Breeding-pork pig dual-link closed-loop based superior line of breeding pig matching screening method

Abstract

The application relates to the technical field of pig breeding management, in particular to a boar dominant strain matching screening method based on a boar-meat pig double-link closed loop. In the method, whole-process production data in the conversion process from boars to commodity meat pigs is acquired, a father strain and a mother strain are coded and identified, and a multi-label sample atlas of the father and mother strains is constructed; reproductive performance data and commodity meat pig production and operation data are associated and mapped, and a boar-meat pig double-link mapping index library is established; the growth contribution of the father, the reproduction contribution of the mother and the combined synergistic gain are analyzed through a comprehensive advantage model, an advantage matching matrix and a comprehensive advantage index are generated; the evaluation weight is dynamically adjusted according to the regional market demand, and is feedback corrected by using historical settlement results, so that the advantage strain matching closed loop optimization is realized. The application effectively improves the precision of boar strain screening and matching combination evaluation, and realizes the synergistic improvement of reproductive advantage and commodity meat pig cultivation benefit.

Description

Technical Field

[0001] This invention relates to the field of breeding pig breeding and management technology, and in particular to a method for screening superior breeding pig breeds based on a closed-loop dual-linkage of breeding and meat pigs. Background Technology

[0002] As the pig farming industry develops towards large-scale, precise, and intelligent operations, genetic improvement of breeding pigs and the selection of superior breeds have become crucial technical aspects for enhancing the production efficiency and economic benefits of pig farming enterprises. As the key genetic basis influencing the production performance of commercial pigs, the combination of paternal and maternal lines directly determines the reproductive performance, growth rate, feed utilization efficiency, and final sales revenue of offspring. In modern breeding systems, a comprehensive analysis of individual breeding pig reproductive performance data, genetic pedigree data, and commercial pig production and management data is typically used to select parent combinations with high matching value. The paternal line usually has a significant impact on the growth performance, carcass quality, and economic traits of commercial pigs, while the maternal line mainly affects reproductive efficiency, farrowing performance, and nursing capacity. Therefore, constructing a breed matching evaluation system that can simultaneously reflect the reproductive advantages of breeding pigs and the economic benefits of commercial pigs is of great significance for optimizing the allocation of breeding pig resources, improving the efficiency of genetic improvement, and reducing breeding decision-making costs.

[0003] Currently, the selection and matching of breeding pig breeds generally adopts a phased statistical analysis model. In the breeding stage, breeders mainly evaluate maternal performance based on reproductive indicators such as sow litter size, number of healthy piglets, and mating success rate. In the commercial pig production stage, the production effect of commercial pigs is evaluated based on indicators such as daily weight gain, feed conversion ratio, survival rate, slaughter performance, and sales settlement results. Data from different stages are usually stored independently by different management systems or manual recording methods, and then correlated and analyzed manually. The existing technical model has at least the following technical defects: First, there is a lack of effective data linkage between the genetic source data of breeding pigs and the production and operation data of commercial pigs, resulting in a disconnect between reproductive indicators and the profitability indicators of commercial pigs. It is difficult to accurately track the contribution of paternal and maternal lines to the performance of offspring commercial pigs, making it impossible to effectively assess the value of superior genetic resources. Second, existing evaluation methods usually use a single reproductive performance or a single production performance indicator as the screening basis, without comprehensively considering the impact of paternal lines on the growth performance of offspring and the contribution of maternal lines to reproductive efficiency, which can easily lead to biases in judging breed advantages. Third, different sales regions have differences in the body size, quality, production cycle, and market price of commercial pigs. The existing fixed matching schemes cannot dynamically adjust the breed combination according to changes in market demand, resulting in insufficient matching between breeding pig production targets and market demand. Fourth, after the breeding pig matching scheme is determined, there is a lack of feedback verification mechanism based on the actual settlement benefits of commercial pigs. It is impossible to use subsequent production results to correct historical matching models, resulting in insufficient accuracy in the evaluation of matching coefficients and difficulty in achieving continuous optimization of the selection of superior breeding pig breeds. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides a method for screening superior breeding pig breeds based on a closed-loop dual-linkage of breeding and meat pigs.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides a method for screening superior breeding pig lines based on a closed-loop breeding-meat pig dual-link system, comprising the following steps: S01: Obtain full-process production data from breeding pigs to commercial meat pigs, and assign unique strain codes to paternal and maternal strain samples of different breeding pig breeds based on the breeding source data and genetic information recorded in the full-process production data, and establish a multi-label sample map of the parent strains of individual breeding pigs. S02: Extract reproductive performance data of individual breeding pigs and production and operation data of individual commercial pigs from the parent-child combination throughout the entire process. Based on the multi-label sample map of the parent breed, and according to ear number, batch traceability information and genetic lineage, perform correlation mapping between production and operation data and reproductive performance data to construct a dual-link mapping index library for the entire life cycle of breeding pig breeds. S03: Construct a comprehensive advantage model for the synergistic analysis of the significant impact of different paternal lines on the growth performance of commercial meat pigs and the significant impact of different maternal lines on the reproductive performance. Then, solve the matching gain of the breeding prior of the paternal line on the mating effect driving weight of the settlement benefit of commercial meat pigs through the comprehensive advantage model, and generate different advantage matching matrices of paternal and maternal lines and their comprehensive advantage index. S04: By calculating the fit between the corresponding market demand in different sales regions and different advantageous matching matrices, the evaluation weights of each index of the comprehensive advantage index are dynamically adjusted according to the meat pig duration assessment index of the demand vector. A comprehensive advantage index ranking table is generated to recommend the advantageous matching matrix with high fit, and the optimal breeding pig breed matching results for different sales regions are obtained. S05: Use historical settlement data of commercial pigs to reverse verify the accuracy of the prediction of the matching coefficients of the breeding pig breeds during actual breeding and make corrections, so as to realize the closed-loop verification of the matching of breeding pig breeds.

[0006] Furthermore, S01 specifically includes the following steps: By querying production and breeding logs, the entire process of breeding and converting breeding pigs into commercial pigs is retrieved from the breeding farm, commercial farm, and fattening farm within the preset breeding time period. Based on the breeding and circulation information recorded in the entire process production data, an identity file is created for each individual breeding pig. By extracting paternal and maternal strain samples from various individual pigs' production and breeding data, as well as the breeding source data corresponding to each strain sample, the reproductive capacity of all paternal and maternal strain samples is aggregated and analyzed based on the breeding source data to extract strain characteristics and obtain a series of parent strain feature vectors. Genetic marker sequencing data and genetic loci of individual pigs corresponding to each paternal and maternal strain sample were obtained by retrieving from the breeding gene bank. Based on the differences in genetic loci, the Euclidean genetic distance between each pair of paternal and maternal strain samples was calculated to obtain the paternal genetic distance matrix and the maternal genetic distance matrix. Define the breed category of pigs as a genetic community unit. Based on the paternal genetic distance matrix or maternal genetic distance matrix, calculate the category genetic cost of the central clustering of the feature vectors of each parent breed within different genetic community units, and obtain the increase in genetic variance. Based on the breeding advantages of breeding pigs and the requirements for matching the breeds, a minimum genetic cost threshold is preset. If the increase in genetic variance is less than the minimum genetic cost threshold, the feature vectors of the two parent breeds are merged by category genetics, the corresponding genetic distance matrix is ​​updated, and finally the breed origin clustering labels of different breeding pig individuals are generated. Referring to the identification files, the breed origin clustering tags are used to perform unique or non-unique breed coding operations on each parent breed sample to establish a multi-label sample map of the parent breeds of individual breeding pigs.

[0007] Furthermore, S02 specifically includes the following steps: By acquiring reproductive performance data of various pig individuals corresponding to parent combinations and production and operation data of corresponding commercial pigs through full-process production data, and establishing genetic phylogenetic charts of different commercial pig individuals based on biological kinship clues registered in the mating and birth information; A multi-layered directed hypergraph framework for constructing a high-dimensional information grid is constructed by introducing a deep learning network. The multi-layered directed hypergraph framework is hierarchically divided according to the indicator dimension of reproductive performance data to form the first directed heterogeneous layer. At the same time, the multi-layered directed hypergraph framework is divided according to the indicator dimension of production and operation data to form the second directed heterogeneous layer. Define individual commercial pigs as a type of performance link node, and embed the kinship direction of the genetic kinship pedigree diagram into the first directed heterogeneous layer as the genetic lineage edge. Then, attach production and operation data as the settlement attribute of a type of performance link node to obtain the strain settlement loading channel diagram of individual commercial pigs. Define individual breeding pigs as nodes in the second-class performance link, and embed the sample index of the multi-label sample map of the parent breed as the breed membership edge into the second directed heterogeneous layer. Add reproductive performance data as the reproductive attribute of the second-class performance link node to obtain the breed reproductive loading channel map of individual breeding pigs. Obtain electronic ear tag information and birth batch information for the management of the breeding pig breeding process. Combine the electronic ear tag information and birth batch information as the root base point. Starting from the root base point, perform a reverse search traversal of the strain settlement loading channel graph along the parent relationship edge in the multi-layer directed hypergraph framework, and output the strain grandparent node. A strain reference matrix is ​​constructed based on the genetic network of the multi-label sample map of the parent strains. Based on the genetic constraints of the strain reference matrix, the strain settlement loading channel graph is decomposed into a low-rank decomposition of the strain genetic origin to generate the strain contribution index of the direct parent corresponding to the ancestral node of the strain. Based on the breed contribution index, the settlement indicator data of individual commercial pigs are mapped to the reproductive indicator database of the parent breeds of breeding pigs, thus constructing a dual-link mapping indicator library of the breeding-meat pigs throughout the entire life cycle of breeding pig breeds.

[0008] Furthermore, S03 specifically includes the following steps: The growth trait data of each commercial pig generated by different paternal line samples within a preset breeding period were extracted using the breeding-meat pig dual-link mapping index library, and the reproductive performance data of each commercial pig generated by different maternal line samples within a preset breeding period were also extracted. To investigate the significant impact of different paternal lines on the growth performance of commercial pigs, the significant impact of different maternal lines on the reproductive performance of commercial pigs, and the combined superiority of different parental lines, growth trait data and reproductive performance data were analyzed and calculated separately to obtain the paternal genetic growth contribution value, the maternal reproductive contribution value, and the combined synergistic gain value of different parental lines for each paternal line sample. A comprehensive advantage model is established based on the paternal genetic growth contribution value, maternal reproductive contribution value, and combined synergistic gain value. This comprehensive advantage model includes a reproductive advantage index, a meat pig benefit index, a market value index, and a stability index. Historical breeding data of commercial meat pigs produced from individual breeding pigs corresponding to each paternal line sample are obtained from the breeding database within a predetermined time period. The historical breeding data is used to evaluate each paternal line sample and assign a prior effect weight to each sample. By analyzing settlement data, we obtain actual offspring settlement performance data of commercial pigs bred based on samples of each paternal line. We introduce a Bayesian inference network and use the likelihood operators of different pig observation dimensions contained in the actual offspring settlement performance data as evidence of the paternal effect to update the posterior probability of the prior effect weights with uncertainty. During the update process, the Bayesian inference network intervenes in the mating genetic boundary matrix that is linked to the paternal line genetic similarity theory and the genetic kinship pedigree. The robust calculation of the smooth constraint Bayesian probability distribution is used to finally obtain the mating effect contribution weight of each paternal line sample to the settlement benefits of commercial meat pigs. The mating effect contribution weights are input into the comprehensive advantage model to solve for the matching gain between different parent lines and samples, generating different advantage matching matrices of parent lines and their comprehensive advantage indices.

[0009] Furthermore, the analysis and calculation of the significant impact of different paternal lines on the growth performance of commercial pigs, the significant impact of different maternal lines on the reproductive performance of commercial pigs, and the combined synergistic gain of different parental lines are performed separately based on growth trait data and reproductive performance data. This yields the paternal genetic growth contribution value for each paternal line sample, the maternal reproductive contribution value for each maternal line sample, and the combined synergistic gain value of different parental lines. Specifically, this includes the following steps: Growth weighting factors are set according to the importance of each growth trait index data for the reproductive development and breed matching of commercial meat pigs. The growth stability index is calculated by combining all growth weighting factors to construct the paternal genetic growth contribution value corresponding to each paternal breed sample. Based on the balancing scale of the economic value of commercial pig breeding according to the breeding performance data, breeding weight factors are set, and breeding grade scores are calculated by combining all breeding weight factors to construct the maternal breeding contribution value corresponding to each maternal breed sample. Using each paternal and maternal strain sample as a statistical unit, the settlement performance data of all commercial pigs bred by the statistical unit is summarized. The actual combination superiority value of the strain is calculated through the settlement performance data, and the theoretical combination superiority value is calculated by combining the paternal genetic growth contribution value and the maternal reproductive contribution value. The combination synergistic gain value of different parent strains is obtained by subtracting the actual combination superiority value from the theoretical combination superiority value.

[0010] Furthermore, S04 specifically includes the following steps: Obtain the market demand for individual commercial pigs in different sales regions, and at the same time obtain the supply update steps of breeding pigs to meet the corresponding market demand in different sales regions; Based on the various market research indicators for pork and the time-varying thresholds and settlement emphasis ratios corresponding to each indicator within the preset breeding period, a regional market target supply and demand vector for individual commercial pork is established in time series. Construct a time-slot sliding window, adjust the time-slot sliding window on the regional market target supply and demand vector according to the dynamic granularity of the supply update step, and calculate the implicit dependency relationship between different pork market inspection indicators within the window to generate an indicator coupling correlation matrix. Meanwhile, within the time-slot sliding window, the evaluation feedback scale of each index item in the comprehensive advantage model is adjusted based on the time-varying limit threshold, and the attention of each pork market inspection indicator and the corresponding different supporting advantage indicators are calculated based on the settlement emphasis ratio. The factor loading matrix of the factor analysis algorithm is adaptively and dynamically adjusted according to the weighting factors of the comprehensive advantage model corresponding to the supporting advantages of the indicator coupling correlation matrix, evaluation feedback scale and attention on the impact of the indicators on the pork market investigation indicators. Finally, a comprehensive advantage index ranking table matching each advantage matching matrix with the corresponding market demand in different sales regions is generated. Calculate the fit degree between each type of advantageous matching matrix and the target supply and demand vector of the regional market. Only extract candidate advantageous matching matrices with a fit degree greater than the preset fit degree threshold. Based on the comprehensive advantage index ranking table, select and recommend regional breeds for candidate advantageous matching matrices to obtain the optimal breed matching results for breeding pigs in different sales regions.

[0011] Furthermore, S05 specifically includes the following steps: The historical settlement data of commercial pigs after the entire process of breeding, raising and marketing in breeding farms, commercial farms and fattening farms are obtained by production and breeding logs, as well as the status characteristics of multiple matching combinations of the parent breeds of the selected historical advantage matching matrix in the preset past time period. Gaussian algorithm is introduced and combined with the biological theory of breeding pigs to construct a nonlinear breeding state space. The state characteristics of each breeding combination are encoded in multiple dimensions through the nonlinear breeding state space to obtain the parent breeding joint state sequence. Based on historical settlement performance data, a historical settlement trajectory matrix of individual commercial pigs is established. Based on the breeding pig-commercial pig breeding income standard benchmark, the income performance base of historical settlement is defined. By analyzing the global income of the historical advantage matching matrix through the income performance base in conjunction with the parent matching joint state sequence, the performance reward function of historical settlement income is defined. By injecting the historical settlement trajectory matrix into the supporting learning model based on maximum entropy inverse reinforcement, the performance reward function is used to maximize the solution of the supporting parameters corresponding to the settlement income fluctuation amplitude exceeding the preset fluctuation amplitude threshold, and the dynamic supporting coefficients that determine the historical settlement performance are output. The pre-trained and deployed pig breed advantage matching prediction model architecture is stripped away from the platform. At the same time, the actual reproductive performance data and real settlement performance data of different parent breeds corresponding to each advantage matching matrix are obtained. Based on the dynamic matching coefficient, the actual reproductive performance data and real settlement performance data are calculated in the whole process to obtain the predicted comprehensive advantage index. Obtain the actual comprehensive advantage index when the actual breeding performance data and the actual settlement performance data of different parent lines are output, calculate the deviation of the predicted comprehensive advantage index from the actual comprehensive advantage index, and obtain the matching deviation value; If the matching deviation is greater than the preset threshold, the matching prediction deviation of the paternal effect weight and the evaluation coefficient of each indicator will be corrected in reverse based on the matching deviation value, thus completing the closed-loop verification of the matching of breeding pig breeds.

[0012] This invention addresses the technical deficiencies in the prior art, and its beneficial technical effects are as follows: This invention utilizes multi-label sample atlases of parental breeds to track the genetic origin, reproductive performance, and offspring production and management data of breeding pigs throughout their entire life cycle. This overcomes the problem of separation between reproductive indicators and commercial pig profitability indicators in traditional breed evaluation processes, improving the accuracy of breed advantage identification. By quantifying the paternal growth genetic contribution, maternal reproductive contribution, and synergistic gains of parental combinations through a comprehensive advantage model, it can uncover the comprehensive impact of different parental breed combinations on indicators such as growth performance, feed conversion ratio, and market profitability of commercial pigs, forming a quantifiable advantage matching matrix. Furthermore, by dynamically adjusting evaluation weights based on regional market demand, it achieves precise matching of breeding pig breed matching programs in different sales regions, improving the adaptability of advantageous breed promotion and application. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0014] Figure 1 A flowchart of the first method for screening superior breeding pig lines based on a closed-loop breeding-meat pig dual-link system is shown. Figure 2 The flowchart of the second method for screening superior breeding pig lines based on the dual-link closed loop of breeding and meat pigs is shown. Detailed Implementation

[0015] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0016] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0017] The first aspect of this invention provides a method for screening superior breeding pig lines based on a closed-loop breeding-meat pig dual-link system, such as... Figure 1 As shown, it includes the following steps: S01: Obtain full-process production data from breeding pigs to commercial meat pigs, and assign unique strain codes to paternal and maternal strain samples of different breeding pig breeds based on the breeding source data and genetic information recorded in the full-process production data, and establish a multi-label sample map of the parent strains of individual breeding pigs. It should be noted that by establishing a multi-label sample map of the parent breeds of individual breeding pigs through this solution, which includes identification, paternal origin, maternal origin, genetic clustering tags, and breed codes, individual breeding pigs at different production and breeding stages can be accurately traced back to the corresponding parent breed origin. This enables digital information management of breeds throughout the entire process from breeding pig selection and reproduction to commercial pork production, providing reliable breed retrieval support for the subsequent correlation mapping between reproductive indicators and pork indicators.

[0018] S02: Extract reproductive performance data of individual breeding pigs and production and operation data of individual commercial pigs from the parent-child combination throughout the entire process. Based on the multi-label sample map of the parent breed, and according to ear number, batch traceability information and genetic lineage, perform correlation mapping between production and operation data and reproductive performance data to construct a dual-link mapping index library for the entire life cycle of breeding pig breeds. It should be noted that by constructing a dual-link mapping index library for breeding and meat pigs through this solution, a closed-loop mapping relationship between "breeding pig breeding data - genetic lines - commercial meat pig settlement results" can be achieved, providing a scientific retrieval index basis for the identification of superior lines, optimization of matching combinations, and precision breeding decisions.

[0019] S03: Construct a comprehensive advantage model for the synergistic analysis of the significant impact of different paternal lines on the growth performance of commercial pigs and different maternal lines on reproductive performance. Then, solve the matching gain of the breeding prior of the paternal line on the mating effect driving weight of the settlement benefit of commercial pigs through the comprehensive advantage model, and generate different advantage matching matrices of paternal and maternal lines and their comprehensive advantage index. S04: By calculating the fit between the corresponding market demand in different sales regions and different advantageous matching matrices, the evaluation weights of each index of the comprehensive advantage index are dynamically adjusted according to the meat pig duration assessment index of the demand vector. A comprehensive advantage index ranking table is generated to recommend the advantageous matching matrix with high fit, and the optimal breeding pig breed matching results for different sales regions are obtained. S05: Use historical settlement data of commercial pigs to reverse verify the accuracy of the prediction of the matching coefficients of the breeding pig breeds during actual breeding and make corrections, so as to realize the closed-loop verification of the matching of breeding pig breeds.

[0020] Furthermore, S01 specifically includes the following steps: By querying production and breeding logs, the entire process of breeding and converting breeding pigs into commercial pigs is retrieved from the breeding farm, commercial farm, and fattening farm within the preset breeding time period. Based on the breeding and circulation information recorded in the entire process production data, an identity file is created for each individual breeding pig. By extracting paternal and maternal strain samples from various individual pigs' production and breeding data, as well as the breeding source data corresponding to each strain sample, the reproductive capacity of all paternal and maternal strain samples is aggregated and analyzed based on the breeding source data to extract strain characteristics and obtain a series of parent strain feature vectors. The breeding source data includes strain name, strain number, source pig herd, source region, parent generation number, parent combination relationship, family number, reproductive performance, growth rate, disease resistance target, SNP marker data and allele information.

[0021] Genetic marker sequencing data and genetic loci of individual pigs corresponding to each paternal and maternal strain sample were obtained by retrieving from the breeding gene bank. Based on the differences in genetic loci, the Euclidean genetic distance between each pair of paternal and maternal strain samples was calculated to obtain the paternal genetic distance matrix and the maternal genetic distance matrix. Define the breed category of pigs as a genetic community unit. Based on the paternal genetic distance matrix or maternal genetic distance matrix, calculate the category genetic cost of the central clustering of the feature vectors of each parent breed within different genetic community units, and obtain the increase in genetic variance. Based on the breeding advantages of breeding pigs and the requirements for matching the breeds, a minimum genetic cost threshold is preset. If the increase in genetic variance is less than the minimum genetic cost threshold, the feature vectors of the two parent breeds are merged by category genetics, the corresponding genetic distance matrix is ​​updated, and finally the breed origin clustering labels of different breeding pig individuals are generated. Referring to the identification files, each parent breed sample is assigned a unique or non-unique breed code using breed origin clustering tags. For breeds with pure genetic origin and clearly belonging to a single independent breeding direction, a unique breed code is assigned. For mixed-related breeds with a common genetic background but multiple production uses, multiple non-unique codes are allowed to coexist based on their genetic community affiliation, production contribution relationship, and supporting application direction. A multi-label sample map of the parent breeds of individual breeding pigs is established.

[0022] It is important to note that the construction of the breeding pig breed coding system is the data foundation for realizing the tracking of the breeding pig's breeding chain, the selection of superior breeds, and the analysis of parent-child pairing relationships. When constructing this system, the first step is to retrieve production data covering the entire process of breeding pig mating, pregnancy, farrowing, weaning, rearing, transfer, and fattening of commercial pigs, based on the production management records of breeding farms, commercial farms, and fattening farms. This is because the superior performance of breeding pigs is not only determined by genetic background but is also influenced by the reproductive process, breeding environment, and production flow. Relying solely on data from a single breeding stage will not be able to fully depict the actual contribution of parent breeds to the production performance of offspring. The tamper-proof binding of the identification file provides each individual breeding pig with a physical anchor point that has pedigree traceability and spatiotemporal consistency with its breeding records, thereby locking the individual's movement trajectory. Different breeds inevitably develop specific reproductive potential, production performance, and genetic stability characteristics during the long-term selection process. Classification based on manual registration names or traditional breed classifications is easily influenced by selection history, farm numbering rules, and human management methods, leading to blurred breed boundaries. Based on the parentage information, mating records, and breed origin records in the identification files, the origins of all paternal and maternal breed samples were traced. Based on this data, the reproductive capacity of different parent breeds was aggregated and analyzed. Specifically, using the parity reproduction records of each breeding pig as input, the parity reproduction data of multiple breeding pigs under the same paternal or maternal breed were normalized over a time window to eliminate scale differences caused by different batches and breeding cycles. Subsequently, a three-level correlation matrix of breed-individual-production indicators was constructed, mapping the reproductive quantity, offspring production performance, and genetic stability indicators of each breed-corresponding breeding pig individual to the feature space. Locally weighted regression was used to fit the productivity variation with parity for indicators such as litter size and weaning litter weight, where reproductive quantity... The indicators detail the strain's propagation capacity, offspring growth indicators characterize genetic gain capacity, and health and survival indicators describe environmental adaptability. Through dynamic weight adjustment, a comprehensive reproductive capacity evaluation value for the strain is formed. Finally, the multidimensional evaluation results corresponding to each strain are vectorized and expressed. Then, the initial capacity level, capacity decay slope, and reproductive lifespan shape parameters of the alienation trajectory are extracted to establish a concrete representation of the multidimensional capacity decay trajectory of the strain. At the two biological dimensions of capacity persistence and reproductive peak expression, the number of offspring, reproductive success rate, offspring survival rate, growth rate, feed conversion performance, and commercial performance indicators of each strain sample are statistically analyzed. The fragmented strain advantage of production and reproductive performance information is represented globally, thereby depicting the different patterns of different strains in coping with reproductive cycle fatigue.

[0023] After extracting breed features, pig breed categories are defined as genetic community units, which are natural clusters formed by individuals with similar genetic backgrounds. Breeding based on genetic distance can easily overlook the distribution of genetic structure within a breed and the potential shared genetic background between breeds; for example, two parental breeds may have a relatively close overall genetic distance, but their individuals may be highly dispersed. Each parental breed feature vector is placed within different genetic community units, and the variance of the Euclidean genetic distance within the merged community is measured as the increase in genetic variance. Specifically, the paternal or maternal genetic distance matrix is ​​converted into a weighted genetic connection graph, where breeds are graph nodes and genetic distance is the connection weight between nodes. Subsequently, the genetic community structure is used to identify the clustering regions of different breed nodes in the genetic space, and the centrality index of each parental breed feature vector within its respective genetic community is calculated. This centrality index reveals the degree of representation of a breed within a genetic community unit, i.e., the average genetic similarity between that breed and other breed samples within the genetic community. For two parental lines to be evaluated for merging, based on the genetic distance distribution between all parental line samples, the changes in the center position within the genetic community, the average genetic distance increment, and the offset of the line feature vector before and after the addition of the current feature vector are calculated. A categorical genetic cost function is constructed. The physical meaning of the increase in genetic variance is the degree of expansion of genetic variance within a community unit when attempting to aggregate the feature vectors of two lines into the same genetic community. The smaller the increment, the more similar the genetic background of the lines, and the less significant the loss of population genetic homogeneity after merging. Quantifying the increase in variance means that the compatibility of the two parental lines is no longer judged solely based on the genetic distance between them, but rather by comprehensively considering the integrity of the genetic structure and the degree of retention of line characteristics after merging. This yields a categorical genetic cost that better conforms to breeding principles and improves the genetic interpretability of line coding. When the categorical genetic cost generated by merging the feature vectors of two parental lines is lower than a preset threshold, it indicates that the two lines are highly similar in terms of genetic structure, production characteristics, and breeding contributions. Merging the categories will not significantly reduce the line distinguishing ability, therefore, a genetic merging operation of similar categories is performed.

[0024] Furthermore, S02 specifically includes the following steps: By acquiring reproductive performance data of various pig individuals corresponding to parent combinations and production and operation data of corresponding commercial pigs through full-process production data, and establishing genetic phylogenetic charts of different commercial pig individuals based on biological kinship clues registered in the mating and birth information; The reproductive performance data includes the number of successful matings, semen batches, conception rate, culling age, total number of piglets born, number of live piglets, number of deformed piglets, average litter weight, piglet birth weight, suckling survival rate, and piglet mortality rate. The production and operation data includes the preferred slaughter weight, minimum lean meat percentage, slaughter grade requirements, sales price level, feed cost, mortality rate, and feed conversion ratio.

[0025] A multi-layered directed hypergraph framework for constructing a high-dimensional information grid is constructed by introducing a deep learning network. The multi-layered directed hypergraph framework is hierarchically divided according to the indicator dimension of reproductive performance data to form the first directed heterogeneous layer. At the same time, the multi-layered directed hypergraph framework is divided according to the indicator dimension of production and operation data to form the second directed heterogeneous layer. Define individual commercial pigs as a type of performance link node, and embed the kinship direction of the genetic kinship pedigree diagram into the first directed heterogeneous layer as the genetic lineage edge. Then, attach production and operation data as the settlement attribute of a type of performance link node to obtain the strain settlement loading channel diagram of individual commercial pigs. The genetic lineage edges not only record the direction of parental relationships but also carry additional information such as genetic contribution probability, kinship level, and breed origin tags, allowing commercial pig nodes to trace their corresponding parent breed origins along the genetic edges. Simultaneously, production and operational data are loaded as settlement attributes for a type of performance link node. Different settlement indicators correspond to different attribute dimensions; for example, weight gain corresponds to economic performance attributes, feed conversion ratio corresponds to feeding efficiency attributes, and slaughter performance corresponds to commercial value attributes. The breed settlement loading channel graph ensures that each commercial pig's settlement performance has a traceable genetic origin path.

[0026] Define individual breeding pigs as nodes in the second-class performance link, and embed the sample index of the multi-label sample map of the parent breed as the breed membership edge into the second directed heterogeneous layer. Add reproductive performance data as the reproductive attribute of the second-class performance link node to obtain the breed reproductive loading channel map of individual breeding pigs. The breed affiliation edge describes the probability relationship of a breeding pig belonging to a specific paternal or maternal breed. When a breeding pig exhibits characteristics from multiple genetic sources, multi-labeling can preserve the correlation of contributions from multiple breeds, avoiding the loss of genetic information caused by traditional single-label encoding. Furthermore, breeding pig reproductive performance data is appended as reproductive attributes to the second-class performance link nodes, including indicators such as mating ability, litter size, offspring quality, and genetic stability, thus forming a breed reproductive loading channel diagram describing "which breed the breeding pig belongs to" and "what reproductive capacity this breed possesses." An entry point for evaluating genetic resources at the breeding pig end is established, providing parental data support for the reverse indicator mapping of subsequent commercial meat pig settlement performance.

[0027] Obtain electronic ear tag information and birth batch information for the management of the breeding pig breeding process. Combine the electronic ear tag information and birth batch information as the root base point. Starting from the root base point, perform a reverse search traversal of the strain settlement loading channel graph along the parent relationship edge in the multi-layer directed hypergraph framework, and output the strain grandparent node. A strain reference matrix is ​​constructed based on the genetic network of the multi-label sample map of the parent strains. Based on the genetic constraints of the strain reference matrix, the strain settlement loading channel graph is decomposed into a low-rank decomposition of the strain genetic origin to generate the strain contribution index of the direct parent corresponding to the ancestral node of the strain. Let the source expression matrix of the target strain be as follows: (The individual pigs used for market meat are listed as candidate grandparent strains), which can be approximated as the product of two low-rank non-negative factor matrices. ,in The latent factor basis matrix of strain characteristics, This represents the contribution coefficient matrix of the brood lines. To incorporate the genetic prior of the brood line reference matrix into the decomposition process, a target loss function with graph regularization constraints is constructed:

[0028] In the formula, To constrain sparsity and select dominant contributing strains, This is the regularization term for the Laplace diagram derived from the strain reference matrix. For matrix trace operations, it forces lines with similar genetic structures or close centrality to have a smooth distribution of contribution coefficients in the decomposition low-dimensional space.

[0029] Based on the breed contribution index, the settlement indicator data of individual commercial pigs are mapped to the reproductive indicator database of the parent breeds of breeding pigs, thus constructing a dual-link mapping indicator library of the breeding-meat pigs throughout the entire life cycle of breeding pig breeds.

[0030] It is important to note that reproductive performance data (reproductive indicator attributes) and production and management data (pig settlement indicator attributes) belong to completely different observation domains. The former reflects the long-term reproductive capacity of the breeding pig population, while the latter characterizes the terminal output efficiency of commercial pigs. There is a natural disconnect between the two in terms of time span, measurement dimensions, and spatial attribution. Directly and simply linking them would break the attribution chain between the breeding contribution of the pig breed and the fattening performance of commercial pigs, causing data link disorder between reproductive indicators and pig settlement indicators. To address this, a graph attention autoencoder is trained using a deep learning network to uniformly encode the entire production process data. Indicators such as litter size, weaning litter weight, and number of effective teats in the reproductive performance data are converted into high-dimensional feature vectors and mapped to a hypergraph space, creating a first directed heterogeneous layer. In this first layer, nodes of different breed types are connected through genetic relationships, and reproductive performance indicators are embedded as node attributes, clarifying the dual relationship of "who produces offspring" and "how well they produce offspring," thus improving the accuracy of the analysis of the reproductive contribution of the parent breed. Similarly, within the hypergraph space, the daily weight gain, feed conversion ratio, backfat thickness, and other indicators of production and operation data are transformed and mapped to create a second directed heterogeneous layer. In this second layer, individual commercial pigs serve as production settlement performance nodes, and different operational indicators serve as node features. A commercial stage evaluation network is formed through production link relationships, and a multi-layer graph attention mechanism is used to calculate the contribution weight of different production indicators to the commercial value performance, ensuring that the stage evaluation of commercial pigs retains genetic background information. A multi-layer directed hypergraph structure is adopted, with both layers sharing the same underlying genetic kinship framework, but each unfolding in an independent attribute space. This allows nodes of the same breed to simultaneously associate genetic relationships, reproductive relationships, and settlement performance relationships, thereby achieving genetic tracing and semantic decoupling of the entire production process from breeding pig breed reproduction to commercial pig settlement, while maintaining a consistent topological structure.

[0031] By combining electronic ear tag information and brood batch information from the breeding process of pigs, a root basis point with spatiotemporal dual anchoring is established. Starting from the unique entry point of the root basis point, a reverse search traversal is performed along the edge of the parent-child relationship to traverse the strain settlement loading channel graph, visiting the paternal node, maternal node, and grandparent strain node layer by layer, until a node with no higher-level strain parent is reached, outputting the strain grandparent node. The electronic ear tag ensures the unique identification of the individual entity, while the brood batch locks in the genetic background of the population within a time window. The combination of the two can effectively resolve traceability ambiguities caused by management operations such as ear tag re-attaching and group transfer, realizing reverse tracing from commercial pigs to their strain genetic ancestry, and providing a correlation mapping action between breeding and pig settlement. After obtaining the strain grandparent node, it is necessary to further quantify the genetic contribution share of each strain grandparent parent to the commercial pig settlement indicators. To this end, a pedigree reference matrix is ​​constructed based on the genetic network of multi-label sample maps of parental breeds. This pedigree reference matrix describes the genetic similarity, production contribution, and reproductive performance relationships among different breeds, serving as a genetic constraint basis. The matrix rows represent parental breed nodes, and the columns represent genetic markers, production indicators, and reproductive characteristics. Pedigree genetic constraints are incorporated into the non-negative matrix factorization process to ensure that the latent factors after factorization remain consistent with the true genetic structure. Specifically, the commercial pig settlement index matrix is ​​decomposed into a pedigree contribution matrix and an individual performance matrix. Through iterative optimization of the non-negative constraint objective function, the settlement performance of each commercial pig node is decomposed into multiple parental breed contribution terms, ultimately obtaining a pedigree contribution index that reflects the actual contribution ratio of different paternal and maternal breeds to the production advantage of commercial pigs. The genetic reference matrix confines the decomposition direction to the genetic space of known breeds.

[0032] Furthermore, S03, as Figure 2 As shown, the specific steps include: The growth trait data of each commercial pig generated by different paternal line samples within a preset breeding period were extracted using the breeding-meat pig dual-link mapping index library, and the reproductive performance data of each commercial pig generated by different maternal line samples within a preset breeding period were also extracted. The growth trait data include carcass quality, slaughter weight, and average daily weight gain. The reproductive performance data include, but are not limited to, number of litters farrowed, number of live piglets farrowed, total number of piglets per litter, number of live piglets per litter, number of healthy piglets per litter, number of deformed piglets per litter, suckling survival rate, and piglet mortality rate.

[0033] To investigate the significant impact of different paternal lines on the growth performance of commercial pigs, the significant impact of different maternal lines on the reproductive performance of commercial pigs, and the combined superiority of different parental lines, growth trait data and reproductive performance data were analyzed and calculated separately to obtain the paternal genetic growth contribution value, the maternal reproductive contribution value, and the combined synergistic gain value of different parental lines for each paternal line sample. A comprehensive advantage model is established based on the paternal genetic growth contribution value, maternal reproductive contribution value, and combined synergistic gain value. This comprehensive advantage model includes a reproductive advantage index, a meat pig benefit index, a market value index, and a stability index. Historical breeding data of commercial meat pigs produced from individual breeding pigs corresponding to each paternal line sample are obtained from the breeding database within a predetermined time period. The historical breeding data is used to evaluate each paternal line sample and assign a prior effect weight to each sample. The historical breeding data includes historical growth trait data of offspring produced by previous matings of the paternal line, settlement performance data of meat pigs, mating and breeding records, and batch performance fluctuation frequency.

[0034] By analyzing settlement data, we obtain actual offspring settlement performance data of commercial pigs bred based on samples of each paternal line. We introduce a Bayesian inference network and use the likelihood operators of different pig observation dimensions contained in the actual offspring settlement performance data as evidence of the paternal effect to update the posterior probability of the prior effect weights with uncertainty. The actual offspring settlement performance data includes offspring growth performance, feed utilization rate, carcass traits and disease resistance information, and utilizes actual offspring settlement performance data; During the update process, the Bayesian inference network intervenes in the mating genetic boundary matrix that is linked to the paternal line genetic similarity theory and the genetic kinship pedigree. The robust calculation of the smooth constraint Bayesian probability distribution is used to finally obtain the mating effect contribution weight of each paternal line sample to the settlement benefits of commercial meat pigs. The comprehensive advantage model is calculated by inputting the contribution weight of mating effect into the comprehensive advantage model to solve the matching gain between different parent lines. The contribution weight of the parent effect is used as the growth performance contribution weight, the reproductive contribution value of the parent is used as the reproductive capacity weight, and the combination synergistic gain value is used as the matching optimization compensation term. Through multi-objective optimization, the comprehensive advantage of different parent combinations is calculated to generate different advantage matching matrices of parent lines and their comprehensive advantage indices.

[0035] It should be noted that the economic traits of commercial pigs, such as growth rate, feed utilization efficiency, and carcass quality, are mainly influenced by the genetic effects of the paternal line, while the maternal line determines more of the sow's reproductive capacity, the initial developmental level of offspring, and the efficiency of herd expansion. Traditional comprehensive advantage evaluation simply adds the effects of the paternal and maternal lines together, which easily leads to unclear attribution of genetic contributions and fails to accurately reflect the actual role of different parent lines in defining the production value of commercial pigs. Based on the characteristics that different paternal lines have a significant impact on the growth performance of commercial pigs and maternal lines have a significant impact on reproductive performance, we calculate the paternal genetic growth contribution value, the maternal reproductive contribution value, and the combined synergistic gain value of different parent lines. The physical meaning of the paternal abnormal growth contribution value is the average additive genetic contribution of the paternal line to the terminal growth indicators of breeding pigs under conditions independent of the maternal background. The meaning of the maternal reproductive contribution value is the average genetic transmission power of the maternal line to reproductive performance. The combined synergistic gain value is used to capture the specific combining power in heterosis of breeding pig lines. Among them, the reproductive advantage index is used to evaluate the long-term propagation capacity of the maternal line, mainly integrating supporting advantage indicators such as farrowing rate, number of weaned pigs, survival rate, average weaning weight per litter, and reproductive cycle; the hog profitability index evaluates the impact of the sire line on the production performance of commercial hogs, aggregating the weighted output of indicators such as daily weight gain, feed conversion ratio, slaughter rate, carcass quality, and mortality rate; the market value index reflects the differences in the economic value of different commercial hog traits in the market, integrating indicators such as average selling price, slaughter grade, and settlement profit; the stability index is used to evaluate the inverse of the coefficient of variation of the reproductive and growth performance of a certain lineage across different batches and environmental conditions over different years, and penalizes environmentally sensitive lines, based on the evaluation and formulation of multiple batch fluctuation coefficients.

[0036] Since the actual contribution of paternal lines to the settlement benefits of commercial pigs is influenced by the heterogeneity of offspring batches, breeding environment, and maternal lines, this scheme evaluates historical breeding data to initially define the stable contribution capacity of different paternal lines in long-term production. Each paternal line is assigned an initial prior effect weight, which represents an estimate of paternal genetic advantage based on historical experience. However, due to the limited number of historical samples and continuous changes in the production environment, this weight still has some uncertainty. Therefore, this scheme further utilizes a Bayesian inference network to encode the prior effect weights as a Dirichlet distribution prior. Likelihood information is constructed for each observation dimension in the actual offspring settlement performance data: Gaussian likelihood for growth performance, ordered multinomial likelihood for carcass grade, and binomial likelihood for disease resistance, forming a multi-source evidence chain. Within the variational inference framework, the effect weights of each paternal line are iteratively updated by minimizing the KL divergence between the prior distribution and the approximate posterior distribution based on this multi-source evidence chain, resulting in a posterior estimate that combines prior laws with contemporary experimental feedback. Actual offspring performance data is used as dynamic observation evidence to update the prior effect weights posteriorly. A breeding genetic boundary matrix is ​​constructed by incorporating paternal line genetic similarity theory and genetic pedigree charts into the Bayesian inference network. This matrix defines the genetic similarity between different paternal lines, serving as a constraint on the probability update process and pulling the effect weights of genetically similar lines towards a common local mean. When two paternal lines have high genetic similarity, their effect weight changes are mutually constrained, effectively avoiding estimation divergence caused by sparse offspring settlement samples of a particular pig line or abnormal environmental fluctuations. The hybrid vigor matrix quantifies the heterosis potential and final economic benefits of each pig line combination, expressing the expected production returns, reproductive efficiency, and genetic stability after different paternal and maternal line combinations.

[0037] Furthermore, S04 specifically includes the following steps: Obtain the market demand for individual commercial pigs in different sales regions, and at the same time obtain the supply update steps of breeding pigs to meet the corresponding market demand in different sales regions; Based on the various market research indicators for pork and the time-varying thresholds and settlement emphasis ratios corresponding to each indicator within the preset breeding period, a regional market target supply and demand vector for individual commercial pork is established in time series. The indicators for the pork market survey include preferred slaughter weight, minimum lean meat percentage, slaughter grade requirements, sales price level, feed cost, mortality rate, and feed conversion ratio. Construct a time-slot sliding window, adjust the time-slot sliding window on the regional market target supply and demand vector according to the dynamic granularity of the supply update step, and calculate the implicit dependency relationship between different pork market inspection indicators within the window to generate an indicator coupling correlation matrix. Meanwhile, within the time-slot sliding window, the evaluation feedback scale of each index item in the comprehensive advantage model is adjusted based on the time-varying limit threshold, and the attention of each pork market inspection indicator and the corresponding different supporting advantage indicators are calculated based on the settlement emphasis ratio. The factor loading matrix of the factor analysis algorithm is adaptively and dynamically adjusted according to the weighting factors of the comprehensive advantage model corresponding to the supporting advantages of the indicator coupling correlation matrix, evaluation feedback scale and attention on the impact of the indicators on the pork market investigation indicators. Finally, a comprehensive advantage index ranking table matching each advantage matching matrix with the corresponding market demand in different sales regions is generated. Calculate the fit degree between each type of advantageous matching matrix and the target supply and demand vector of the regional market. Only extract candidate advantageous matching matrices with a fit degree greater than the preset fit degree threshold. Based on the comprehensive advantage index ranking table, select and recommend regional breeds for candidate advantageous matching matrices to obtain the optimal breed matching results for breeding pigs in different sales regions.

[0038] It should be noted that different sales regions are influenced by differences in consumption habits, slaughter standards, price levels, breeding costs, and market preferences. Some regions focus more on high slaughter weight and rapid growth performance, while others focus more on lean meat percentage, carcass grade, and feed utilization efficiency. By constructing a regional market target supply and demand vector for individual commercial pigs in different sales regions based on market demand, each component of this vector corresponds to a composite code of the target value range (time-varying threshold) and settlement weight (settlement emphasis ratio) of a pig market assessment indicator in that time series segment. The time-varying threshold represents the dynamic demand threshold that each pig market assessment indicator should adjust to in a particular sales region's supply and demand for large-sized pigs. Market demand for indicators such as slaughter weight and lean meat percentage exhibits seasonal fluctuations and cyclical trends. For example, the minimum lean meat percentage in a certain sales region during a certain quarter may be dynamically increased to 55% (threshold boundary) following market trends. The dynamic attribute of the "threshold" effectively avoids the smoothing effect of market demand peaks and troughs on the ranking of breed advantages. Furthermore, the contribution ratio of each indicator to regional revenue is calculated based on market settlement data, forming the settlement emphasis ratio. The supply update step size refers to the interval at which the supply of commercial pigs for different sales regions is updated to reflect the demand for breeding pigs. Implicit correlations exist among the indicators in the pork market. For example, increased slaughter weight typically affects feed costs and feed conversion ratios, while improved lean meat percentage may influence sales prices and slaughter grade evaluations. Since the different evaluation weights of the comprehensive advantage index need to be dynamically adjusted according to market demand, the market coexistence dependence of these indicators must be considered. Specifically: within each time slot sliding window of a certain length, the empirical distribution of each indicator is nonparametrically fitted based on the dynamic granularity of the supply update step size. The optimal family of Copula functions is selected to characterize the tail dependency structure between indicators. Copula entropy is calculated as a measure of nonlinear dependency strength, thereby decoupling the marginal distribution from the dependency structure. This ensures that the strong correlation between slaughter weight and slaughter grade, among other indicators, is accurately captured and not diluted by long-tail noise. The time-varying dependency matrices of all windows are aggregated to generate an indicator coupling correlation matrix.

[0039] During sliding window scanning, the evaluation feedback scale of each index item in the comprehensive advantage model is adjusted based on time-varying thresholds. When the market tolerance range of a certain indicator narrows, the curvature of the scoring function of the corresponding index item increases, amplifying small performance differences near the threshold boundary; when the tolerance range widens, the curvature decreases to avoid oversensitivity. For example, in the high-end meat market, the feedback scale of indices related to lean meat percentage and carcass quality is increased. Based on the settlement emphasis ratio, the attention matrix between each market inspection index and the corresponding supporting advantage indicators of each index item in the comprehensive advantage model is calculated. This matrix quantifies the attention allocation weight of each market indicator to reproductive advantage, hog profitability, market value, and stability, and serves as the basis for dynamic adjustment of the evaluation weights of different index components of the comprehensive advantage model. For example, the market value index is easily affected by the dominant fluctuations in sales price levels, while the hog profitability index is affected by feed conversion ratio. The adaptive dynamic adjustment calculation of the trade-off factors specifically involves: extracting principal factors from the index coupling correlation matrix to obtain the potential driving factor structure of market demand; then projecting the attention matrix onto this factor space to calculate the loading coefficients of each factor on the four comprehensive advantage index items. Changes in the loading coefficients directly reflect the strength of the influence of potential market demand factors on various dimensions of advantage evaluation. This allows for dynamic adjustments to the trade-off factors, replacing the fixed index weights in the general model with adaptive parameters driven by the potential demand structure of the regional market. Finally, the adjusted trade-off factors are substituted into the comprehensive advantage model to recalculate the comprehensive advantage index order for each advantage-supporting matrix.

[0040] Furthermore, S05 specifically includes the following steps: The historical settlement data of commercial pigs after the entire process of breeding, raising and marketing in breeding farms, commercial farms and fattening farms are obtained by production and breeding logs, as well as the status characteristics of multiple matching combinations of the parent breeds of the selected historical advantage matching matrix in the preset past time period. The historical settlement performance data includes, but is not limited to, the number of farmers raising commercial pigs, the number of piglets received, the average weight of piglets received, the market transport rate, the market rate, the first-grade product rate, the average market weight, and the corrected feed conversion ratio.

[0041] Gaussian algorithm is introduced and combined with the biological theory of breeding pigs to construct a nonlinear breeding state space. The state characteristics of each breeding combination are encoded in multiple dimensions through the nonlinear breeding state space to obtain the parent breeding joint state sequence. Based on historical settlement performance data, a historical settlement trajectory matrix of individual commercial pigs is established. Based on the breeding pig-commercial pig breeding income standard benchmark, the income performance base of historical settlement is defined. By analyzing the global income of the historical advantage matching matrix through the income performance base in conjunction with the parent matching joint state sequence, the performance reward function of historical settlement income is defined. By injecting the historical settlement trajectory matrix into the supporting learning model based on maximum entropy inverse reinforcement, the performance reward function is used to maximize the solution of the supporting parameters corresponding to the settlement income fluctuation amplitude exceeding the preset fluctuation amplitude threshold, and the dynamic supporting coefficients that determine the historical settlement performance are output. The pre-trained and deployed pig breed advantage matching prediction model architecture is stripped away from the platform. At the same time, the actual reproductive performance data and real settlement performance data of different parent breeds corresponding to each advantage matching matrix are obtained. Based on the dynamic matching coefficient, the actual reproductive performance data and real settlement performance data are calculated in the whole process to obtain the predicted comprehensive advantage index. Obtain the actual comprehensive advantage index when the actual breeding performance data and the actual settlement performance data of different parent lines are output, calculate the deviation of the predicted comprehensive advantage index from the actual comprehensive advantage index, and obtain the matching deviation value; If the matching deviation is greater than the preset threshold, the matching prediction deviation of the paternal effect weight and the evaluation coefficient of each indicator will be corrected in reverse based on the matching deviation value, thus completing the closed-loop verification of the matching of breeding pig breeds.

[0042] It should be noted that pig breed mating exhibits typical nonlinear combinatorial characteristics. The paternal and maternal effects not only involve additive genetic contributions but also complex interactions between dominant and epistatic effects. Feature enhancement of the Gaussian state space is achieved by incorporating biological constraints related to breed mating. For example, paternal genetic effects are used as constraints on growth performance, maternal reproductive effects as constraints on reproductive performance, and parental synergistic gains as constraints on combinatorial advantage, ensuring the state space conforms to real breeding patterns and establishing a joint state sequence for parental mating of pig breeds. Simultaneously, a complete observation chain (historical settlement trajectory matrix) is constructed based on historical settlement performance data, from mating state to settlement result. After obtaining the joint state sequence and settlement trajectory, the global benefit generated by each mating decision in the historical advantage mating matrix is ​​analyzed by coordinating the benefit performance base with the parental mating joint state sequence, and a performance reward function for historical settlement benefits is defined. This reward function is not a simple linear mapping of the final settlement amount. Instead, it takes into account the input of breeding pig production, the market value of meat pigs, and the contribution of breed genetic advantages, and adopts an exponential utility function. It uses a risk aversion coefficient to penalize the fluctuation of settlement income, so that the reward function can reflect the true preference structure of breed breeding decisions between the economic benefits of commercial meat pigs and the stability of breeding pig production.

[0043] However, the evaluation weights in the historical advantage matching matrix, such as the trade-off coefficient between reproductive advantage and piglet benefits, were preset based on historical experience in the previous comprehensive advantage model. Their adaptability to matching returns under different environments and market conditions at different times has not been rigorously verified. Directly using these fixed weights to guide future matching will fail to distinguish between "weight setting bias" and "true genetic effects of the breed." Therefore, the historical settlement trajectory matrix is ​​injected into a matching learning model based on maximum entropy inverse reinforcement to calculate its dynamic matching coefficients. Specifically, the model treats the joint state sequence of parental matching as the state access frequency of an expert strategy. With the goal of maximizing the performance reward function, and under the constraint of maximizing the strategy distribution entropy, it solves the problem alternately through value iteration and strategy optimization, extracting matching parameters that lead to significant positive fluctuations in settlement returns, i.e., fluctuations greater than a preset fluctuation threshold. If the matching deviation exceeds the preset threshold, it indicates that the predicted current matching parameters cannot reflect the true optimal production law of the pig breed combination, resulting in low matching prediction accuracy. Therefore, the matching prediction deviations of the paternal effect weights and the evaluation coefficients of each indicator are corrected.

[0044] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for selecting superior breeding pig lines based on a closed-loop breeding-meat pig dual-link system, characterized in that, Includes the following steps: S01: Obtain full-process production data from breeding pigs to commercial meat pigs, and assign unique strain codes to paternal and maternal strain samples of different breeding pig breeds based on the breeding source data and genetic information recorded in the full-process production data, and establish a multi-label sample map of the parent strains of individual breeding pigs. S02: Extract reproductive performance data of individual breeding pigs and production and operation data of individual commercial pigs from the parent-child combination throughout the entire process. Based on the multi-label sample map of the parent breed, and according to ear number, batch traceability information and genetic lineage, perform correlation mapping between production and operation data and reproductive performance data to construct a dual-link mapping index library for the entire life cycle of breeding pig breeds. S03: Construct a comprehensive advantage model for the synergistic analysis of the significant impact of different paternal lines on the growth performance of commercial meat pigs and the significant impact of different maternal lines on the reproductive performance. Then, solve the matching gain of the breeding prior of the paternal line on the mating effect driving weight of the settlement benefit of commercial meat pigs through the comprehensive advantage model, and generate different advantage matching matrices of paternal and maternal lines and their comprehensive advantage index. S04: By calculating the fit between the corresponding market demand in different sales regions and different advantageous matching matrices, the evaluation weights of each index of the comprehensive advantage index are dynamically adjusted according to the meat pig duration assessment index of the demand vector. A comprehensive advantage index ranking table is generated to recommend the advantageous matching matrix with high fit, and the optimal breeding pig breed matching results for different sales regions are obtained. S05: Use historical settlement data of commercial pigs to reverse verify the accuracy of the prediction of the matching coefficients of the breeding pig breeds during actual breeding and make corrections, so as to realize the closed-loop verification of the matching of breeding pig breeds.

2. The method for screening superior breeding pig lines based on a closed-loop breeding-meat pig dual-linkage as described in claim 1, characterized in that, S01 specifically includes the following steps: By querying production and breeding logs, the entire process of breeding and converting breeding pigs into commercial pigs is retrieved from the breeding farm, commercial farm, and fattening farm within the preset breeding time period. Based on the breeding and circulation information recorded in the entire process production data, an identity file is created for each individual breeding pig. By extracting paternal and maternal strain samples from various individual pigs' production and breeding data, as well as the breeding source data corresponding to each strain sample, the reproductive capacity of all paternal and maternal strain samples is aggregated and analyzed based on the breeding source data to extract strain characteristics and obtain a series of parent strain feature vectors. Genetic marker sequencing data and genetic loci of individual pigs corresponding to each paternal and maternal strain sample were obtained by retrieving from the breeding gene bank. Based on the differences in genetic loci, the Euclidean genetic distance between each pair of paternal and maternal strain samples was calculated to obtain the paternal genetic distance matrix and the maternal genetic distance matrix. Define the breed category of pigs as a genetic community unit. Based on the paternal genetic distance matrix or maternal genetic distance matrix, calculate the category genetic cost of the central clustering of the feature vectors of each parent breed within different genetic community units, and obtain the increase in genetic variance. Based on the breeding advantages of breeding pigs and the requirements for matching the breeds, a minimum genetic cost threshold is preset. If the increase in genetic variance is less than the minimum genetic cost threshold, the feature vectors of the two parent breeds are merged by category genetics, the corresponding genetic distance matrix is ​​updated, and finally the breed origin clustering labels of different breeding pig individuals are generated. Referring to the identification files, the breed origin clustering tags are used to perform unique or non-unique breed coding operations on each parent breed sample to establish a multi-label sample map of the parent breeds of individual breeding pigs.

3. The method for screening superior breeding pig lines based on a closed-loop breeding-meat pig dual-linkage as described in claim 1, characterized in that, S02 specifically includes the following steps: By acquiring reproductive performance data of various pig individuals corresponding to parent combinations and production and operation data of corresponding commercial pigs through full-process production data, and establishing genetic phylogenetic charts of different commercial pig individuals based on biological kinship clues registered in the mating and birth information; A multi-layered directed hypergraph framework for constructing a high-dimensional information grid is constructed by introducing a deep learning network. The multi-layered directed hypergraph framework is hierarchically divided according to the indicator dimension of reproductive performance data to form the first directed heterogeneous layer. At the same time, the multi-layered directed hypergraph framework is divided according to the indicator dimension of production and operation data to form the second directed heterogeneous layer. Define individual commercial pigs as a type of performance link node, and embed the kinship direction of the genetic kinship pedigree diagram into the first directed heterogeneous layer as the genetic lineage edge. Then, attach production and operation data as the settlement attribute of a type of performance link node to obtain the strain settlement loading channel diagram of individual commercial pigs. Define individual breeding pigs as nodes in the second-class performance link, and embed the sample index of the multi-label sample map of the parent breed as the breed membership edge into the second directed heterogeneous layer. Add reproductive performance data as the reproductive attribute of the second-class performance link node to obtain the breed reproductive loading channel map of individual breeding pigs. Obtain electronic ear tag information and birth batch information for the management of the breeding pig breeding process. Combine the electronic ear tag information and birth batch information as the root base point. Starting from the root base point, perform a reverse search traversal of the strain settlement loading channel graph along the parent relationship edge in the multi-layer directed hypergraph framework, and output the strain grandparent node. A strain reference matrix is ​​constructed based on the genetic network of the multi-label sample map of the parent strains. Based on the genetic constraints of the strain reference matrix, the strain settlement loading channel graph is decomposed into a low-rank decomposition of the strain genetic origin to generate the strain contribution index of the direct parent corresponding to the ancestral node of the strain. Based on the breed contribution index, the settlement indicator data of individual commercial pigs are mapped to the reproductive indicator database of the parent breeds of breeding pigs, thus constructing a dual-link mapping indicator library of the breeding-meat pigs throughout the entire life cycle of breeding pig breeds.

4. The method for screening superior breeding pig lines based on a closed-loop breeding-meat pig dual-linkage as described in claim 1, characterized in that, S03 specifically includes the following steps: The growth trait data of each commercial pig generated by different paternal line samples within a preset breeding period were extracted using the breeding-meat pig dual-link mapping index library, and the reproductive performance data of each commercial pig generated by different maternal line samples within a preset breeding period were also extracted. To investigate the significant impact of different paternal lines on the growth performance of commercial pigs, the significant impact of different maternal lines on the reproductive performance of commercial pigs, and the combined superiority of different parental lines, growth trait data and reproductive performance data were analyzed and calculated separately to obtain the paternal genetic growth contribution value, the maternal reproductive contribution value, and the combined synergistic gain value of different parental lines for each paternal line sample. A comprehensive advantage model is established based on the paternal genetic growth contribution value, maternal reproductive contribution value, and combined synergistic gain value. This comprehensive advantage model includes a reproductive advantage index, a meat pig benefit index, a market value index, and a stability index. Historical breeding data of commercial meat pigs produced from individual breeding pigs corresponding to each paternal line sample are obtained from the breeding database within a predetermined time period. The historical breeding data is used to evaluate each paternal line sample and assign a prior effect weight to each sample. By analyzing settlement data, we obtain actual offspring settlement performance data of commercial pigs bred based on samples of each paternal line. We introduce a Bayesian inference network and use the likelihood operators of different pig observation dimensions contained in the actual offspring settlement performance data as evidence of the paternal effect to update the posterior probability of the prior effect weights with uncertainty. During the update process, the Bayesian inference network intervenes in the mating genetic boundary matrix that is linked to the paternal line genetic similarity theory and the genetic kinship pedigree. The robust calculation of the smooth constraint Bayesian probability distribution is used to finally obtain the mating effect contribution weight of each paternal line sample to the settlement benefits of commercial meat pigs. The mating effect contribution weights are input into the comprehensive advantage model to solve for the matching gain between different parent lines and samples, generating different advantage matching matrices of parent lines and their comprehensive advantage indices.

5. The method for screening superior breeding pig lines based on a closed-loop breeding-meat pig dual-linkage as described in claim 4, characterized in that, The method for analyzing the significant impact of different paternal lines on the growth performance of commercial pigs, the significant impact of different maternal lines on the reproductive performance of commercial pigs, and the combined superiority of different parent lines involves separate analysis and calculation based on growth trait data and reproductive performance data. This yields the paternal genetic growth contribution value for each paternal line sample, the maternal reproductive contribution value for each maternal line sample, and the combined synergistic gain value of different parent lines. The specific steps include: Growth weighting factors are set according to the importance of each growth trait index data for the reproductive development and breed matching of commercial meat pigs. The growth stability index is calculated by combining all growth weighting factors to construct the paternal genetic growth contribution value corresponding to each paternal breed sample. Based on the balancing scale of the economic value of commercial pig breeding according to the breeding performance data, breeding weight factors are set, and breeding grade scores are calculated by combining all breeding weight factors to construct the maternal breeding contribution value corresponding to each maternal breed sample. Using each paternal and maternal strain sample as a statistical unit, the settlement performance data of all commercial pigs bred by the statistical unit is summarized. The actual combination superiority value of the strain is calculated through the settlement performance data, and the theoretical combination superiority value is calculated by combining the paternal genetic growth contribution value and the maternal reproductive contribution value. The combination synergistic gain value of different parent strains is obtained by subtracting the actual combination superiority value from the theoretical combination superiority value.

6. The method for screening superior breeding pig lines based on a closed-loop breeding-meat pig dual-linkage as described in claim 1, characterized in that, S04 specifically includes the following steps: Obtain the market demand for individual commercial pigs in different sales regions, and at the same time obtain the supply update steps of breeding pigs to meet the corresponding market demand in different sales regions; Based on the various market research indicators for pork and the time-varying thresholds and settlement emphasis ratios corresponding to each indicator within the preset breeding period, a regional market target supply and demand vector for individual commercial pork is established in time series. Construct a time-slot sliding window, adjust the time-slot sliding window on the regional market target supply and demand vector according to the dynamic granularity of the supply update step, and calculate the implicit dependency relationship between different pork market inspection indicators within the window to generate an indicator coupling correlation matrix. Meanwhile, within the time-slot sliding window, the evaluation feedback scale of each index item in the comprehensive advantage model is adjusted based on the time-varying limit threshold, and the attention of each pork market inspection indicator and the corresponding different supporting advantage indicators are calculated based on the settlement emphasis ratio. The factor loading matrix of the factor analysis algorithm is adaptively and dynamically adjusted according to the weighting factors of the comprehensive advantage model corresponding to the supporting advantages of the indicator coupling correlation matrix, evaluation feedback scale and attention on the impact of the indicators on the pork market investigation indicators. Finally, a comprehensive advantage index ranking table matching each advantage matching matrix with the corresponding market demand in different sales regions is generated. Calculate the fit degree between each type of advantageous matching matrix and the target supply and demand vector of the regional market. Only extract candidate advantageous matching matrices with a fit degree greater than the preset fit degree threshold. Based on the comprehensive advantage index ranking table, select and recommend regional breeds for candidate advantageous matching matrices to obtain the optimal breed matching results for breeding pigs in different sales regions.

7. The method for screening superior breeding pig lines based on a closed-loop breeding-meat pig dual-linkage as described in claim 1, characterized in that, S05 specifically includes the following steps: The historical settlement data of commercial pigs after the entire process of breeding, raising and marketing in breeding farms, commercial farms and fattening farms are obtained by production and breeding logs, as well as the status characteristics of multiple matching combinations of the parent breeds of the selected historical advantage matching matrix in the preset past time period. Gaussian algorithm is introduced and combined with the biological theory of breeding pigs to construct a nonlinear breeding state space. The state characteristics of each breeding combination are encoded in multiple dimensions through the nonlinear breeding state space to obtain the parent breeding joint state sequence. Based on historical settlement performance data, a historical settlement trajectory matrix of individual commercial pigs is established. Based on the breeding pig-commercial pig breeding income standard benchmark, the income performance base of historical settlement is defined. By analyzing the global income of the historical advantage matching matrix through the income performance base in conjunction with the parent matching joint state sequence, the performance reward function of historical settlement income is defined. By injecting the historical settlement trajectory matrix into the supporting learning model based on maximum entropy inverse reinforcement, the performance reward function is used to maximize the solution of the supporting parameters corresponding to the settlement income fluctuation amplitude exceeding the preset fluctuation amplitude threshold, and the dynamic supporting coefficients that determine the historical settlement performance are output. The pre-trained and deployed pig breed advantage matching prediction model architecture is stripped away from the platform. At the same time, the actual reproductive performance data and real settlement performance data of different parent breeds corresponding to each advantage matching matrix are obtained. Based on the dynamic matching coefficient, the actual reproductive performance data and real settlement performance data are calculated in the whole process to obtain the predicted comprehensive advantage index. Obtain the actual comprehensive advantage index when the actual breeding performance data and the actual settlement performance data of different parent lines are output, calculate the deviation of the predicted comprehensive advantage index from the actual comprehensive advantage index, and obtain the matching deviation value; If the matching deviation is greater than the preset threshold, the matching prediction deviation of the paternal effect weight and the evaluation coefficient of each indicator will be corrected in reverse based on the matching deviation value, thus completing the closed-loop verification of the matching of breeding pig breeds.

Images