tolerance evaluation task, performing representation learning on the interaction graph using a graph neural network, and screening key heat tolerance evaluation indicators; S106, using the domain knowledge graph to perform knowledge enhancement on the key heat resistance evaluation indicators, generating extended heat resistance evaluation indicators to supplement the key heat resistance evaluation indicators, and selecting the best key heat resistance evaluation indicator combination; S108, constructing a comprehensive evaluation model for heat tolerance of rice based on the optimal combination of key heat tolerance evaluation indicators, collecting indicator parameter values ​​of the rice to be tested as model input, and obtaining a comprehensive evaluation result of heat tolerance of rice to be tested.

[0020] It should be noted that the relevant description text of rice heat tolerance evaluation is obtained using rice heat tolerance evaluation as a keyword, the pre-trained BERT model is used for word embedding, and the attention mechanism is used to obtain the global context of different words in the relevant description text to obtain the corresponding global context embedding; the semantic relative distance is obtained according to the position information of different word vectors and keyword vectors in the relevant description text, and the words less than the preset semantic relative distance threshold are selected as the local context. The local context is dynamically weighted to obtain the local context embedding, and the calculation formula of the semantic relative distance d is: , Pi and Pg represent the position information of word vector and keyword vector respectively, and m represents the length of keyword vector.

[0021] The global context embedding and local context embedding are concatenated and fed into a fully connected layer to obtain the corresponding semantic feature sequence. This semantic feature sequence is then combined with the IVF and HNSW algorithms to construct a search index. The HNSW structure is then applied to an inverted index to generate a rice heat tolerance evaluation index. The IVF algorithm divides the feature space into several partitions, each represented by a centroid. For a query feature vector, the distance to all centroids is calculated, and an exact match search is performed in the partitions containing the n nearest centroids. The HNSW algorithm constructs a multi-layered NSW structure. Starting from any node at the top layer, the node closest to the query point is searched in each layer as the entry node for the next layer, continuing to the bottom layer. The HNSW algorithm is then applied to the nearest neighbor centroid search of the IVF algorithm to achieve algorithm integration. The similarity between data is obtained using the rice heat tolerance evaluation index combined with the cosine distance measurement function. The closer the distance value is to 0, the more it meets the preset standard. Rice heat tolerance evaluation instance data that meets the similarity standard is obtained, and the rice heat tolerance evaluation instance data is structured. The heat tolerance evaluation indicators involved are extracted from the structured rice heat tolerance evaluation instance data to construct a heat tolerance evaluation indicator set.

[0022] Figure 2 A flow chart of the embodiment for screening key heat resistance evaluation indicators is shown.

[0023] According to an embodiment of the present invention, the interactive relationship between the heat tolerance evaluation index and the rice heat tolerance evaluation task is obtained to construct an interaction graph, and a graph neural network is used to perform representation learning on the interaction graph to screen key heat tolerance evaluation indicators, specifically: S202, performing usage frequency statistics on the heat resistance evaluation indicators in the heat resistance evaluation indicator set, performing graph processing on the heat resistance evaluation indicator set, using the heat resistance evaluation indicators and the rice heat resistance evaluation tasks as nodes, setting connection relationships between the nodes according to the interactive relationships between different rice heat resistance evaluation tasks and the heat resistance evaluation indicators, generating an interaction graph, and setting edge weights in the interaction graph using the usage frequency; S204, performing representation learning on the interaction graph using a graph neural network to obtain feature representations of heat tolerance evaluation index nodes and rice heat tolerance evaluation task nodes, calculating an inner product value using a feature vector of the heat tolerance evaluation index node and a feature vector of the rice heat tolerance evaluation task node, and characterizing the degree of influence of the heat tolerance evaluation index node and the rice heat tolerance evaluation task using the inner product value; S206, and use the inner product value and edge weight of each heat resistance evaluation index node to perform aggregation in the graph neural network layer to obtain a single-layer neighborhood update representation of the heat resistance evaluation index node, and use the single-layer neighborhood representation of the heat resistance evaluation index node and the single-layer neighborhood update representation of the heat resistance evaluation index node; S208, obtain the heat resistance evaluation index node and the final feature vector of the heat resistance evaluation index node by stacking several layers of graph neural network, use the CTR estimation model to calculate the matching degree of the final feature vector of the heat resistance evaluation index node and the final feature vector of the heat resistance evaluation index node, and screen the key heat resistance evaluation index based on the matching degree.

[0024] It should be noted that in the graph neural network layer, the inner product value and edge weight of each heat tolerance evaluation indicator node are used for aggregation. After obtaining the node representation after the first layer of aggregation update, the neighboring nodes are used as new nodes for outward expansion, and then convolution is performed to obtain the updated feature representation of the node in the next graph neural network layer. Through residual stacking of several layers of graph neural networks, the heat tolerance evaluation indicator node and the final feature vector of the heat tolerance evaluation indicator node are obtained. The CTR prediction model is used to calculate the inner product matching degree of the final feature vector of the heat tolerance evaluation indicator node and the final feature vector of the heat tolerance evaluation indicator node, and a preset number of key heat tolerance evaluation indicators are selected. Preferred key heat tolerance evaluation indicators include heat stress index, anther morphology, fruit set rate, antioxidant enzyme activities such as catalase (CAT), superoxide dismutase (SOD), and peroxidase (POD), endogenous hormone levels such as methyl jasmonate (MeJA), jasmonic acid-isoleucine (JAIle), and dihydrojasmonic acid, and gene expression levels of superoxide dismutase gene OsSOD8 and peroxidase gene PODA2.

[0025] It is important to note that the domain knowledge graph is used to enhance the knowledge of key heat tolerance evaluation indicators. The domain knowledge graph related to rice heat tolerance evaluation is accessed, and the corresponding knowledge graph subgraph is obtained from the domain knowledge graph using the key heat tolerance evaluation indicators. The question "obtaining extended indicators for key heat tolerance evaluation indicators" is used as the input. A pre-trained network, such as the BERT model, is used to vectorize the question and extract semantic features. During each reasoning hop, the question representation and the indicator vector from the previous hop are first combined through a multi-head attention mechanism to obtain fine-grained information from the question vectorization, generating intermediate reasoning information. The obtained attention weights are used to indicate which part of the current question is more important. Indicator vectors are generated for different parts of the question to guide dynamic reasoning and path expansion. Based on the intermediate reasoning information, intermediate indicator vectors are generated. During each reasoning hop, more attention is paid to the current information. A sigmoid activation function is used to output a random value, which is multiplied by the historical reasoning information to filter out redundant information from the previous hop. An indicator vector is generated by combining past and present key reasoning information. The attention weights of each word vector in each question are combined with the intermediate indicator vector to obtain the final indicator vector.

[0026] The final indicator vector is used to perform local path matching on the knowledge graph subgraph, and the entity relationship in the local path is imported into the BERT model to obtain the local path embedding; the indicator vector is used to perform similarity scoring on the local path embedding, and the entity score is updated using the similarity score. The previous jump indicator vector and the local path embedding are multiplied and passed through a fully connected layer, and the score is normalized using a sigmoid activation function. The entity score of the previous jump and the local path score are multiplied to obtain the entity score of the current jump. The attention weight of each jump is obtained according to all indicator vectors, and the entity score of each jump is stacked and multiplied by the attention weight of each jump and summed to obtain the final entity score. According to the final entity score, the entities that meet the standards are screened to generate an expanded heat resistance evaluation index to supplement the key heat resistance evaluation index. Preferably, the fruit set rate is used for knowledge enhancement, and relevant knowledge is obtained through the knowledge graph, for example: the fruit set rate change trend under different temperature treatment conditions is obtained, and the slope of the overall trend line (linear fit) is calculated. The ratio of the overall slope of the rice variety to the overall slope of the control variety is used as the heat resistance coefficient, and the heat resistance of the variety to be tested is judged according to the heat resistance coefficient. Therefore, the overall trend line of the fruit set rate change trend is used as an expanded heat resistance evaluation indicator of the fruit set rate.

[0027] It should be noted that the key heat-resistant evaluation indicators after supplementation are obtained and encoded, and initial heat-resistant evaluation indicator combinations are randomly generated to construct an initial population. The heat-resistant evaluation indicator combination is optimized through a genetic algorithm, and a classification model is constructed using an SVM model. The feature vector corresponding to the initial heat-resistant evaluation indicator combination is imported into the classification model, and the S-fold cross-validation strategy is used to obtain the classification accuracy. The fitness function is constructed using the classification accuracy; the fitness of the heat-resistant evaluation indicator combination is calculated according to the fitness function, and a preset number of heat-resistant evaluation indicator combinations are selected based on the fitness to replicate, thereby expanding the population and obtaining the optimization direction of the heat-resistant evaluation indicator combination; and the selected heat-resistant evaluation indicator combinations are cross-recombined, and important heat-resistant evaluation indicators are merged to accelerate the optimization selection of feature combinations; in addition, the encoding of individual heat-resistant evaluation indicators is changed to increase the possibility of new feature combinations and avoid the occurrence of local optimality. By iteratively optimizing the population, the optimal solution is obtained when the iterative optimization reaches the planting criterion, indicating that the classification accuracy of the heat-resistant evaluation indicator combination has reached an ideal state, and the best key heat-resistant evaluation indicator combination is output.

[0028] Figure 3 A flowchart of an embodiment for constructing a comprehensive evaluation model for heat tolerance of rice is shown.

[0029] According to an embodiment of the present invention, a comprehensive evaluation model for heat tolerance of rice is constructed based on the optimal combination of key heat tolerance evaluation indicators, specifically: S302, extracting single indicator evaluation data from the rice heat tolerance evaluation instance data according to each key heat tolerance evaluation indicator in the optimal key heat tolerance evaluation indicator combination, and generating a data label using the rice heat tolerance evaluation result; S304, constructing a comprehensive evaluation model for heat tolerance of rice using a conditional variational autoencoder, using labeled single-indicator evaluation data for model training, obtaining a latent variable distribution of training samples through an encoder, and reconstructing data using the latent variable distribution through a decoder; S306, combining the rice heat tolerance evaluation result label with the latent variable distribution to train a discriminator, obtaining estimated single-indicator evaluation data under the specific rice heat tolerance evaluation result label generated by data reconstruction, and using the discriminator to determine the data label, and obtaining a trained conditional variational autoencoder through alternating training; S308: Import the estimated single-indicator evaluation data generated by the conditional variational autoencoder into a multi-layer perceptron to obtain the deviation from the test data, obtain the rice heat tolerance evaluation result label classification result based on the deviation, and output the rice heat tolerance comprehensive evaluation model when the classification accuracy meets the preset requirements.

[0030] It should be noted that the conditional variational autoencoder decodes the latent variables output by the encoder and the corresponding category labels of the data by inputting them into the decoder. When generating data, the category labels and the decoder control the type of generated data. The discriminator is introduced into the conditional variational autoencoder to enhance the model's generation capabilities and promote the effective use of input category label information during data generation. The conditional variational autoencoder is trained, and the reconstruction training process includes training the conditional variational autoencoder and training the discriminator using single-metric evaluation data and reconstructed data. The generation training process uses specific rice heat tolerance evaluation result labels to generate estimated single-metric evaluation data of a specific category, and the discriminator determines its category.

[0031] The optimal combination of key heat tolerance evaluation indicators is used to collect indicator parameter values ​​for the rice to be tested. These indicator parameter values ​​are then imported into a comprehensive rice heat tolerance evaluation model to obtain estimated indicator parameter values ​​for each rice heat tolerance evaluation result label. A multi-layer perceptron is used to obtain the residual between the estimated indicator parameter values ​​and the indicator parameter values, and the rice heat tolerance evaluation result label corresponding to the minimum residual is obtained. The heat tolerance evaluation result of the rice to be tested is then output. The preferred rice heat tolerance evaluation result labels are categorized as strong heat tolerance, relatively strong heat tolerance, average heat tolerance, relatively weak heat tolerance, and weak heat tolerance. The rank order can be flexibly modified based on actual conditions to facilitate phenotypic observation and description.

[0032] Figure 4 A block diagram of a comprehensive evaluation system for heat tolerance of rice is shown.

[0033] The second embodiment of the present invention provides a comprehensive evaluation system 4 for heat tolerance of rice, comprising: an evaluation index screening module 401, an evaluation index combination optimization module 402, a data acquisition module 403, a rice heat tolerance evaluation module 404, and an evaluation result output module 405; The evaluation index screening module is responsible for extracting heat tolerance evaluation indicators from rice heat tolerance evaluation instance data, constructing an interaction graph based on the interactive relationship between each heat tolerance evaluation indicator and the rice heat tolerance evaluation task, using a graph neural network to perform representation learning on the interaction graph, and screening key heat tolerance evaluation indicators; The evaluation index combination optimization module uses the domain knowledge graph to enhance the knowledge of key heat-resistant evaluation indicators, constructs extended heat-resistant evaluation indicators to supplement the key heat-resistant evaluation indicators, and selects the best key heat-resistant evaluation indicator combination through feature selection; The data acquisition module is responsible for collecting the index parameter values ​​of the rice to be tested according to the best key heat resistance evaluation index combination; The rice heat tolerance evaluation module constructs a comprehensive evaluation model for rice heat tolerance based on the data reconstruction model combined with the best key heat tolerance evaluation index combination, and imports the index parameter values ​​to obtain the comprehensive evaluation result of rice heat tolerance of the tested rice; The evaluation result output module is responsible for outputting and visually displaying the comprehensive evaluation result of the heat resistance of the rice to be tested in a predicted manner.

[0034] A third aspect of the present invention provides a computer-readable storage medium, which includes a program for a comprehensive evaluation method for heat tolerance of rice. When the program for a comprehensive evaluation method for heat tolerance of rice is executed by a processor, the steps of the comprehensive evaluation method for heat tolerance of rice are implemented.

[0035] In the several embodiments provided in this application, it should be understood that the disclosed methods and systems can be implemented in other ways. The system embodiments described above are merely illustrative. For example, the division of the modules is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple modules or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms. In addition, the functional modules in the various embodiments of the present invention can all be integrated into one processing module, or each module can be a separate module, or two or more modules can be integrated into one module; the above-mentioned integrated modules can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0036] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0037] Alternatively, if the integrated modules of the present invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (such as a personal computer, server, or network device) to execute all or part of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as removable storage devices, ROM, RAM, magnetic disks, or optical disks.

[0038] The above description is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A comprehensive evaluation method for heat tolerance of rice, characterized in that: The following steps are involved: Retrieving and obtaining rice heat tolerance evaluation instance data, extracting heat tolerance evaluation indicators from the rice heat tolerance evaluation instance data, and constructing a heat tolerance evaluation indicator set; constructing an interaction graph based on the interactive relationship between each heat tolerance evaluation indicator in the heat tolerance evaluation indicator set and the rice heat tolerance evaluation task, using a graph neural network to perform representation learning on the interaction graph, and screening key heat tolerance evaluation indicators; Use domain knowledge graphs to enhance the knowledge of key heat resistance evaluation indicators, generate extended heat resistance evaluation indicators to supplement key heat resistance evaluation indicators, and select the best key heat resistance evaluation indicator combination; A comprehensive evaluation model of rice heat resistance is constructed based on the optimal combination of key heat resistance evaluation indicators. The indicator parameter values ​​of the rice to be tested are collected as model input to obtain the comprehensive evaluation results of the rice heat resistance of the rice to be tested.

2. A comprehensive evaluation method for heat tolerance of rice according to claim 1, characterized in that: Retrieve and obtain rice heat tolerance evaluation instance data, extract heat tolerance evaluation indicators from the rice heat tolerance evaluation instance data, and construct a heat tolerance evaluation indicator set, specifically: Use "rice heat tolerance evaluation" as a keyword to obtain relevant descriptive text about rice heat tolerance evaluation. Use the pre-trained BERT model for word embedding. Use the attention mechanism to obtain the global context of different words in the relevant descriptive text and obtain the corresponding global context embedding. Obtaining semantic relative distances in the relevant description text based on position information of different word vectors and keyword vectors, selecting words with a semantic relative distance less than a preset threshold as local contexts, and dynamically weighting the local contexts to obtain local context embedding; The global context embedding and the local context embedding are spliced ​​and imported into a fully connected layer to obtain a corresponding semantic feature sequence, the semantic feature sequence is used in combination with the IVF algorithm and the HNSW algorithm to construct a retrieval index, and the HNSW structure is applied to the inverted index to generate a rice heat tolerance evaluation index; Using the rice heat tolerance evaluation index combined with the cosine distance measurement function to obtain similarity between data, obtaining rice heat tolerance evaluation instance data that meets the similarity standard, and performing structured processing on the rice heat tolerance evaluation instance data; The heat tolerance evaluation indicators involved are extracted from the structured rice heat tolerance evaluation instance data to construct a heat tolerance evaluation indicator set.

3. The comprehensive evaluation method for heat tolerance of rice according to claim 1, characterized in that: The interaction relationship between heat tolerance evaluation indicators and rice heat tolerance evaluation tasks was obtained to construct an interaction graph. A graph neural network was used to learn the representation of the interaction graph and screen key heat tolerance evaluation indicators, specifically: Performing usage frequency statistics on the heat resistance evaluation indicators in the heat resistance evaluation indicator set, performing graph processing on the heat resistance evaluation indicator set, using the heat resistance evaluation indicators and rice heat resistance evaluation tasks as nodes, setting connection relationships between the nodes according to the interactive relationships between different rice heat resistance evaluation tasks and heat resistance evaluation indicators, generating an interaction graph, and setting edge weights in the interaction graph using the usage frequency; Using a graph neural network to perform representation learning on the interaction graph, obtaining feature representations of heat tolerance evaluation index nodes and rice heat tolerance evaluation task nodes, calculating an inner product value using a feature vector of the heat tolerance evaluation index node and a feature vector of the rice heat tolerance evaluation task node, and characterizing the degree of influence of the heat tolerance evaluation index node and the rice heat tolerance evaluation task through the inner product value; And in the graph neural network layer, the inner product value and edge weight of each heat-resistant evaluation index node are used for aggregation to obtain the single-layer neighborhood update representation of the heat-resistant evaluation index node, and the single-layer neighborhood representation of the heat-resistant evaluation index node and the single-layer neighborhood update representation of the heat-resistant evaluation index node are used; By stacking several layers of graph neural networks, the heat resistance evaluation index node and the final feature vector of the heat resistance evaluation index node are obtained, and the CTR estimation model is used to calculate the matching degree of the final feature vector of the heat resistance evaluation index node and the final feature vector of the heat resistance evaluation index node, and the key heat resistance evaluation indicators are screened based on the matching degree.

4. A comprehensive evaluation method for heat tolerance of rice according to claim 1, characterized in that: Use domain knowledge graph to enhance the knowledge of key heat resistance evaluation indicators, generate extended heat resistance evaluation indicators to supplement the key heat resistance evaluation indicators, specifically: Access the knowledge graph related to rice heat tolerance evaluation, and use key heat tolerance evaluation indicators to obtain the corresponding knowledge graph subgraph in the domain knowledge graph; Taking the problem "obtaining extended indicators of key heat resistance evaluation indicators" as input, the problem is vectorized and features are extracted. A multi-head attention mechanism is used to obtain fine-grained information from the vectorized representation of the problem and obtain intermediate reasoning information. Generate an intermediate indicator vector based on the intermediate reasoning information, combine attention with the intermediate indicator vector to obtain a final indicator vector, use the final indicator vector to perform local path matching on the knowledge graph subgraph, and import the entity relationship in the local path into the BERT model to obtain the local path embedding; The local path embedding is scored for similarity using the indicator vector, and the entity score is updated using the similarity score. The entity scores of each hop are stacked and multiplied by the attention and summed to obtain the final entity score. According to the final entity score, entities that meet the standards are screened to generate an expanded heat resistance evaluation index to supplement the key heat resistance evaluation index.

5. The comprehensive evaluation method for heat tolerance of rice according to claim 1, characterized in that: Select the best combination of key heat resistance evaluation indicators, specifically: Obtaining the supplemented key heat-resistance evaluation indicators for encoding, randomly generating an initial heat-resistance evaluation indicator combination to construct an initial population, optimizing the heat-resistance evaluation indicator combination through a genetic algorithm, constructing a classification model using an SVM model, importing the feature vector corresponding to the initial heat-resistance evaluation indicator combination into the classification model, obtaining the classification accuracy, and constructing a fitness function using the classification accuracy; Calculating the fitness of the heat-resistance evaluation index combination according to the fitness function, selecting a preset number of heat-resistance evaluation index combinations for replication based on the fitness to achieve population expansion, and cross-recombining the selected heat-resistance evaluation index combinations to change the coding of individual heat-resistance evaluation indicators therein; The optimal solution is obtained by iteratively optimizing the population and outputting the best combination of key heat resistance evaluation indicators.

6. A comprehensive evaluation method for heat tolerance of rice according to claim 1, characterized in that: A comprehensive evaluation model for heat tolerance of rice was constructed based on the optimal combination of key heat tolerance evaluation indicators, specifically: Extracting single-indicator evaluation data from rice heat tolerance evaluation instance data based on each key heat tolerance evaluation indicator in the optimal key heat tolerance evaluation indicator combination, and generating data labels using the rice heat tolerance evaluation results; A comprehensive evaluation model for rice heat tolerance was constructed using a conditional variational autoencoder. Labeled single-indicator evaluation data was used for model training. The encoder obtained the latent variable distribution of the training samples, and the decoder used this latent variable distribution to reconstruct the data. The rice heat tolerance evaluation result label is combined with the latent variable distribution to train the discriminator, obtain the estimated single indicator evaluation data under the specific rice heat tolerance evaluation result label generated by data reconstruction, and use the discriminator to determine the data label. Through alternating training, a trained conditional variational autoencoder is obtained; The estimated single-indicator evaluation data generated by the conditional variational autoencoder is imported into a multi-layer perceptron to obtain the deviation from the test data. The rice heat tolerance evaluation result label classification result is obtained based on the deviation. When the classification accuracy meets the preset requirements, the comprehensive evaluation model of rice heat tolerance is output.

7. A comprehensive evaluation method for heat tolerance of rice according to claim 1, characterized in that: The index parameter values ​​of the tested rice are collected as model input to obtain the comprehensive evaluation results of the heat tolerance of the tested rice, specifically: Using the best key heat tolerance evaluation index combination to collect index parameter values ​​of the tested rice, the index parameter values ​​are imported into the comprehensive evaluation model of rice heat tolerance to obtain the estimated index parameter values ​​under each rice heat tolerance evaluation result label; The residual between the estimated index parameter value and the index parameter value is obtained through a multi-layer perceptron, the rice heat tolerance evaluation result label corresponding to the minimum residual is obtained, and the heat tolerance evaluation result of the tested rice is output.

8. A comprehensive evaluation system for heat tolerance of rice, characterized in that: Implementing the comprehensive evaluation method for heat tolerance of rice according to any one of claims 1 to 7, the system comprises: an evaluation index screening module, an evaluation index combination optimization module, a data acquisition module, a rice heat tolerance evaluation module and an evaluation result output module; The evaluation index screening module is responsible for extracting heat tolerance evaluation indicators from rice heat tolerance evaluation instance data, constructing an interaction graph based on the interactive relationship between each heat tolerance evaluation indicator and the rice heat tolerance evaluation task, using a graph neural network to perform representation learning on the interaction graph, and screening key heat tolerance evaluation indicators; The evaluation index combination optimization module uses the domain knowledge graph to enhance the knowledge of key heat-resistant evaluation indicators, constructs extended heat-resistant evaluation indicators to supplement the key heat-resistant evaluation indicators, and selects the best key heat-resistant evaluation indicator combination through feature selection; The data acquisition module is responsible for collecting the index parameter values ​​of the rice to be tested according to the best key heat resistance evaluation index combination; The rice heat tolerance evaluation module constructs a comprehensive evaluation model for rice heat tolerance based on the data reconstruction model combined with the best key heat tolerance evaluation index combination, and imports the index parameter values ​​to obtain the comprehensive evaluation result of rice heat tolerance of the tested rice; The evaluation result output module is responsible for outputting and visually displaying the comprehensive evaluation result of the heat resistance of the rice to be tested in a predicted manner.

Images