High-temperature weather cow stress resistance breeding regulation method based on intelligent big data analysis

Abstract

The application provides a high-temperature weather dairy cow stress resistance breeding regulation method based on intelligent big data analysis, and relates to the field of dairy cow breeding technology. It comprises the following steps: constructing a multi-dimensional dairy cow stress resistance data collection system, collecting dairy cow basic information, phenotypes, molecular and environmental stress data under high-temperature environment; preprocessing and feature engineering of the initial dairy cow data for stress resistance orientation, and constructing a high-temperature stress resistance feature library; constructing a breeding regulation model based on the stress resistance feature vector, and optimizing according to molecular marker assisted selection and distributed learning; obtaining a personalized stress resistance breeding regulation scheme based on the optimized stress resistance breeding regulation model; coping with unknown high-temperature stress and new stress resistance demand, starting zero-base learning and emergency breeding strategy. The application adopts an XGBoost-BN hybrid model to construct a breeding regulation model, migrates learning and optimizes the model through distributed collaborative filtering learning, and realizes continuous optimization and adaptive improvement of the breeding regulation system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of charging fault diagnosis, and in particular to a high-temperature weather dairy cow stress resistance breeding regulation method based on intelligent big data analysis. BACKGROUND

[0002] Holstein cows have poor heat tolerance, and when the temperature exceeds 26℃, the cows will have obvious heat stress response, milk yield will decrease, disease resistance will decrease, and severe cases can lead to an increase in the culling rate of cows, so it is of great significance to strengthen the research on the formation of the mechanism of cow stress resistance, cultivate a core group of stress resistance, and develop targeted stress resistance feed to alleviate stress response and improve production.

[0003] Dairy cow mastitis is one of the most common, most difficult to prevent and treat, and most expensive dairy cow diseases worldwide, and is also one of the main diseases affecting the development of the dairy industry.

[0004] Traditional breeding methods mainly rely on phenotypic observation and pedigree analysis, and have defects such as long cycle, low selection efficiency, and difficulty in accurately mining stress resistance related genetic loci. With the rapid development of big data technology, artificial intelligence and molecular biology, it is possible to use intelligent big data analysis to integrate multi-dimensional data of dairy cows and achieve accurate regulation of stress resistance breeding. However, in the prior art, the data collection dimension is single, and is mostly limited to basic phenotypic data such as milk yield and body temperature, lacking integration and analysis of molecular level data such as genome and transcriptome and dynamic data of environmental stress; at the same time, the breeding regulation model is mostly a static model, which cannot dynamically adjust the breeding strategy according to real-time high-temperature environmental changes and individual differences of dairy cows, resulting in poor regulation effect and difficulty in meeting the needs of accurate and efficient dairy cow stress resistance breeding under high-temperature weather. SUMMARY

[0005] The present application provides a high-temperature weather dairy cow stress resistance breeding regulation method based on intelligent big data analysis to solve the defects in the prior art.

[0006] In one aspect, the present application provides a high-temperature weather dairy cow stress resistance breeding regulation method based on intelligent big data analysis, comprising:

[0007] S1: Constructing a multi-dimensional dairy cow stress resistance data collection system, collecting dairy cow basic information, phenotypes, molecules and environmental stress data under high-temperature environment, and obtaining initial dairy cow data;

[0008] S2: Preprocessing and feature engineering of the initial dairy cow data for stress resistance orientation, constructing a high-temperature stress resistance feature library, and obtaining a standardized stress resistance feature vector;

[0009] S3: Constructing a breeding regulation model according to the stress resistance feature vector, optimizing the breeding regulation model according to molecular markers and distributed learning, and obtaining an optimized stress resistance breeding regulation model;

[0010] S4: breeding control is carried out according to the optimized stress resistance breeding control model to obtain a core group stage breeding result, a dynamic adjustment is made to the control scheme according to the breeding result, and a personalized stress resistance breeding control scheme is obtained;

[0011] S5: in response to unknown high-temperature stress and new stress resistance requirements, a zero-base learning and emergency breeding strategy is started on the basis of the personalized stress resistance breeding control scheme.

[0012] According to the high-temperature weather dairy cow stress resistance breeding control method based on intelligent big data analysis provided by the application, in step S1, the specific steps of obtaining the initial dairy cow data are:

[0013] S11: determine the multi-dimensional dairy cow stress resistance data object.

[0014] S12: deploy multi-type data collection equipment.

[0015] S13: establish a data transmission and storage mechanism, the collection equipment transmits real-time data to the edge computing node, after format conversion and redundant data elimination, upload to the distributed database, set up off-site double backup to ensure data security, obtain the initial dairy cow data.

[0016] According to the high-temperature weather dairy cow stress resistance breeding control method based on intelligent big data analysis provided by the application, in step S2, the specific steps of obtaining the standardized stress resistance feature vector are:

[0017] S21: the stress resistance data is cleaned from the abnormal value and the missing value data, and the cleaned dairy cow data is obtained.

[0018] S22: eliminate the dimension difference of the cleaned dairy cow data, and obtain the standard dairy cow data.

[0019] S23: extract and select stress resistance features from the standard dairy cow data, and construct a high-temperature stress resistance feature library.

[0020] S24: the core features are weighted and fused by using the static weighting method based on feature importance, the weight is dynamically allocated according to the high-temperature scene, and the standardized stress resistance feature vector is obtained.

[0021] According to the high-temperature weather dairy cow stress resistance breeding control method based on intelligent big data analysis provided by the application, in step S23, the specific steps of constructing the high-temperature stress resistance feature library are:

[0022] S231: extract indexes directly reflecting stress resistance from the standard dairy cow data.

[0023] S232: according to the stress resistance index, the stress resistance correlation is strengthened by feature combination.

[0024] S233: Screen the correlation features related to stress resistance by using variance, calculate the feature importance according to the decision tree algorithm, retain the features with importance greater than a preset threshold, and construct a high-temperature stress resistance feature library.

[0025] According to the high-temperature weather dairy cow stress resistance breeding regulation method based on intelligent big data analysis provided by the application, in step S3, the specific steps for obtaining the optimized stress resistance breeding regulation model are:

[0026] S31: Calculate the individual stress resistance comprehensive score according to the stress resistance feature vector, preliminarily select the stress resistance core data according to the score, and divide the stress resistance core data into a training set, a validation set and a test set by using stratified sampling.

[0027] S32: Use the XGBoost-BN hybrid model to construct an initial stress resistance breeding regulation model, and fuse molecular markers and phenotype data.

[0028] S33: Adopt transfer learning and distributed collaborative filtering learning to optimize the model, adapt to the stress resistance breeding scene, and obtain an optimized stress resistance breeding regulation model.

[0029] According to the high-temperature weather dairy cow stress resistance breeding regulation method based on intelligent big data analysis provided by the application, in step S32, the specific steps for constructing the initial stress resistance breeding regulation model using the XGBoost-BN hybrid model are:

[0030] S321: Standardize the molecular marker data and phenotype data of the training set data, output the standardized feature matrix, and preliminarily screen the molecular marker-phenotype potential causal relationship to obtain a molecular marker causal relationship list.

[0031] S322: Input the standardized feature matrix into the XGBoost module for training, take the stress resistance comprehensive score as the prediction target, optimize the model hyperparameters through cross-validation, obtain a stress resistance score prediction model, calculate and output the feature importance ranking based on the decision tree node split gain.

[0032] S323: Adjust the directed edges between nodes based on the molecular marker causal relationship list and the feature importance ranking, calculate the conditional probability table of each node through the training set data, optimize the Bayesian network structure, and obtain a stress resistance causal network.

[0033] S324: Fuse the stress resistance score prediction model and the stress resistance causal network, and output an initial stress resistance breeding regulation model.

[0034] According to the high-temperature weather dairy cow stress resistance breeding regulation method based on intelligent big data analysis provided by the application, in step S33, the specific steps for adopting transfer learning and distributed collaborative filtering learning to optimize the model are:

[0035] S331: The XGBoost-BN model trained using normal temperature data is used as a migration source, and the phenotype and general molecular marker processing data are migrated to the initial model to output the XGBoost-BN model to be optimized.

[0036] S332: The master node performs sharding processing based on the XGBoost-BN model to be optimized, allocates a dedicated model shard to each pasture edge node, and obtains a sharded model module and a node allocation scheme.

[0037] S333: Each pasture edge node trains the XGBoost-BN model to be optimized according to the sharded model module and the node allocation scheme, and outputs a sharded parameter update amount and a feature embedding vector.

[0038] S334: The sharded parameter update amount and the feature embedding vector are subjected to sharding consistency verification, the global parameters are aggregated according to double weights, and the complete model is reorganized to output the optimized resistance breeding regulation model.

[0039] According to the high-temperature weather dairy cow resistance breeding regulation method based on intelligent big data analysis provided by the application, in step S4, the specific steps of outputting the adjusted regulation scheme are:

[0040] S41: Fine breeding regulation is performed according to the optimized resistance breeding regulation model, and the core group stage selection result is output.

[0041] S42: The general production group improvement regulation is implemented according to the core group stage selection result, and the improvement regulation implementation record is output.

[0042] S43: A multi-dimensional evaluation system is established based on the improvement regulation implementation record to verify the effect of the regulation scheme, and the verification result data is output.

[0043] S44: The regulation scheme is dynamically adjusted according to the verification result data, and the adjusted regulation scheme is output.

[0044] According to the high-temperature weather dairy cow resistance breeding regulation method based on intelligent big data analysis provided by the application, in step S5, the specific steps of zero-based learning are:

[0045] S51: Collect real-time resistance characteristics, phenotypes and environmental data of dairy cows, and when the cosine similarity is less than the preset threshold, determine that it is an unknown resistance problem, classify the unknown resistance problem, and obtain the resistance problem classification.

[0046] S52: According to the resistance problem classification, the association between abnormal phenotypes and known feature chains is mined through a meta-learning algorithm to construct a new problem-strategy association model.

[0047] S53: Constructing a physical rule base according to a new problem-strategy association model, integrating high-temperature physiological metabolic rules, molecular regulation rules, and breeding emergency rules, and generating emergency strategies according to different types of problems.

[0048] According to the high-temperature weather dairy cow stress resistance breeding control method based on intelligent big data analysis provided by the application, in step S52, the specific steps of constructing the new problem-strategy association model are:

[0049] S521: Extracting abnormal phenotype characteristics in the non-stress problem classification.

[0050] S522: Mapping the abnormal phenotype characteristics to the known feature space, and using the meta-learning algorithm to find the association between the abnormal phenotype and the known feature chain.

[0051] S523: Extracting the conserved stress resistance mechanism of the close relative species.

[0052] S524: Combining the association between the abnormal phenotype and the known feature chain with the conserved stress resistance mechanism of the close relative species to construct a new problem-strategy association model.

[0053] The high-temperature weather dairy cow stress resistance breeding control method based on intelligent big data analysis provided by the application has the following beneficial effects:

[0054] The XGBoost-BN hybrid model architecture provided by the application combines the advantages of machine learning and causal reasoning. The XGBoost module realizes high-precision stress resistance score prediction through decision tree integration, while the Bayesian network constructs a genotype-phenotype-stress resistance causal network, which can intuitively show how molecular markers affect stress resistance by affecting phenotype. This hybrid model not only can predict stress resistance level, but also can analyze the formation mechanism of stress resistance, providing double basis for breeding strategy formulation.

[0055] The application adopts a model optimization strategy combining transfer learning and distributed collaborative filtering learning, which significantly improves the generalization ability and adaptability of the model. By transferring general knowledge from the normal temperature environment dairy cow breeding model, the dependence of the high-temperature stress resistance model on data volume is reduced. Distributed collaborative training realizes the collaborative use of multi-farm data through sharding processing and feature embedding vector sharing, while protecting data privacy, so that the model can adapt to the climate characteristics and breeding conditions of different regions.

[0056] The application establishes a multi-level breeding regulation system from core group selection to general production group improvement. Through molecular marker monitoring, parent selection optimization and offspring stage evaluation, precise genetic improvement of stress resistance traits is realized. At the same time, combined with feed nutrition regulation and environmental optimization management measures, a breeding strategy of genetic-environmental synergy is formed. This multi-level and multi-angle regulation method ensures the stability and sustainability of the breeding effect.

[0057] The application proposes a zero-base learning mechanism to cope with unknown high temperature stress. Through abnormal phenotype feature extraction, related species conservation mechanism mining and new problem-strategy association model construction, it can quickly respond to new stress resistance problems. Combined with physical rule library and IF-THEN production rule, targeted emergency strategy is generated, and through federated learning, the model is dynamically updated, which significantly improves the adaptability and robustness of the breeding system. Breakthrough the traditional limitations, provide ideas and strategies to cope with new challenges, and update the global model after the effect verification of the implementation of the emergency strategy, enhance the ability of the system to cope with complex and variable environment, and ensure the stable development of the dairy cattle breeding industry in high temperature environment.

[0058] An XGBoost-BN hybrid model is used to build a breeding regulation model, which combines molecular markers and phenotype data to take advantage of different models and make the results more interpretable and practical. Transfer learning and distributed collaborative filtering learning are used to optimize the model, reducing the initialization workload and adapting the model to local data rules to improve generalization ability and overall performance. Based on the optimized model, individualized regulation schemes for different groups are given, and corresponding measures are implemented according to the characteristics of different groups. It can also dynamically adjust the scheme according to real-time data to ensure the sustainability and stability of the regulation effect. In the face of unknown high temperature stress and new stress resistance needs, it has a perfect response mechanism, can classify unknown problems, mine correlations and generate emergency strategies, and supplement new sample data after implementation to enhance the ability to cope with unknown challenges, and realize the continuous optimization and adaptability improvement of the breeding regulation system. BRIEF DESCRIPTION OF DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0060] Figure 1 is a module schematic diagram of the high-temperature weather dairy cow stress resistance breeding regulation method based on intelligent big data analysis provided by the embodiment of the present application;

[0061] Figure 2 is a flowchart of constructing a breeding regulation model according to the stress resistance feature vector. DETAILED DESCRIPTION

[0062] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below according to the drawings in the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the protection scope of the present application.

[0063] Embodiment 1: The following is based on Figures 1-2 The present application discloses a high-temperature weather dairy cow resistance breeding regulation method based on intelligent big data analysis.

[0064] As Figures 1-2 shown, the high-temperature weather dairy cow resistance breeding regulation method based on intelligent big data analysis provided by the embodiments of the present application comprises the following steps.

[0065] S1: Construct a multi-dimensional dairy cow resistance data acquisition system, acquire basic information, phenotype, molecular and environmental stress data of dairy cows under high-temperature environment, and obtain initial dairy cow data.

[0066] S1.1: Determine the data acquisition object and group construction, select different breeds such as Holstein cows and West Holstein crossbred cows, cover healthy dairy cows of different ages and parity to form a test group, select 300-500 individuals for each breed, and at the same time, incorporate a crossbred cow group for milk and meat, so as to ensure that the sample can cover different resistance genetic backgrounds and lay a foundation for subsequent resistance core group cultivation.

[0067] S1.2: Deploy multiple types of data acquisition equipment to realize full-dimensional acquisition of resistance-related data.

[0068] Basic information acquisition: record the unique identification of dairy cows through electronic ear tags, and the breeding site code, according to the pasture management system, input the initial health status and past medical history, and focus on marking the crossbreeding generation of West Holstein crossbred cows and the resistance records of the parents.

[0069] Phenotype data acquisition: real-time acquisition of dairy cow rectal temperature, heart rate, activity, feed intake, milk yield, milk quality, observation of respiratory rate, standing / laying time ratio and other heat stress behavior characteristics through image recognition technology, and recording of production performance changes during the summer high-temperature period, and calculation of milk yield maintenance rate.

[0070] Molecular data collection: Collect blood samples from dairy cows every quarter, use gene chip technology to detect SNP (single nucleotide polymorphism) sites in the genome, and focus on screening heat stress resistance related molecular markers and mastitis resistance SNP markers. Collect liver and mammary gland tissue samples during the peak period of high temperature stress, obtain transcriptome data through RNA sequencing, and extract the expression information of heat shock protein genes and immune related genes.

[0071] Environmental stress data collection: 5-8 environmental monitoring stations are evenly arranged in the pasture to collect air temperature, relative humidity, light intensity and wind speed, and calculate the temperature and humidity index.

[0072]

[0073] In the formula, THI is the temperature and humidity index, is the basic weight coefficient of temperature on heat stress, is the correction coefficient of humidity and temperature interaction, is the basic offset constant, T is the temperature, and RH is the relative humidity.

[0074] Real-time record high temperature stress intensity, THI> 72 is mild stress, THI> 80 is moderate stress, THI> 89 is severe stress, and record special environmental parameters such as duration of extreme high temperature and diurnal temperature difference.

[0075] S13: Establish data transmission and storage mechanism, collect equipment through 4G / 5G wireless network Real-time data transmission to edge computing node, format conversion, redundant data elimination, upload to distributed database, set up off-site double backup to ensure data safety, finally get initial data set containing more than 100,000 records, data set needs to be associated with cow individual identification, collection time, environmental parameters, form individual-environment-traits three-dimensional data association structure.

[0076] S2: Pretreatment and feature engineering of initial data set for stress resistance, build high temperature stress resistance feature library, get standardized stress resistance feature vector.

[0077] S21: Abnormal value and missing value data cleaning of stress resistance data.

[0078] S211: Based on 3σ principle and box plot method to identify abnormal data, such as rectal temperature > 41℃ or < 37℃ under high temperature, milk yield drops by more than 50%, according to breeding log to judge the reason, data caused by equipment failure is directly excluded, and cow sudden disease data is marked as special stress sample.

[0079] S212: Discrete data, specifically disease records, are filled with mode, and continuous data, specifically body temperature and gene expression, are filled with linear interpolation for short-term missing data and K-Nearest Neighbor interpolation for long-term missing data, ensuring the integrity of the stress resistance feature data.

[0080] S22: Standardize the cleaning cow data to eliminate dimensional differences and adapt to stress resistance analysis:

[0081] Phenotypic data (body temperature, milk yield maintenance rate) are standardized using Z-score.

[0082] Molecular data (gene expression, SNP typing data) are converted using logarithmic conversion.

[0083]

[0084] In the formula, x is the gene expression, and y is the converted standardized data value.

[0085] Environmental data are directly standardized using Min-Max normalization to ensure balanced weights of each stress resistance influencing factor.

[0086] S23: Stress resistance feature extraction and selection, construction of high temperature stress resistance feature library:

[0087] S231: Extract indicators that directly reflect stress resistance from preprocessed data, including body temperature stability coefficient under high temperature stress, heat stress resistance gene expression, mastitis resistance SNP marker typing results, milk yield maintenance rate, and milk fat rate change rate.

[0088] S232: Strengthen stress resistance association through feature combination, such as THI x heart rate reflecting the interaction of environmental stress and physiological response, heat stress resistance gene expression x milk yield maintenance rate constructing stress resistance-production synergy feature, and SNP marker combination score.

[0089] S233: Use variance screening to select features significantly related to stress resistance, calculate feature importance based on random forest algorithm, retain the top 80% of important features, eliminate highly correlated features, and construct a high temperature stress resistance feature library.

[0090] S24: Use static weighting method based on feature importance to weight and integrate core features, dynamically assign weights according to high temperature scenarios, and obtain standardized stress resistance feature vector.

[0091] S3: Construct breeding regulation model based on stress resistance feature vector, optimize based on molecular marker assisted selection and distributed learning, and obtain stress resistance core group and optimized model.

[0092] S31: Calculate the individual stress resistance comprehensive score according to the stress resistance feature vector, preliminarily select the stress resistance core data according to the score, and divide the stress resistance core data into a training set, a validation set and a test set by stratified sampling.

[0093] S32: Use an XGBoost-BN hybrid model (extreme gradient boosting-Bayesian network hybrid model) to construct an initial stress resistance breeding regulation model, and fuse molecular markers and phenotype data.

[0094] S321: Data preprocessing and feature alignment, output standardized feature matrix and causal relationship preliminary screening results of the adaptive model.

[0095] According to the divided training set, the molecular marker data and the phenotype data are processed again. For molecular marker data, low-frequency invalid SNP sites are filtered out by using minor allele frequency, and core markers significantly related to stress resistance are retained. For phenotype data, extreme values are removed by box-type Figure Two subcheck to ensure data reliability. Then, the molecular marker data and the phenotype data are aligned according to the unique identification of the individual cow to generate a standardized feature matrix with unified dimensions. At the same time, combined with the knowledge of animal husbandry, 10-15 groups of molecular marker-phenotype potential causal relationships are preliminarily screened out.

[0096] S322: Train the XGBoost (extreme gradient boosting) model, and output the feature importance ranking and stress resistance score prediction model.

[0097] The standardized feature matrix is used as the input of the XGBoost model, the model hyperparameters are set, the stress resistance comprehensive score is used as the prediction target, and the model is trained by 5-fold cross-validation. In each round of cross-validation, 80% of the training set data is used to construct a decision tree ensemble, and 20% of the data is used to verify the prediction error. The tree weight and branch rule are optimized by minimizing the square sum of the stress resistance score prediction error.

[0098] The formula for minimizing the square of the stress resistance score prediction error is:

[0099]

[0100] In the formula, is the true score, is the predicted score.

[0101] After training, the feature importance ranking and the preliminary stress resistance score prediction model are output.

[0102] S323: Optimize the Bayesian network structure based on the XGBoost feature importance, and output the genotype-phenotype-stress resistance causal network.

[0103] Based on the potential causal relationship between molecular markers and phenotypes and the importance of features, the structure of the Bayesian network is optimized. According to the importance of features, the connection between nodes is adjusted. If the importance of molecular marker A is higher than that of phenotype B, and molecular marker A supports the influence of phenotype B, a directed edge A→B is constructed, specifically HSP70 expression→body temperature stability coefficient. If the importance of phenotype C is higher than that of stress resistance score, a directed edge C→stress resistance score is constructed, specifically, such as milk yield maintenance rate→stress resistance score. Using the training set data, the conditional probability table between nodes is calculated by the maximum likelihood estimation method, and finally a complete genotype-phenotype-stress resistance causal network is formed, which can directly show how molecular markers affect stress resistance through phenotypes.

[0104] S324: Fuse the XGBoost prediction result with the Bayesian network causal reasoning to output the initial stress resistance breeding regulation model.

[0105] The XGBoost stress resistance score prediction model and the Bayesian network causal network are fused: on the one hand, the XGBoost model is used to predict the stress resistance score of the input standardized feature vector, and the current stress resistance level of the individual is obtained. On the other hand, based on the Bayesian network causal reasoning, combined with the feature importance output by XGBoost, the key influencing factors of individual stress resistance are analyzed, and targeted breeding strategies are generated: for parental selection, individuals carrying superior genotype combinations are preferentially recommended. For molecular marker selection, 3-5 core SNP sites that need to be monitored are marked. The stress resistance score prediction and breeding strategy recommendation modules are packaged and integrated to form the initial stress resistance breeding regulation model, which provides a basic framework for subsequent distributed collaborative filtering learning optimization.

[0106] S33: Use the transfer learning and distributed collaborative filtering learning optimization model to adapt to the stress resistance breeding scene, and obtain the stress resistance core group and the optimization model.

[0107] S331: Select the XGBoost-BN model for normal temperature environment dairy cow breeding as the transfer source. This model is trained based on 500,000 normal temperature dairy cow data and has mature phenotype data processing logic and general molecular marker analysis capability. The phenotype data, general molecular marker processing related XGBoost decision tree structure, and Bayesian network prior probability associated with general molecular markers such as SNP of basic immune genes are extracted from the transfer source and transferred to the initial stress resistance breeding regulation model. Only the high-temperature stress related module parameters are reinitialized, and the XGBoost-BN model to be optimized initialized by transfer learning is output.

[0108] S332: The master node performs sharding processing based on the XGBoost-BN model to be optimized, assigns exclusive model shards to each pasture edge node, and obtains the sharding model module and node allocation scheme.

[0109] The master node is based on the XGBoost-BN model to be optimized, and is processed by module function + feature type two-dimensional slicing. For the XGBoost module, it is divided into three slices according to the feature attribute: molecular marker group, phenotype group, and environment group. For the Bayesian network module, it is divided into two core causal chain slices according to the causal logic: genotype→ phenotype and phenotype→ stress resistance. For the PLS-DA (partial least squares discriminant analysis) module integrated in the model, it is divided into different dimension slices according to the principal component interpretation degree. Combined with the local data types of each pasture edge node, each node is assigned a dedicated model slice to ensure that each node only processes modules matching the local data, avoiding resource consumption of full model transmission, and finally outputting the sliced model modules and the corresponding node allocation scheme.

[0110] S333: Local collaborative training optimizes slicing parameters, and outputs slicing parameter update and feature embedding vector.

[0111] After each pasture edge node obtains the dedicated model slice according to the node allocation scheme, local training is carried out combined with local stress resistance data. For XGBoost slicing, the slicing gradient descent algorithm is used to calculate only the decision tree branch gradient corresponding to the local feature group, and the stress resistance score prediction error sum of squares is used as the loss function to iteratively optimize the tree weight and splitting rule. For Bayesian network slicing, the local expectation maximization algorithm is used to update only the conditional probability table corresponding to the local causal chain to ensure that the causal relationship adapts to the local data law. After each round of training, the nodes do not transmit raw data, but share the local feature mapping into low-dimensional feature embedding vectors — for example, A pasture shares the molecular marker feature embedding vector to assist B pasture in optimizing the analysis of the molecular marker-environment interaction effect of the environment group slice. After 5-10 iterations, the node-optimized slicing parameter update and the matching feature embedding vector are output.

[0112] S334: Global slicing parameter aggregation and reorganization, output preliminary optimized stress resistance breeding regulation model.

[0113] The master node first receives the node slicing parameter update output by S333, and starts the slicing consistency verification process: for the same type of slicing of different nodes, the consistency is judged by calculating the parameter cosine similarity, if the similarity < 0.8, it means that the slicing optimization has deviation, and the corresponding node is returned to execute the local training again. After verification, the global parameter is calculated using the slicing contribution degree weighted aggregation strategy: taking the slicing data volume proportion and the slicing prediction accuracy as the dual weight.

[0114]

[0115] wherein, is the weight of the kth slice, is the number of slicing samples, M is the total number of the same type of fragments, is the weight coefficient of the proportion of the sample size when calculating the fragment weight, is the weight coefficient of the fragment accuracy rate when calculating the fragment weight.

[0116] The weighted sum obtains the global parameter update amount of each fragment. The global parameters of all fragments are reorganized according to the original model structure, the integrity of the XGBoost-BN model is restored, and the preliminary optimized resistance breeding regulation model is output.

[0117] S4: Obtain individualized resistance breeding regulation scheme based on the optimized model, implement and verify the effect in groups.

[0118] S41: Use the output of the optimization model to carry out fine breeding regulation of the core group, and output the stage selection results of the core group.

[0119] Molecular marker monitoring: every half year, the core group individuals are rechecked for resistance SNP markers, the change of gene frequency is tracked, and the drift of dominant genotype is avoided.

[0120] Parental selection optimization: based on the model obtained kinship coefficient and resistance prediction value, two selection modes of high resistance x high resistance and high resistance x medium resistance are developed, and the proportion of high resistance x high resistance combination is not less than the preset threshold, to ensure the genetic gain of resistance of offspring.

[0121] Offspring selection: the offspring of the core group are subjected to stage resistance evaluation, 6-month-old individuals are screened by molecular markers, 12-month-old individuals are screened according to summer heat stress phenotype, and finally individuals with resistance score ≥75 points are reserved into the core group reserve team.

[0122] S42: Implement improvement regulation of ordinary production group according to the stage selection results of core group, and output the improvement regulation implementation record.

[0123] Hybrid improvement: for low resistance Holstein cow group, introduce West Holstein hybrid frozen semen for hybrid improvement, the resistance score of the first generation hybrid offspring needs to be improved to more than 65 points, and the second generation hybrid offspring needs to reach more than 70 points.

[0124] Feed nutrition synergistic regulation: for ordinary production group, the heat stress resistant feed formula recommended by the model is used, and TMR (total mixed ration) is used for accurate feeding, to improve the physiological state in high temperature period and provide environmental support for resistance gene expression.

[0125] Environmental optimization: install temperature and humidity sensors in the cowshed to automatically control the fan and spraying system, control the temperature in the cowshed below 28℃, and reduce the interference of environmental stress on breeding effect.

[0126] S43: Establish a multi-dimensional evaluation system based on the records of improved regulation implementation to verify the effectiveness of regulation schemes and output verification result data.

[0127] Short-term validation: Monitor the physiological and production indicators of the core group and the general production group. The incidence of clinical mastitis in the core group needs to be reduced to below 2%.

[0128] Long-term validation: The stress resistance score and stress resistance gene frequency of the offspring of the core population were detected. The milk yield of the hybrid offspring of the ordinary production population increased by ≥5% compared with the parents.

[0129] Molecular-level validation: The expression levels of heat stress resistance genes in the core population were detected by RT-qPCR (reverse transcription and real-time quantitative polymerase chain reaction), which were ≥20% higher than those in the general population, and the expression levels of mastitis resistance genes were adapted to physiological needs.

[0130] S44: Dynamically adjust the control scheme to adapt to changes in population resistance: Collect real-time data of the core population and ordinary production population every month, re-input into the optimization model, and if the change in resistance score is greater than 10 points, adjust the parent selection ratio or hybridization improvement strategy.

[0131] S5: To address unknown high-temperature stress and new resistance requirements, we initiated a zero-foundation learning and emergency breeding strategy.

[0132] S51: Based on real-time data and multi-condition judgment, output the classification results of unknown resilience problems.

[0133] Real-time stress resistance characteristics, current phenotypic status, and environmental monitoring data of dairy cows were collected. When the cosine similarity between the real-time stress resistance characteristics of dairy cows and all samples in the model training set was found to be less than a preset threshold, and abnormal phenotypes were observed, while environmental monitoring revealed novel stress factors, these criteria were used to identify an unknown stress resistance problem. Subsequently, this unknown problem was classified using an expert system integrating knowledge from animal husbandry and molecular biology experts, outputting classification results into physiological stress type, molecular regulation type, and environmental interaction type, providing a basis for subsequent steps.

[0134] S52: Using the classification results output in the previous step, we can mine associations through zero-based learning and output a new problem-strategy association model.

[0135] S521: Once an unknown stress resistance problem is identified and corresponding classification results are obtained, representative and key feature information is extracted from the abnormal phenotypes that appear, such as neurological symptoms exhibited by dairy cows in extreme environments, like convulsions and behavioral abnormalities, or changes in blood biochemical indicators, such as abnormal electrolyte concentrations and altered metabolite levels. These features are then normalized to fit them into the known stress resistance feature space.

[0136] S522: After mapping the abnormal phenotype features to the known feature space, a meta-learning algorithm is used to find the correlation between them. The meta-learning algorithm has the characteristics of being able to quickly learn and adapt to new tasks, and here it can analyze the internal relationship between the current abnormal phenotype features and those known features based on a large amount of existing data and knowledge patterns, quickly correlate the feature chain similar to neurological symptoms-oxidative stress-ion balance, etc. Through algorithm discovery, it is found that the occurrence of neurological symptoms may be related to changes in oxidative stress response in the body, and oxidative stress further affects the ion balance state, thereby revealing the underlying physiological and biochemical reaction chain behind the abnormal phenotype, providing a basis for subsequent understanding of the nature of the problem and developing strategies.

[0137] S523: Extract the conserved stress resistance mechanisms of related species.

[0138] Considering that yellow cattle, buffaloes, and dairy cows belong to related species, they may have similar stress resistance mechanisms when dealing with environmental stress such as high temperature. Collect research results on high-temperature stress resistance of these related species, which include physiological, molecular, and other aspects. From numerous research results, extract relatively conserved stress resistance mechanisms, that is, mechanisms that exist universally and function in different related species. Antioxidant enzyme system regulation can regulate the activity of antioxidant enzymes in the body, eliminate excess free radicals, and maintain the redox balance in cells. These conserved mechanisms reflect some inherent physiological and molecular regulation methods that ruminants may share when dealing with high-temperature stress.

[0139] S524: After associating the abnormal phenotype with the known feature chain and extracting the conserved stress resistance mechanisms of related species, combine this information with the existing dairy cow stress resistance molecular marker database. The dairy cow stress resistance molecular marker database records the association between specific molecular markers and dairy cow stress resistance performance.

[0140] Take the association of abnormal phenotype and known feature chain as the input layer of the model, and take the association features such as neurological symptoms-oxidative stress-ion balance as input parameters. Take the extracted conserved stress resistance mechanisms of related species as the middle layer, which establishes a connection between the input features and potential regulation targets. Finally, combine the dairy cow molecular marker database and calculate the potential effect score of different coping strategies for solving the current new problem based on the association data in the database, thereby generating a series of association rules of abnormal phenotype-conserved mechanism-recommended strategy, and finally constructing a new problem-strategy association model.

[0141] S53: Based on the new problem-strategy association model, construct a physical rule library, integrate high-temperature physiological metabolism rules, molecular regulation rules, and breeding emergency rules. According to different types of problems, generate corresponding emergency strategies:

[0142] For physiological stress type problems, generate emergency cooling + drug intervention to relieve symptoms, and screen for strong resistance core group of individuals to freeze sperm, and carry out emergency mating in advance to retain the emergency strategy of anti-stress genes.

[0143] For molecular regulation type problems, develop an emergency strategy using gene activators to activate silenced anti-stress genes, and monitor gene expression changes based on transcriptome sequencing to provide data for subsequent molecular marker updates.

[0144] For environmental interaction type problems, generate an emergency strategy to modify the environment of the cowshed, introduce short-term extreme environment resistant breed frozen sperm crossbreeding, and quickly improve the resistance of the group.

[0145] These emergency strategies generated for different types of problems are output as a whole to prepare for subsequent strategy verification.

[0146] After implementing the emergency strategy, closely observe the changes in relevant indicators within 48 hours. If abnormal phenotypes are found to be alleviated and the resistance score is improved by ≥20 points, the emergency strategy is determined to be effective. Subsequently, the characteristic data of this unknown problem, the emergency strategy used, and the effect produced are recorded as new samples, which are supplemented to the training set. Based on the newly expanded training set, the updated global model is updated to have the ability to handle such new types of resistance problems. Finally, the updated global model is output to enhance the ability of the entire breeding control system to cope with unknown challenges.

[0147] Embodiment 2: Dairy cow resistance core group cultivation, mastitis resistance mechanism and comprehensive prevention and control technology Research on the specific implementation process of the intelligent big data analysis-based high-temperature weather dairy cow resistance breeding control method is as follows:

[0148] 1. 27 new resistance molecular markers were screened, and 400 resistance core groups were established. The transfer and production performance identification of West Holstein crossbred cows were completed, and the screening of dairy cow resistance molecular markers was successfully completed through high-throughput sequencing technology combined with whole genome association analysis technology. 10058 resistance markers were obtained, and 27 significantly different SNPs were annotated, of which 3 SNPs were directly related to immune response. The screening of resistance molecular markers was initially completed, and at present, blood samples and milk samples have been collected on a large scale in Si Hong Aide Ranch, and West Holstein crossbred cows and other Holstein cows are being verified for resistance molecular markers. A core group of 400 dairy cows with resistance markers was established.

[0149] 2. Establish standards for quality and safety control technology of fresh milk quality key points and meat quality control standards. The residual aflatoxin M1 in fresh milk shows seasonal distribution, which should be increased in the winter dairy ration.

[0150] The dosage of mycotoxin adsorbent. Mycotoxin adsorbents have no adverse effects on dairy cow production performance, and the use of compound mycotoxin adsorbents can reduce mycotoxin residues in dairy cows and milk. Silica milk liners can improve factors including milk yield, average milk flow rate, SCC (somatic cell count), and teat tip score. Silica milk liners do not negatively affect the total microbial count in raw milk or the incidence of mastitis. Fully enclosed cowsheds are less affected by external temperature fluctuations, maintaining a stable environment and minimizing the impact on dairy cow milk production performance.

[0151] Muscle and fat samples were collected from West Dutch crossbred cattle and Holstein fattening cattle. Meat quality analysis revealed that West Dutch crossbred cattle were significantly superior to Holstein cattle in terms of tenderness, meat color, and intramuscular fat content. Mature bovine intramuscular adipocytes were cultured in vitro, and changes in fat content during lipolysis were detected using Oil Red staining and extraction methods. The release of glycerol and free fatty acids from the cells was measured using colorimetric methods. The expression of key lipolysis genes PPARα (peroxisome proliferator-activated receptor α), LPL (lipoprotein lipase), and HSL mRNA (hormone-sensitive lipase messenger RNA) and protein was analyzed using RT-qPCR (reverse transcription combined with real-time quantitative polymerase chain reaction) and Western blot.

[0152] 3. Screening Study for Resistance Markers in Bovine Mastitis: Blood samples, milk samples, and milk samples from cows with clinical mastitis were collected from demonstration farms. 16S rDNA (16S ribosomal DNA) gene sequencing analysis (complying with the "Regulations on the Management of Human Genetic Resources of the People's Republic of China" and related laws) was used to identify the types of pathogens in the milk of cows with clinical mastitis and those with high somatic cell counts. It was found that the TNF (tumor necrosis factor) level in the blood of cows with clinical mastitis was significantly higher than that of healthy cows. Based on this, drug sensitivity tests were conducted on these pathogens against six commonly used antibiotics (ampicillin, ciprofloxacin, lincomycin, gentamicin, streptomycin, and penicillin) (the use of antibiotics complied with the "Technical Guidelines for Clinical Trials of Antimicrobial Drugs" and related laws) (the drug sensitivity tests complied with the "Law of the People's Republic of China on the Prevention and Control of Infectious Diseases" and related laws). The results were fed back to guide drug use in production, reducing the incidence of clinical mastitis in milk by 41.22%. Resistance genes of isolated Klebsiella pneumoniae were also detected. A mastitis pathogen bank was established to prepare for future trials of specific mastitis vaccines and further research.

[0153] Combined with the results of pathogen identification and specific pathogen challenge test in vivo (challenge test in vivo conforms to the provisions of the "Pathogenic Microorganism Laboratory Biosafety Management Regulations" and related laws), 10 candidate genes for bovine mastitis resistance were screened, including IL8 (interleukin 8), CXCR1 (CXC chemokine receptor 1), LF (lactoferrin), TLR1 (Toll-like receptor 1), TLR4 (Toll-like receptor 4), IL17A (interleukin 17A), ILF3 (interleukin enhancer binding factor 3), and CDH13 (cadherin 13). A total of 45 SNPs (single nucleotide polymorphisms) were identified using PCR (polymerase chain reaction) sequencing and time-of-flight mass spectrometry, and linkage analysis and genotype analysis were performed. On this basis, the production traits of test cattle from 2017 to 2019, such as lactation performance and clinical mastitis records, were collected. Multifactor analysis, Logstic (logistic regression) analysis, and multi-gene dimensionality reduction methods were used to analyze the significant SNP sites affecting bovine mastitis resistance, and cell experiments were used to verify the results in vitro. According to the significant SNP sites affecting mastitis resistance screened in the early stage, pedigree, phenotype values, and genetic values were combined in production to implement gene marker-assisted selection and matching.

[0154] 4. Research and application of antibacterial peptide high-efficiency expression plasmid in the treatment of bovine mastitis The recombinant plasmid was transfected into primary mammary epithelial cells, and after infection with Staphylococcus aureus, the cell survival rate was significantly higher than that of the control group without transfection of the recombinant plasmid, and the cell death rate was also significantly lower than that of the control group. Transfection of the above two recombinant plasmids into mice at the peak of lactation showed that the expression levels of genes TAP (antigen processing and transport body) and Lyz (lysozyme) in the mammary glands of some mice increased significantly, while no expression of target genes was detected in the hearts, livers, spleens, lungs, and kidneys of mice. Combined with various physiological and biochemical indicators of mice, it was proved that the recombinant plasmid had no significant effect on the behavior, weight changes, blood biochemical indicators, and organ indices of mice, proving the safety of the constructed recombinant plasmid.

[0155] 5. Development of a film-forming teat dip for dairy cows (the use of the teat dip conforms to the provisions of the "Veterinary Biological Products Management Regulations" and related laws). A new film-forming teat dip for dairy cows was developed based on the optimization of chitosan and povidone-iodine as the main disinfecting components, and two clinical experiments were conducted in demonstration bases (clinical trials conform to the provisions of the "Biomedical New Technology Clinical Research and Clinical Transformation Application Management Regulations" and related laws). The results of the milk area comparison experiment and the cow group comparison experiment showed that the new film-forming teat dip developed with chitosan and povidone-iodine as the main disinfecting components had obvious effects on the prevention and treatment of subclinical mastitis in dairy cows and had no significant impact on milk quality. It is a simple, green, safe, and effective teat dip.

[0156] In summary, the embodiment provides a high-temperature weather dairy cow resistance breeding regulation method based on intelligent big data analysis, constructs a breeding regulation model by adopting an XGBoost-BN hybrid model, fuses molecular markers and phenotype data, and optimizes the model by using transfer learning and distributed collaborative filtering learning, and the beneficial effects are as follows:

[0157] Based on the optimized model, individualized resistance breeding regulation schemes for the core group and the ordinary production group are given, fine molecular marker monitoring, parent matching optimization and offspring breeding, etc. are performed on the core group, and hybrid improvement, feed nutrition synergistic regulation and environmental optimization, etc. are adopted for the ordinary production group, so that targeted regulation can be implemented according to the characteristics and needs of different groups, and the resistance and production performance of the group are effectively improved. A mechanism for dynamically adjusting the regulation scheme is established, the resistance change of the group is monitored according to real-time data, and the parent matching ratio or hybrid improvement strategy is adjusted in time, so that the regulation scheme can flexibly adapt to the dynamic changes of the resistance of the group, and the sustainability and stability of the breeding regulation effect are ensured.

[0158] For unknown high-temperature stress and new resistance needs, a perfect response mechanism is provided, unknown resistance problem classification results are output by multi-condition judgment, a new problem-strategy association model is constructed by using zero-base learning to mine correlations, and finally corresponding emergency strategies are generated. This layer-by-layer progressive approach can effectively deal with new complex resistance problems, and enhance the ability of the entire breeding regulation system to respond to unknown challenges. After implementing the emergency strategy, new sample data can be supplemented to the training set, so that the model can continuously learn new knowledge and gradually improve the processing ability of various resistance problems, realizing the continuous optimization and adaptability improvement of the entire breeding regulation system.

[0159] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions essentially or say the part that makes a contribution to the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of the embodiments or some parts of the embodiments.

[0160] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit it. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced equivalently. The modification or replacement does not make the essence of the corresponding technical solution deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A high-temperature weather dairy cow stress resistance breeding regulation method based on intelligent big data analysis, characterized in that, The method comprises the following steps: S1: Construct a multi-dimensional dairy cow resistance data collection system, collect basic information, phenotype, molecular and environmental stress data of dairy cows in high temperature environment, and obtain initial dairy cow data; S2: Preprocess and feature engineering of the initial dairy cow data for resistance orientation, construct a high-temperature resistance feature library, and obtain a standardized resistance feature vector; S3: Construct a breeding control model according to the resistance feature vector, optimize the breeding control model according to molecular markers and distributed learning, and obtain an optimized resistance breeding control model; S31: Calculate the individual resistance comprehensive score according to the resistance feature vector, preliminarily select the resistance core data according to the score, and divide the resistance core data into a training set, a validation set and a test set by stratified sampling; S32: An initial resistance breeding control model is constructed using an XGBoost-BN hybrid model, and molecular markers and phenotype data are fused; S321: Standardize the molecular marker data and phenotype data of the training set data, output a standardized feature matrix, and preliminarily screen the molecular marker-phenotype potential causal relationship to obtain a molecular marker causal relationship list; S322: The standardized feature matrix is input into the XGBoost module for training, the resistance comprehensive score is taken as the prediction target, the model hyperparameters are optimized through cross-validation, a resistance score prediction model is obtained, the feature importance ranking is calculated and output based on the decision tree node split gain; S323: The molecular marker causal relationship list and the feature importance ranking are adjusted to adjust the directed edges between nodes, and the conditional probability table of each node is calculated through the training set data to optimize the structure of the Bayesian network, and a resistance causal network is obtained; S324: The resistance score prediction model and the resistance causal network are fused to output an initial resistance breeding control model; S33: The model is optimized by transfer learning and distributed collaborative filtering learning, which is adapted to the resistance breeding scene, and an optimized resistance breeding control model is obtained; S4: According to the optimized resistance breeding control model, breeding control is carried out to obtain the stage selection results of the core group, the control scheme is dynamically adjusted according to the selection results, and a personalized resistance breeding control scheme is obtained; S5: In response to unknown high-temperature stress and new resistance needs, zero-base learning and emergency breeding strategies are started based on the personalized resistance breeding control scheme; S51: Collect real-time resistance features, phenotypes and environmental data of dairy cows, and when the cosine similarity is less than a preset threshold, determine that it is an unknown resistance problem, classify the unknown resistance problem, and obtain an unknown problem classification; S52: According to the unknown problem classification, an association model of new problem-strategy is constructed by using a meta-learning algorithm to mine the association between abnormal phenotypes and known feature chains; S521: Extract abnormal phenotype features in the unknown problem classification; S522: Map the abnormal phenotype features to the known feature space, and use a meta-learning algorithm to find the association between abnormal phenotypes and known feature chains; S523: Extract the conserved resistance mechanism of a close relative species; S524: Combine the association of the abnormal phenotype with the known feature chain with the conserved stress resistance mechanism of the close relative species to construct a new problem-strategy association model; S53: Construct a physical rule library according to the new problem-strategy association model, integrate high-temperature physiological metabolic rules, molecular regulation rules, and breeding emergency rules, and generate emergency strategies according to different types of problems.

2. The high temperature weathered cow resistance breeding regulation method based on intelligent big data analysis according to claim 1, characterized in that, In step S1, the specific steps for obtaining the initial dairy cow data are as follows: S11: Determine the multi-dimensional dairy cow stress resistance data object; S12: Deploy multi-type data acquisition equipment; S13: Establish a data transmission and storage mechanism, collect real-time data from the acquisition equipment to the edge computing node, perform format conversion and redundant data elimination, and then upload to the distributed database. Set up off-site double backup to ensure data security, and obtain the initial dairy cow data.

3. The high temperature weathered cow resistance breeding regulation method based on intelligent big data analysis according to claim 1, characterized in that, In step S2, the specific steps for obtaining the standardized stress resistance feature vector are as follows: S21: Perform abnormal value and missing value data cleaning on the stress resistance data to obtain cleaned dairy cow data; S22: Eliminate the dimension difference of the cleaned dairy cow data to obtain standard dairy cow data; S23: Extract stress resistance features from the standard dairy cow data to construct a high-temperature stress resistance feature library; S24: Use a static weighting method based on feature importance to weight and fuse the core features, dynamically assign weights according to the high-temperature scenario, and obtain a standardized stress resistance feature vector.

4. The high temperature weathered cow resistance breeding regulation method based on intelligent big data analysis according to claim 3, characterized in that, In step S23, the specific steps for constructing the high-temperature stress resistance feature library are as follows: S231: Extract indicators that directly reflect stress resistance from the standard dairy cow data; S232: Strengthen stress resistance association through feature combination according to stress resistance indicators; S233: Use variance screening to select relevant features, calculate feature importance based on decision tree algorithm, and retain features with importance greater than a preset threshold to construct a high-temperature stress resistance feature library.

5. The high temperature weathered cow stress tolerance breeding regulation method based on intelligent big data analysis according to claim 1, characterized in that, In step S33, the specific steps for optimizing the model using transfer learning and distributed collaborative filtering learning are as follows: S331: Use the XGBoost-BN model trained with normal temperature data as the transfer source, migrate the phenotype and general molecular marker processing data to the initial model, and output the optimized XGBoost-BN model; S332: The master node performs sharding based on the optimized XGBoost-BN model, assigns a dedicated model shard to each pasture edge node, and obtains a shard model module and node allocation scheme; S333: Each pasture edge node trains the optimized XGBoost-BN model based on the shard model module and node allocation scheme, and outputs the shard parameter update amount and feature embedding vector; S334: Perform shard consistency verification on the shard parameter update amount and feature embedding vector, aggregate global parameters according to double weights, reorganize into a complete model, and output the optimized stress resistance breeding regulation model.

6. The high temperature weathered cow resilience breeding regulation method based on intelligent big data analysis according to claim 1, characterized in that, In step S4, the specific steps for outputting the adjusted regulation scheme are as follows: S41: Fine-tune breeding regulation based on the optimized stress resistance breeding regulation model, and output the core group stage selection result; S42: Implement general production group improvement regulation based on the core group stage selection result, and output the improvement regulation implementation record; S43: Based on the improved regulation implementation record, a multi-dimensional evaluation system is established to verify the effect of the regulation scheme, and the verification result data is output; S44: Dynamically adjust the regulation scheme according to the verification result data, and output the adjusted regulation scheme.

Images