Optimal parent combination matching method based on ML algorithm, electronic equipment and medium

Abstract

The invention belongs to the technical field of animal and plant breeding, and discloses an optimal parent combination matching method based on an ML algorithm, electronic equipment and a storage medium, and the optimal parent combination matching method based on the ML algorithm comprises the following steps: obtaining gender information, genotype data and phenotype data of a seed sample; performing cross validation on the seed sample based on an ML algorithm to obtain an optimal model; performing permutation and combination on the seed samples according to the gender information to obtain a parent combination condition; calculating the breeding ability of all parent combination conditions; an optimal model is adopted, and an optimal parent combination is recognized according to the breeding capacity. According to the method, the LGBM algorithm is utilized to automatically extract and select important features related to target traits from massive seed sample data, and a prediction model suitable for agricultural breeding is constructed. The model is optimized through a cross validation method, and the accuracy of the model is improved. And the certainty of breeding is improved by calculating the breeding capability of all parent combination conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of animal and plant breeding, and in particular relates to an optimal parent combination selection method based on an ML algorithm, an electronic device and a storage medium. Background Art

[0002] Agricultural breeding is a complex and time-consuming task, and traditional methods mainly rely on expert experience and a large number of field trials. However, with the development of genomics and information technology, more and more large-scale data can be collected and analyzed, and traditional methods are inefficient and difficult to make full use of these data.

[0003] In the related art, when traditional models are used for seed selection, the operating efficiency and accuracy are not high when the number of markers and samples is large. Summary of the invention

[0004] In view of this, the present invention provides an optimal parent combination selection method based on ML algorithm, an electronic device and a storage medium to solve the problems of low selection efficiency and low accuracy in the prior art.

[0005] In a first aspect, the present invention provides a method for selecting an optimal parent combination based on an ML algorithm, comprising: obtaining sex information, genotype data and phenotypic data of a seed sample; cross-validating the seed sample based on an ML algorithm to obtain an optimal model; arranging and combining the seed samples according to the sex information to obtain a parent combination; calculating the breeding ability of all the parent combinations; using the optimal model and identifying the optimal parent combination based on the breeding ability.

[0006] In some embodiments, the seed samples are cross-validated based on the ML algorithm to obtain the optimal model, specifically including: constructing a preliminary model based on the seed samples using the LGBM machine algorithm; cross-validating the preliminary model to obtain the true situation of the model; calculating the optimized hyperparameters based on the true situation of the model using the Bayesian algorithm; and obtaining the optimal model based on the optimized hyperparameters.

[0007] In some embodiments, the seed samples are cross-validated based on the ML algorithm to obtain the optimal model, which specifically also includes: constructing an intermediate model based on the optimized hyperparameters; using a cross-validation method to identify four indicators of the intermediate model; and obtaining the optimal model based on the four indicators.

[0008] In some embodiments, the number of cross-validation is ten times.

[0009] In some embodiments, any of the methods for selecting the optimal parent combination based on the ML algorithm further includes: cleaning the genotype data of the seed sample to obtain quality-type genotype data; cleaning the phenotypic data of the seed sample to obtain quality-type phenotypic data; obtaining a quality sample based on the quality-type genotype data and the quality-type phenotypic data; wherein the optimal model is obtained based on the quality sample.

[0010] In some embodiments, the optimal parent combination selection method based on the ML algorithm further includes: evaluating the optimal model using the root mean square error and the Pearson correlation coefficient to obtain an evaluation graph.

[0011] In some embodiments, the seed samples are arranged and combined according to the gender information to obtain the parent combination, which specifically includes: simulating the genotype probability of each combination offspring; calculating the breeding value of each combination according to the genotype probability to obtain the parent combination.

[0012] In some embodiments, the optimal model is used, and the optimal parent combination is identified according to the breeding ability, specifically comprising: sorting the breeding values ​​from high to low to obtain a sorting result; and identifying the optimal parent combination according to the sorting result.

[0013] In a second aspect, the present invention also provides an electronic device, comprising: at least one processor; a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: obtain the gender information of the seed sample; cross-validate the seed sample based on the ML algorithm to obtain the optimal model; arrange and combine the seed samples according to the gender information to obtain the parent combination; calculate the breeding ability of all the parent combinations; adopt the optimal model and identify the optimal parent combination based on the breeding ability.

[0014] In a third aspect, the present invention provides a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements any one of the optimal parent combination selection methods based on the ML algorithm.

[0015] The beneficial effects of the present invention are:

[0016] The present invention uses the LGBM algorithm to automatically extract and select important features related to target traits from massive seed sample data, and construct a prediction model suitable for agricultural breeding. The model is optimized by the cross-validation method to improve the accuracy of the model. The breeding capacity of all parent combinations is calculated to improve the certainty of breeding. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods 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 This is a flow chart of an optimal parent combination matching method based on an ML algorithm according to an embodiment of the present invention;

[0019] Figure 2 This is a flow chart of an optimal parent combination matching method based on an ML algorithm according to another embodiment of the present invention;

[0020] Figure 3 A model marker importance ranking diagram of an optimal parent combination matching method based on an ML algorithm according to an embodiment of the present invention;

[0021] Figure 4 A distribution diagram of observed values ​​and predicted values ​​under ten-fold crossover of an optimal parent combination matching method based on an ML algorithm according to an embodiment of the present invention;

[0022] Figure 5 A histogram of offspring breeding values ​​of all parent combinations of the optimal parent combination selection method based on the ML algorithm in one embodiment of the present invention;

[0023] Figure 6 The figure is a block diagram of an electronic device according to an embodiment of the present invention.

[0024] Description of reference numerals:

[0025] 110. Processor;

[0026] 120. Memory. DETAILED DESCRIPTION

[0027] In the history of selective breeding with related technologies, we have gone through the exploration from empirical breeding to breeding theories and methods, including selection theory, pure line theory, backcross breeding, reincarnation breeding, mutagenesis breeding, single seed transmission, and ideal plant type; then to molecular marker-assisted selection breeding, we have explored various molecular markers, such as amplified fragment length polymorphism marker-assisted selection (AFLP), microsatellite marker-assisted selection (SSR) and single nucleotide polymorphism marker-assisted selection (SNP); with the development of sequencing technology, the sequencing throughput is getting higher and higher, the cost is getting lower and lower, and the computer computing power is constantly improving, which has created technical conditions for the development of new breeding technology and set off a wave of genomic selection (GS) breeding.

[0028] Genomic selection breeding can effectively solve the limitations of difficult-to-measure traits, high luck, long time consumption, high technical difficulty and other factors, and accelerate the pace of breeding. Genomic selection breeding is a breeding method that uses marker-assisted selection with high-density molecular genetic markers covering the entire genome. With the introduction of the concept of "smart breeding", smart breeding has attracted more and more attention from breeders and is gradually moving towards application. Smart breeding will become the core driving force for the development of the seed industry. From traditional breeding to molecular breeding to smart breeding, the "scientific" component of breeding is increasing. The "scientific" component mainly comes from the construction and upgrading of the smart breeding platform, establishing a crop breeding data extraction, mining, storage, analysis, and sharing database based on germplasm resource information, integrating multi-omics data for joint analysis, breaking through the underlying supporting technologies such as biological big data acquisition, analysis, and mining, and establishing an integrated smart breeding strategy of "data-technology-algorithm-decision-making", improving breeding accuracy, shortening the breeding cycle, and establishing an efficient crop smart breeding system.

[0029] In order to accelerate the development of excellent promotion varieties, the selection of the best parent combination is the prerequisite for the formation of excellent promotion varieties. In modern agricultural production, the development of excellent promotion varieties is of great significance for improving crop yields, resistance to pests and diseases, and adapting to different environmental conditions. In order to achieve this goal, the selection of the best parent combination is crucial.

[0030] The present invention proposes an optimal parent combination selection method based on the LGBM algorithm, aiming to solve the problems of low efficiency, low success rate and high cost in the traditional breeding process and promote the rapid development of agricultural breeding technology.

[0031] Based on the lightGBM algorithm, a complete solution from data input to optimal parent combination selection results is established, which can quickly and accurately obtain the desired optimal parent combination solution with one click.

[0032] The machine learning lightGBM algorithm is a very excellent machine learning algorithm at this stage. LightGBM (Light Gradient Boosting Machine) is a framework based on the GBDT algorithm (using weak classifiers to iteratively train to obtain the optimal model), supports efficient parallel training, and has faster training speed, lower memory consumption, better accuracy, supports distributed and can quickly process massive data, etc.:

[0033] 1) Optimize speed and memory usage (histogram-based algorithms optimization).

[0034] 2) Mutually exclusive feature bundling algorithm (fusing and bundling some features can reduce the number of features and reduce the dimension).

[0035] 3) Sparse optimization (only the histogram needs to be constructed for sparse features, the genotype matrix is ​​a sparse matrix, and only a small number of sites are valid sites).

[0036] 4) Accuracy optimization (Leaf-wise (Best-first) Tree Growth strategy and Optimal Split for Categorical Features).

[0037] 5) Optimization of distributed learning (Feature Parallel, Data Parallel, Voting Parallel).

[0038] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0039] Combine the following Figures 1 to 6 , describing an embodiment of the present invention.

[0040] like Figure 1 and Figure 2 As shown, according to an embodiment of the present invention, on the one hand, a method for selecting an optimal parent combination based on an ML algorithm is provided, comprising the following steps:

[0041] Step S101: Obtaining the gender information, genotype data and phenotype data of the seed sample.

[0042] Step S103: Cross-validate the seed samples based on the ML algorithm to obtain the optimal model.

[0043] Step S105: Arrange and combine the seed samples according to the gender information to obtain the parent combination.

[0044] Step S107: Calculate the breeding capabilities of all parent combinations.

[0045] Step S109: Use the optimal model and identify the optimal parent combination based on breeding ability.

[0046] The seeds of the present invention are suitable for most material backgrounds, such as most plants such as rice, corn, wheat, cotton, pigs, cattle, sheep, chickens, ducks, geese and other animals, and most types of variation, such as SNP, InDel and SV of resequencing or chip data, CNV data, etc. This application can effectively reduce cluster resource consumption, reduce labor costs and time costs, improve the accuracy of breeding, and promote the progress of breeding.

[0047] Gender information may include male or female. In practical applications, seed samples may be referred to as parent samples.

[0048] This application uses the powerful feature selection capability of the LGBM algorithm to automatically extract and select important features related to the target trait from a large amount of seed sample data, and construct a prediction model suitable for agricultural breeding. The model is optimized by the cross-validation method to improve the accuracy of the model. Then, by calculating the breeding ability of all parent combinations, the breeding uncertainty, breeding time cost, and breeding manpower cost are reduced. As a result, a certain excellent trait in a donor variety is integrated into a recipient variety with most excellent traits but poor target traits, so that the excellent traits are integrated to form a more excellent promotion variety.

[0049] Further, step S103, cross-validating the seed samples based on the ML algorithm to obtain the optimal model, specifically includes the following steps:

[0050] Step S1031: Based on the seed sample, a preliminary model is constructed using the LGBM machine algorithm;

[0051] Step S1033: cross-validate the preliminary model to obtain the true situation of the model;

[0052] Step S1035: Calculate the optimized hyperparameters using the Bayesian algorithm according to the actual situation of the model;

[0053] Step S1036: Obtain the optimal model by optimizing the hyperparameters.

[0054] In this embodiment, the preliminary model constructed by the lightGBM algorithm is a preliminary GS model, and 10-fold cross validation is used to reflect the actual situation of the model, and then the Bayes algorithm is used to optimize and find the optimal model hyperparameters. Among them, the LGBM algorithm is used for training and testing, and the Bayesian optimization method is used to find the optimal hyperparameters to build the optimal model, which can obtain a better evaluation value.

[0055] Furthermore, step S103, cross-validating the seed samples based on the ML algorithm to obtain the optimal model, specifically includes the following steps:

[0056] Step S1037: constructing an intermediate model according to the optimized hyperparameters;

[0057] Step S1038: using a cross-validation method to identify four indicators of the intermediate model;

[0058] Step S1039: Obtain the optimal model based on the four indicators.

[0059] The number of cross validations is ten.

[0060] In this embodiment, the intermediate model is a model obtained by further constructing the GS model, and the 10-fold cross-validation method is used to calculate the four indicators of the optimized model, which include: Pearson Correlation Coefficient (PCC), Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE). The data set is divided into ten subsets (folds), one subset is used as a validation set in each fold, and the remaining subsets are used as training sets, and this process is repeated ten times. Finally, the results of the ten validations are averaged to obtain the overall performance indicators of the model.

[0061] Furthermore, the optimal parent combination selection method based on the ML algorithm also includes the following steps:

[0062] Step S201: cleaning the genotype data of the seed sample to obtain quality genotype data;

[0063] Step S203: cleaning the phenotypic data of the seed sample to obtain quality phenotypic data;

[0064] Step S205: obtaining a quality sample according to the quality genotype data and the quality phenotype data;

[0065] Among them, the optimal model is obtained based on quality samples.

[0066] In this embodiment, the genotypes of the parents are filled to obtain a genotype file without missing values, and the samples with missing values ​​in the phenotype file are removed to obtain a phenotype file without missing values. The phenotype data and genotype data of the parents are cleaned to obtain high-quality data without missing values.

[0067] Among them, the root square error and Pearson correlation coefficient are used to evaluate the optimal model and obtain the evaluation graph. The machine learning algorithm is used for training, the 10-fold cross-validation method is used, the LGBM machine algorithm is used for Bayesian optimization of hyperparameters, the root square error and Pearson correlation coefficient are used to evaluate the model, and statistics and graphs are drawn. The genotype probability of each combination offspring is simulated, and the breeding value of each combination is calculated, and then the optimal parent combination result of the specified threshold is screened.

[0068] Further, based on step S105, the seed samples are arranged and combined according to the gender information to obtain the parent combination, which specifically includes the following steps:

[0069] Step S1051: Simulate the genotype probability of each combination offspring;

[0070] Step S1053: Calculate the breeding value of each combination according to the genotype probability to obtain the parent combination situation.

[0071] In this embodiment, all combinations of paternal and maternal parents are taken into consideration and combined, and the genotype probability of the offspring is calculated for each combination, thereby obtaining a huge offspring genotype probability matrix of the parent combination and obtaining its matrix file; finally, the offspring genotype probability matrix file is substituted into the constructed optimal model, the breeding ability of all parent combinations is calculated, and then the best few or top1% or top5% parent combinations are selected for breeding.

[0072] Further, based on step S109, the optimal model is adopted, and the optimal parent combination is identified according to the breeding ability, which specifically includes the following steps:

[0073] Step S1091: sorting from high to low according to the breeding values ​​to obtain a sorting result;

[0074] Step S1093: Identify the optimal parent combination according to the sorting results.

[0075] like Figure 6 As shown, according to an embodiment of the present invention, on the other hand, an electronic device is provided, including: at least one processor 110; a memory 120 communicatively connected to the at least one processor, wherein the memory 120 stores instructions executable by the at least one processor 110, and the instructions are executed by the at least one processor 110, so that the at least one processor 110 can perform the following steps:

[0076] Step S101: Obtaining the gender information of the seed sample, where the gender information includes genotype data and phenotype data.

[0077] Step S103: Cross-validate the seed samples based on the ML algorithm to obtain the optimal model.

[0078] Step S105: Arrange and combine the seed samples according to the gender information to obtain the parent combination.

[0079] Step S107: Calculate the breeding capabilities of all parent combinations.

[0080] Step S109: Use the optimal model and identify the optimal parent combination based on breeding ability.

[0081] According to an embodiment of the present invention, on the other hand, a storage medium is provided, storing a computer program, and when the computer program is executed by a processor, any one of the optimal parent combination selection methods based on the ML algorithm is implemented.

[0082] The computer executes the instructions to perform the following steps:

[0083] Step S101: Obtaining the gender information of the seed sample, where the gender information includes genotype data and phenotype data.

[0084] Step S103: Cross-validate the seed samples based on the ML algorithm to obtain the optimal model.

[0085] Step S105: Arrange and combine the seed samples according to the gender information to obtain the parent combination.

[0086] Step S107: Calculate the breeding capabilities of all parent combinations.

[0087] Step S109: Use the optimal model and identify the optimal parent combination based on breeding ability.

[0088] The software of the present invention obtains high-quality phenotypic and genotypic data of the father and mother after data cleaning; then uses the lightGBM algorithm to build a model, and after Bayesian algorithm optimization, finds the optimal hyperparameters to build the optimal model; then simulates the genotype probabilities of all offspring of different parent combinations, and jointly calculates the average breeding values ​​of different parent combinations through the genotype probabilities of the offspring and the model file; finally, selects the parent combination with the best breeding value performance for parent construction of breeding new varieties, and generally takes the top 5% or top 1% of the breeding value parent combination to construct the material for variety breeding.

[0089] ###Software Description###

[0090]

[0091] Software parameter description:

[0092] -h helps explain the documentation.

[0093] -vcf parental genotype data vcf file.

[0094] -pheno Parental phenotype data file.

[0095] -sex parents' gender information file.

[0096] -model model algorithm, the default is LGBM algorithm.

[0097] -out Output directory for result files.

[0098] -select selects the number of combinations before the optimal parent. If it is an integer n, select the first n combinations before the optimal parent; if it is a floating point number t, select t*100% combinations before the optimal parent.

[0099] -low If this parameter is not used, the screening starts from the highest breeding value; if this parameter is used, the screening starts from the lowest breeding value.

[0100] -analysis analysis procedure. The first step is genotype data cleaning; the second step is phenotype data cleaning; the third step is machine learning algorithm training; the fourth step is to select the optimal parent combination.

[0101] -config configuration file, no need to use this parameter, it has already been configured

[0102] -parallel number of threads, default is 4.

[0103] The software input file format is as follows:

[0104] Parental genotype format:

[0105]

[0106] Parental phenotype file format:

[0107] The first column is the sample name, and the second column is the phenotype value;

[0108]

[0109] Format of the gender document of the parents:

[0110] The first column is the sample name, and the second column is the gender.

[0111]

[0112] Results of the test data:

[0113] Using the phenotypic and variation data of 6210 samples of public corn, 50 samples were randomly selected as female plants and 139 samples as male plants. The model was constructed, and the genotype probability of the offspring was simulated. The ability value of each parent combination was calculated, and finally the best parent combination was selected.

[0114] Figure 3 It is a ranking chart of the constructed model mark importance. The horizontal axis is the mark importance value from high to low from left to right, the left vertical axis is the mark importance value, and the right vertical axis is the accumulated mark importance value.

[0115] Figure 4 This is the distribution diagram of the observed values ​​and predicted values ​​of the model constructed by parent-based training under ten-fold crossover. The horizontal axis is the observed value, the vertical axis is the predicted value, and the black solid line is the trend line simulated by univariate linear regression. The figure contains the corresponding formula, RMSE value and PCC value. The PCC value is 0.428.

[0116] Figure 5 It is a histogram of the breeding values ​​of offspring of all parent combinations. The horizontal axis is the predicted breeding value distribution and the vertical axis is the frequency.

[0117] Table 1 is the evaluation table of the four indicators of the constructed model, with MSE of 0.85, RMSE of 0.92, MAE of 0.76, and PCCs of 0.43.

[0118] Table 1

[0119] MSE RMSE MAE PCCs 0.852181613 0.923136833 0.763593244 0.427663477

[0120] Table 2 is a partial result of the offspring breeding value table for each parent combination. From the table, you can sort from high to low, and select the results of the "top 5%" or "top 1%" or "top 10" combinations as needed to construct the material for breeding offspring. Usually, the top 1% combination is selected to breed offspring to obtain offspring with high breeding values ​​to achieve the purpose of variety improvement. For example, you can use CUBIC_100_X_Tester_25 material and CUBIC_107_X_Tester_25 material as the parent combination to construct the material for offspring breeding.

[0121] Table 2 Offspring breeding values ​​of each parent combination (partial)

[0122] ID Breed_value CUBIC_100_X_Tester_25.x.CUBIC_107_X_Tester_25 0.832627823 CUBIC_100_X_Tester_25.x.CUBIC_109_X_Tester_25 0.350533724 CUBIC_100_X_Tester_25.x.CUBIC_110_X_Tester_25 0.183775324 CUBIC_100_X_Tester_25.x.CUBIC_116_X_Tester_25 0.27458238 CUBIC_100_X_Tester_25.x.CUBIC_124_X_Tester_25 -1.35221969 CUBIC_100_X_Tester_25.x.CUBIC_127_X_Tester_25 -0.791093617 CUBIC_100_X_Tester_25.x.CUBIC_134_X_Tester_25 -1.252725172 CUBIC_100_X_Tester_25.x.CUBIC_136_X_Tester_25 -0.644840799 CUBIC_100_X_Tester_25.x.CUBIC_145_X_Tester_25 -0.40031842 CUBIC_100_X_Tester_25.x.CUBIC_14_X_Tester_25 0.323767095 CUBIC_100_X_Tester_25.x.CUBIC_162_X_Tester_25 -0.899258775 CUBIC_100_X_Tester_25.x.CUBIC_166_X_Tester_25 -0.091380995

[0123] The present invention is not only applicable to high-density markers (SNPs, InDels, SVs) for high-throughput sequencing, but can also be used for chip and simplified genome sequencing needs, providing a rapid breeding platform for breeders to select the optimal parent combination to cultivate their own varieties.

[0124] The technical problem to be solved by the present invention is how to efficiently and accurately select the optimal parent combination in the agricultural breeding process based on the LGBM (LightGBM, lightweight gradient boosting machine) algorithm, thereby accelerating the breeding process and improving the breeding effect. The powerful feature selection ability of the LGBM algorithm is used to automatically extract and select important features related to the target traits from massive data to construct a prediction model suitable for agricultural breeding. The model is optimized by cross-validation, hyperparameter tuning and other methods to improve the model accuracy. The optimal parent combination is then selected by simulating the breeding values ​​of the offspring of different parent combinations, which reduces breeding uncertainty, reduces the time cost of breeding, and reduces the labor cost of breeding.

[0125] The main beneficial effects of the present invention are as follows:

[0126] 1. Use the excellent machine learning LGBM algorithm to improve accuracy.

[0127] 2. The accuracy of the model is improved by finding the optimal hyperparameters through the Bayesian optimization algorithm.

[0128] 3. Parent combination screening can be performed without investigating the phenotype of the offspring.

[0129] 4. High operating efficiency, the optimal parent combination can be obtained in a short time.

[0130] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0131] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for selecting optimal parent combinations based on ML algorithm, characterized in that: include: Obtain sex information, genotype data, and phenotypic data of seed samples; Cross-validating the seed samples based on the ML algorithm to obtain an optimal model; Arrange and combine the seed samples according to the gender information to obtain parent combination information; Calculate the breeding ability of all parent combinations; The optimal model is used to identify the optimal parent combination based on the breeding ability.

2. The optimal parent combination matching method based on ML algorithm according to claim 1, characterized in that: The seed samples are cross-validated based on the ML algorithm to obtain the optimal model, which specifically includes: Based on the seed samples, a preliminary model is constructed using the LGBM machine algorithm; Cross-validating the preliminary model to obtain the true situation of the model; According to the actual situation of the model, the Bayesian algorithm is used to calculate the optimized hyperparameters; The optimal model is obtained according to the optimized hyperparameters.

3. The optimal parent combination matching method based on ML algorithm according to claim 2 is characterized in that: Cross-validating the seed samples based on the ML algorithm to obtain the optimal model specifically includes: Building an intermediate model according to the optimized hyperparameters; The four indicators of the intermediate model were identified using cross-validation method; The optimal model is obtained according to the four indicators.

4. The optimal parent combination matching method based on ML algorithm according to claim 2, characterized in that: The number of cross validations is ten.

5. The optimal parent combination matching method based on ML algorithm according to any one of claims 1 to 4, characterized in that: Also includes: Cleaning the genotype data of the seed sample to obtain quality genotype data; Cleaning the phenotypic data of the seed sample to obtain quality phenotypic data; Obtaining a quality sample according to the quality genotype data and the quality phenotype data; Wherein, the optimal model is obtained based on the quality samples.

6. The optimal parent combination matching method based on ML algorithm according to claim 5, characterized in that: Also includes: The optimal model is evaluated using the root mean square error and the Pearson correlation coefficient to obtain an evaluation graph.

7. The optimal parent combination matching method based on ML algorithm according to any one of claims 1 to 4, characterized in that: The seed samples are arranged and combined according to the gender information to obtain parent combination information, which specifically includes: Simulate the genotype probabilities of the offspring for each combination; The breeding value of each combination is calculated according to the genotype probability to obtain the parent combination situation.

8. The optimal parent combination matching method based on ML algorithm according to claim 7, characterized in that: The optimal model is used, and the optimal parent combination is identified according to the breeding ability, specifically including: Sorting the breeding values ​​from high to low to obtain a sorting result; The optimal parent combination is identified according to the sorting result.

9. An electronic device, characterized in that: include: at least one processor (110); a memory (120) communicatively coupled to the at least one processor (110); The memory (120) stores instructions executable by the at least one processor (110), and the instructions are executed by the at least one processor (110) to enable the at least one processor (110) to: Obtain gender information of seed samples; Cross-validating the seed samples based on the ML algorithm to obtain an optimal model; Arrange and combine the seed samples according to the gender information to obtain parent combination information; Calculate the breeding ability of all parent combinations; The optimal model is used to identify the optimal parent combination based on the breeding ability.

10. A storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a processor, the optimal parent combination matching method based on the ML algorithm as described in any one of claims 1 to 8 is implemented.

Images