Intelligent breeding methods, systems, equipment, media, and program products for breeding pigs.

By constructing a unified database and conducting multi-dimensional analysis, the mismatch between boar semen recommendations and regional sow populations and market demand was resolved, achieving accuracy in semen recommendations and optimizing profitability, thereby improving the breeding efficiency and profitability of large-scale pig farming groups.

CN123118147APending Publication Date: 2026-11-17HENAN MUYUAN INTELLIGENT TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611405206.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-10
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

When large-scale pig farming groups establish breeding pig farms and commercial pig farms across regions, the recommended boar semen does not match the quality characteristics of regional sow populations and end-market demands. The selection of breeding lines lacks actual profit data analysis, resulting in the failure to fully realize genetic potential and improve the input-output ratio.

Method used

By collecting data from multiple sources, constructing a unified database, cleaning and labeling the data, constructing sow quality profiles and market user profiles, calculating comprehensive recommendation scores based on profit data analysis, screening candidate semen and outputting semen recommendation schemes, and optimizing sow breeding and boar breeding decisions.

Benefits of technology

This approach achieves precise matching between boar semen and actual regional needs, improves the matching accuracy of semen recommendations, shortens the identification and elimination cycle of low-efficiency breeding lines, and enhances regional profitability and breeding effectiveness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118147A_ABST
    Figure CN123118147A_ABST
Patent Text Reader

Abstract

This application discloses a method, system, equipment, medium, and program product for intelligent breeding pig mating, belonging to the field of smart farming technology. The method includes: collecting multi-source raw data, including breeding, production, semen inventory, farming input-output, market, and environmental data, and constructing a unified database after processing; conducting multi-dimensional quality evaluation of sow populations in each region to construct a regional sow quality profile; extracting terminal market preference indicators for each region to construct a market user profile vector; conducting multi-dimensional analysis of breeding line profitability data for each region and outputting the optimal breeding line recommendation scheme; upon receiving mating requests from a target region, screening candidate semen based on the above profile, vector, and scheme, calculating a comprehensive recommendation score, and outputting semen recommendation schemes from high to low scores, so that semen recommendations simultaneously match regional sow population characteristics, terminal market preferences, and the optimal breeding line profitability target, thereby improving the matching degree between semen recommendations and actual regional needs.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of smart breeding technology, specifically to a method, system, equipment, medium, and program product for intelligent selection and mating of breeding pigs. Background Technology

[0002] In the context of large-scale pig farming groups establishing breeding pig farms and commercial pig farms across regions, in order to rationally allocate boar semen, select and promote breeding lines in multiple regions, and arrange sow breeding and boar breeding resources accordingly, it is usually necessary to combine the sow population status, end-market demand, and actual operating profit data in different regions to improve regional mating results and the competitiveness of commercial pigs in the market.

[0003] In related technologies, the recommended allocation of boar semen is usually based primarily on the ranking of individual boar breeding indices and semen inventory status, supplemented by inbreeding judgment based on pedigree records; the selection of breeding lines typically relies on the experience assessment of breeders or comparison of single breeding indices; sow selection and boar breeding are generally conducted independently based on reproductive performance indices and growth or meat quality trait-related indices, respectively. These methods are mainly based on individual indices or experience-based judgments, failing to fully consider the quality structure of the regional sow population, the differences in regional end-market demand for commercial pig products, and the actual profitability of each breeding line in its respective region. Furthermore, there is no effective data linkage between the selection and boar breeding stages, making it difficult for related decisions to correspond to actual regional conditions and operational results.

[0004] Therefore, when cross-regional precise breeding and optimization of breeding lines are required, problems such as mismatch between the recommended boar semen results and the actual breeding needs of the region, difficulty in objectively identifying the profitability of breeding lines, and lack of coordination between the direction of sow breeding and the direction of boar breeding may arise, which restricts the realization of genetic potential and the improvement of input-output ratio. Summary of the Invention

[0005] In view of this, this application aims to provide a method, system, equipment, medium and program product for intelligent breeding of boars, in order to solve the technical problem of how to achieve accurate matching of boar semen with the actual needs of various regions.

[0006] In a first aspect, this invention provides a method for intelligent selection and mating of breeding pigs, comprising: collecting multi-source raw data, including breeding data, production data, semen inventory data, livestock input-output observation data, market data, and environmental data; performing data cleaning, individual identifier association, regional tagging, and time alignment processing on the multi-source raw data to construct a unified database; based on the breeding data and production data in the database, performing multi-dimensional quality evaluation on the sow population in each region to construct a regional sow quality profile; extracting preference indicators from the terminal market in each region based on the market data in the database to construct a market user profile vector; performing multi-dimensional analysis on the profitability data of breeding lines in each region based on the livestock input-output observation data in the database to output the optimal breeding line recommendation scheme for each region; and upon receiving mating requests from a target region, screening candidate semen based on the regional sow quality profile, the market user profile vector, and the optimal breeding line recommendation scheme, calculating a comprehensive recommendation score for each candidate semen, and outputting a semen recommendation scheme in descending order of comprehensive recommendation score.

[0007] In one embodiment, a multi-dimensional quality evaluation is performed on the sow population in each region, including: calculating the comprehensive quality score of individual sows: weighted normalization is performed on reproductive performance indicators, growth performance indicators, body shape and appearance indicators, and health status indicators to obtain reproductive performance score, growth performance score, body shape and appearance score, and health status score; the reproductive performance score, growth performance score, body shape and appearance score, and health status score are multiplied by their respective preset weights and then summed to obtain the individual sow quality score; calculating the regional quality index: the sum of the comprehensive quality scores of all sows in the region is divided by the total number of sows in the region to obtain the regional quality index, which is used to characterize the average quality level of the sow population in the region; calculating the regional quality dispersion: calculating the regional... The standard deviation of the quality scores of all individual sows within the region is divided by the regional quality index to obtain the regional quality dispersion, which is used to characterize the uniformity of sow quality in the region. A regional breed structure vector is constructed by statistically analyzing the proportion of sows of each breed in the region, generating a regional breed structure vector to characterize the breed composition characteristics of the region. A regional parity distribution vector is constructed by statistically analyzing the proportion of sows of each parity in the region, generating a regional parity distribution vector to characterize the parity structure characteristics of the region. A regional breeding index mean vector is constructed based on the population mean corresponding to each breeding index dimension of the regional sow population, and the population standard deviation corresponding to each breeding index dimension is recorded to characterize the breeding index distribution characteristics of the regional sow population.

[0008] In one embodiment, constructing a market user profile vector includes: extracting lean meat percentage preference index, meat quality grade preference index, growth rate preference index, body shape and appearance preference index, and weight specification preference index from the regional terminal market; normalizing each market preference index to a unified numerical range to form a market user profile vector; and calculating the score of each preference index based on the price difference of different grades of products in the regional market.

[0009] In one embodiment, a multi-dimensional analysis of the profitability data of breeding systems in various regions is performed based on aquaculture input-output observation data to output the optimal breeding system recommendation scheme for each region. This includes: for each breeding system in each region, a weighted summation of three dimensions—profitability level, profitability stability, and profitability growth trend—is used to calculate the profitability score of the breeding system; wherein, the profitability level is based on the average gross profit per head of the corresponding breeding system in the region, and is obtained by normalizing the maximum and minimum average gross profit per head of all breeding systems in the region; the profitability stability is calculated based on the coefficient of variation of average gross profit per head of multiple batches; and the profitability growth trend is based on the average gross profit per head of multiple batches. The slope of profit change is obtained through nonlinear mapping; cross-regional weighted aggregation calculation is performed on the profitability scores of the same breeding system in different regions to obtain the group's comprehensive profitability score for the breeding system; the weight of each region participating in the weighted aggregation is determined according to the corresponding region's output or profit contribution ratio; the profitability ranking of the breeding systems is obtained by sorting them from high to low according to the group's comprehensive profitability score; the breeding system-regional fit score is calculated by combining the profitability of the breeding system, the matching degree of sow quality, and the matching degree of market users, and the optimal breeding system and alternative breeding system schemes for the corresponding region are output according to the fit score.

[0010] In one embodiment, the process of screening candidate semen includes: screening candidate semen that is available in stock, within its expiration date, and meets the semen quality standards; performing inbreeding constraint filtering: filtering candidate semen whose expected offspring inbreeding coefficient exceeds a preset threshold based on pedigree calculation; performing breeding line matching filtering: screening boar semen that is compatible with the breed and strain of the target area and the optimal breeding line; and performing regional adaptation filtering: screening boar semen that is compatible with the sow breed structure of the target area.

[0011] In one embodiment, the comprehensive recommendation score for each candidate semen sample is obtained by weighted summation of multiple sub-scores, including breeding index regional fit score, semen quality comprehensive score, sow quality fit score, market user matching score, breeding line contribution score, and semen freshness score; wherein: breeding index regional fit score: calculated using Gaussian kernel function to determine the degree of fit between the boar's various breeding indices and the breeding index distribution of the sow population in the target area; the closer the boar's index is to the regional sow population index distribution, the higher the fit score; semen quality comprehensive score: calculated by weighted summation based on continuous detection parameters of sperm motility, sperm concentration, and normal sperm morphology rate; sow quality fit score: obtained by comprehensively considering breed fit, parity fit, and sow quality complementarity, characterizing the degree of matching between the candidate boar and the quality characteristics of the sow population in the target area; market user matching score: calculated by obtaining multiple candidate boars... The genetic performance score for each product feature dimension is obtained by mapping the estimated breeding value or breeding index of the candidate boar; the market user preference score for each product feature dimension in the target region is obtained; the deviation between the boar's genetic performance score and the regional market user preference score for each product feature dimension is calculated, and the average deviation for all product feature dimensions is calculated to determine the market user matching score based on the average deviation; the breeding line contribution score is calculated as follows: the optimal breeding line corresponding to the target region is obtained, and the corresponding role score is assigned according to the hybridization role played by the candidate boar in the hybridization architecture of the optimal breeding line to obtain the breeding line contribution score; the more core the hybridization role of the candidate boar in the optimal breeding line, the higher the breeding line contribution score; the semen freshness score is calculated using an exponential decay method based on the remaining effective days of semen, and the shorter the remaining effective period of semen, the lower the freshness score.

[0012] In one embodiment, when outputting a semen recommendation scheme, a predetermined number of candidate semen samples ranked first are output, along with detailed sub-scores, inbreeding coefficients, and mating system role information for each candidate semen sample, and a complete recommendation decision log is saved.

[0013] In one embodiment, the method further includes sow breeding decision-making and boar breeding planning, wherein: sow breeding decision-making includes: using the optimal breeding line in the target area as the optimization target, calculating the sow breeding priority score by comprehensively considering the individual quality of the sow, the role suitability of the breeding line, and the profit potential of the offspring, and outputting a list of recommended sows for retention and culling based on the breeding priority score; boar breeding planning includes: predicting the semen demand of different breeds of boars based on the regional sow inventory, mating plan, and optimal breeding line; calculating the boar breeding priority score by comprehensively considering the boar breeding index, the role of the breeding line, the urgency of semen demand, and the contribution of population genetic diversity, and outputting a list of boar breeding priorities; the contribution of population genetic diversity is calculated based on the kinship between the candidate boar and the existing breeding pig population, and the more distant the kinship between the candidate boar and the existing breeding pig population, the higher the contribution of genetic diversity.

[0014] In one embodiment, the method further includes: effect tracking and iterative optimization: tracking production performance data after mating execution, calculating the deviation between the actual profit and expected profit of the breeding line; when the actual profit of a certain breeding line in the corresponding region is continuously negative relative to the expected profit, automatically adjusting the profit score weight of the breeding line in the region and reducing the recommendation priority of the breeding line; based on historical recommendation effect data, using the Bayesian optimization method to adaptively adjust the weight of each item in the comprehensive recommendation score; using the maximization of the average gross profit per head after the implementation of the recommendation scheme as the objective function, constraining the value of each weight to be in the range of [0,1] and the sum of all weights to be 1, searching for the optimal weight combination; recalculating and updating the sow quality profile of each region according to a preset period.

[0015] A second aspect of this invention provides an intelligent breeding pig mating system, comprising: a data acquisition layer for acquiring raw data from a multi-source system, including breeding data, production data, semen inventory data, livestock input-output observation data, market data, and environmental data; a data governance layer for cleaning, standardizing, and integrating the acquired multi-source data, including a data cleaning module, an individual identifier association module, a regional tagging module, and a time alignment module; an intelligent analysis layer including a regional sow quality evaluation module, a market user profiling module, a profit data analysis module, and a boar semen precision recommendation engine; and a decision engine layer for converting the calculation results of the intelligent analysis layer into executable decision schemes, including scheme generation, scheme approval, scheme issuance, and effect tracking functions.

[0016] In one embodiment, the individual identification association module of the data governance layer associates breeding data, production data, semen data and pedigree data to the individual level through ear tags to construct a full life cycle data archive for each pig; the regional label module labels each pig and each production record with regional, farm area and breeding line type labels; and the time alignment module aligns breeding events, production events and financial events by monthly and batch dimensions to construct a time series dataset.

[0017] In one embodiment, the intelligent analysis layer further includes: a sow breeding decision-making module, a boar breeding planning module, and a breeding system optimization module.

[0018] In one embodiment, the system further includes an application terminal layer, comprising a management terminal and a mobile terminal, for displaying regional sow quality profiles, breeding line profitability rankings, semen recommendation schemes, sow breeding suggestions, and boar breeding plans.

[0019] A third aspect of the present invention provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the at least one processor to perform the intelligent breeding pig selection method provided in any implementation of the first aspect.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the intelligent breeding pig selection method provided in any implementation of the first aspect.

[0021] A fifth aspect of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the intelligent breeding pig selection method provided in any implementation of the first aspect.

[0022] The intelligent breeding method for breeding pigs provided by this invention collects breeding data, production data, semen inventory data, livestock input-output observation data, market data, and environmental data, and performs data cleaning, individual identifier association, regional tagging, and time alignment processing to construct a unified database. This solves the problem of fragmented multi-source data that prevents comprehensive consideration of regional sow quality, market demand, and livestock profitability. The method constructs regional sow quality profiles based on breeding and production data, constructs market user profile vectors based on market data, and outputs optimal breeding line recommendations for each region based on livestock input-output observation data. Upon receiving breeding requests from a target region, candidate semen is first screened, then a comprehensive recommendation score for each candidate semen is calculated, and semen recommendations are output in descending order of score. This ensures that semen recommendations simultaneously match regional sow population characteristics, end-market preferences, and optimal breeding line profitability targets, thereby improving the matching degree between semen recommendations and actual regional needs. This avoids the problem of poor regional adaptability caused by recommendations based solely on boar index ranking and shortens the identification and elimination cycle for low-efficiency breeding lines. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the specific embodiments or related technologies of this application, the drawings used in the description of the specific embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the intelligent breeding method for pigs provided in this application embodiment.

[0025] Figure 2 A schematic diagram of an intelligent breeding pig selection system is provided for an embodiment of this application.

[0026] Figure 3 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] After large-scale pig farming groups establish breeding and commercial pig farms across multiple regions, they need to rationally allocate boar semen to each region for breeding. However, current technologies primarily rely on boar breeding index rankings and inventory status for semen recommendations, failing to consider differences in sow quality characteristics and end-market demands across different regions. Breeding line selection depends on experience or a single breeding index, lacking systematic analysis based on actual profitability data. Furthermore, breeding data, production data, and pig farming input-output observation data for each region are stored in separate systems, lacking a unified data integration platform. Research has revealed that the fundamental reason for the mismatch between semen recommendations and actual regional needs lies in the fact that recommendation decisions do not incorporate multi-dimensional information such as regional sow quality characteristics, end-market preferences, and the actual profitability of breeding lines.

[0029] Exemplary methods To address the problems existing in the prior art, this application provides an intelligent breeding pig selection method. This method can be executed by a server, a data platform for a breeding enterprise, a dedicated analysis platform, or an electronic device including a processor and memory. Figure 1 As shown, the method specifically includes: S101, Collect multi-source raw data, including: breeding data, production data, semen inventory data, aquaculture input-output observation data, market data, and environmental data.

[0030] In this application, multi-source raw data refers to a data set from two or more business systems, monitoring systems, or manual input terminals. It can be structured data, such as numerical breeding indices, profit figures, and inventory status, or semi-structured data, such as batch ledgers, grading records, and survey forms.

[0031] Specifically, breeding data can come from a breeding management system, including pedigree data of breeding pigs, breeding indices (paternal index, maternal index, meat quality index, growth index, reproductive index, etc.), and EBV (Estimated Breeding Value); production data can come from a pig farm production management system, including individual sow files (breed, parity, body condition score, breeding records), mating records, farrowing records, and piglet growth data for each region; semen inventory data can come from a semen management system, including boar semen collection records, quality testing data output from the CASA (Computer-Assisted Semen Analysis) system, and inventory status (batch number, storage location, expiration date, number of available portions); and livestock input-output observation data can come from ERP (Enterprise Resource Planning) systems. Enterprise Resource Planning (ERP) systems or financial management systems can be used to collect data on feeding costs, feed costs, veterinary drug costs, labor costs, sales revenue, and gross profit. Market data can come from market information systems and may include hog prices, pork price grading, and consumer preference survey data. Environmental data can come from environmental monitoring systems and may include temperature, humidity, and feeding management level scores. This approach unifies heterogeneous data previously scattered across different business systems into subsequent analysis, reducing recommendation bias caused by a single data source.

[0032] In some embodiments, the aforementioned data can be collected daily, weekly, monthly, in batches, or event-triggered; it can also be collected through interface calls, database synchronization, file import, or message queues. More generally, the source forms of multi-source raw data are not limited to those listed above; any data that reflects at least some information related to breeding, production, inventory, finance, market, and environment can be used as input for subsequent processing. This open definition can cover deployment methods with different levels of information technology infrastructure in group-based farming scenarios.

[0033] S102, perform data cleaning, individual identifier association, regional label marking, and time alignment processing on the multi-source raw data to construct a unified database.

[0034] Specifically, in the data cleaning stage, missing values, outliers, and duplicate records can be handled, and data formats and coding standards can be standardized. For example, if the ear tag format of the same boar is inconsistent in different systems, the ear tags can be uniformly converted to a preset length and coding rule. Records of vitality values, concentration values, litter size, or gross profit that exceed physiologically reasonable ranges can be marked as outliers and processed through removal, correction, or imputation. This approach helps reduce the interference of dirty data on subsequent scoring models.

[0035] In some embodiments, individual identifier association can link breeding data, production data, semen data, and pedigree data to the individual level via ear tags, forming a full lifecycle data profile for each pig. For example, a sow's reproductive record in the production management system can be mapped to its breeding index in the breeding management system and its family relationship in the pedigree database; a boar's semen batch quality data and inventory data can also be aggregated around its ear tag. Regional labeling can label each pig and each production record with regional, farm, and breeding type tags to support cross-regional statistics and local recommendations. Time alignment can align breeding events, production events, and financial events according to monthly and batch dimensions to construct a time-series dataset. For example, the mating time, farrowing time, fattening and slaughter time, and final profit record of a batch can be aligned to the same analysis window. It should be noted that the aforementioned individual identifier association provides a unified object basis for regional labeling and time alignment, while time alignment provides comparable time-series data conditions for subsequent profit trend analysis, recommendation effect tracking, and weight optimization, making the unified database computable and traceable.

[0036] S103, based on breeding and production data in the database, conducts multi-dimensional quality evaluation of sow populations in each region and constructs a regional sow quality profile.

[0037] In one embodiment, this profile can describe a regional sow population from multiple levels, including individual quality, population mean, population dispersion, breed composition, parity structure, and breeding index distribution. Unlike using a single index of only one sow, this regional profile reflects the population characteristics at the regional level, and is therefore more suitable as input conditions for regional semen recommendations.

[0038] As a specific implementation method, the first step is to calculate the comprehensive quality score of individual sows: Reproductive performance indicators, growth performance indicators, body conformation indicators, and health status indicators are weighted and normalized to obtain reproductive performance scores, growth performance scores, body conformation scores, and health status scores. These scores are then multiplied by their respective preset weights and summed to obtain the individual sow quality score. Through six dimensions—individual quality comprehensive score, regional quality index, regional quality dispersion, regional breed structure vector, regional parity distribution vector, and regional breeding index mean vector—the regional sow quality characteristics are comprehensively characterized from the individual level to the group level. This provides refined group profile data support for the subsequent calculation of sow quality matching scores and regional matching scores for semen recommendation.

[0039] Reproductive performance indicators can include one or more of the following: total litter size, number of live piglets, weaning litter weight, and conception rate. These indicators directly reflect the sow's reproductive capacity and are among the most heavily weighted dimensions in sow quality evaluation. Their quantitative scoring allows the system to identify individual sows and regional groups with excellent reproductive performance, providing a basis for quality-complementary matching in semen recommendations. Growth performance indicators can include one or more of the following: age at 100 kg, daily weight gain, and feed conversion efficiency. These indicators reflect the growth potential and feed utilization efficiency of the sow's offspring. Including them in quality evaluation helps assess the sow's genetic contribution to the growth performance of commercial pigs, enabling recommendations to comprehensively optimize both reproductive and growth performance. Body conformation indicators can include one or more of the following: body length, body height, backfat thickness, and leg and rump scores. These indicators reflect the sow's body structure and condition. Sows with good body conformation typically have a longer service life and greater reproductive sustainability. Including them in quality evaluation helps assess the sow's long-term productive value. Health status indicators can include one or more of the following: limb and hoof scores, disease records, and immune status. These indicators reflect the sow's health level and disease resistance. Sows in good health have higher breeding success rates and lower culling risks. Incorporating these indicators into quality assessment helps identify individuals with higher health risks and improves the comprehensiveness of the assessment.

[0040] Weighted normalization can be achieved by mapping each indicator value to the [0,1] interval and then summing them by weight. Weighted normalization eliminates differences in units and numerical ranges between different indicators, allowing them to be weighted and combined on a uniform numerical scale, ensuring the comparability and rationality of quality scores. The individual quality score of a sow can be calculated using the following formula: (1) in, Classified by reproductive performance, Classified by growth performance, Based on body shape and appearance, Classified by health status; , , , These are the weights for the corresponding dimensions, and the sum of the weights is 1. By using a weighted summation method, the scores of the four dimensions are combined into a single individual quality score. This allows the system to measure the overall quality level of each sow with a unified numerical indicator, providing a quantifiable and rankable individual evaluation basis for subsequent regional herd quality evaluation and sow breeding decisions.

[0041] For example, the preset weight can be set to , , , Among these factors, reproductive performance has the highest weight, reflecting the quality evaluation orientation of sows with reproductive performance as the core. Setting the weight of reproductive performance to the highest value is consistent with the production positioning of sows as the core of the breeding group, ensuring that sows with outstanding reproductive capacity receive higher comprehensive evaluations in the quality score; growth performance, body shape and appearance, and health status are used as auxiliary dimensions to ensure the comprehensiveness of the quality evaluation.

[0042] After obtaining the comprehensive quality scores of each sow within the region, the regional quality index can be further calculated. The regional quality index is obtained by summing the comprehensive quality scores of all sows within the region and dividing by the total number of sows in that region. This index characterizes the average quality level of the sow population within the region. The calculation of the regional quality index can be expressed as follows: (2) Where N is the total number of sows in the region. The i-th sow is given an overall individual quality score. The regional quality index is calculated by averaging the individual quality scores of all sows within the region, aggregating individual-level quality evaluations into a group-level quality indicator. This allows the system to compare the average quality level of sow groups in different regions, providing a quantitative basis for calculating sow quality matching in breeding system-regional fit analysis.

[0043] Based on the regional quality index, the regional quality dispersion is further calculated. The standard deviation of the quality scores of all individual sows within the region is calculated, and then divided by the regional quality index to obtain the regional quality dispersion, which characterizes the uniformity of sow quality within the region. The calculation of regional quality dispersion can be expressed as follows: (3) in, RQD represents the standard deviation of individual sow quality scores in the region, while RQI is the regional quality index for that region. A smaller RQD indicates higher uniformity of sow quality. Regional quality dispersion is measured using the coefficient of variation (CV). Regions with lower quality dispersion indicate a more uniform sow population quality, suitable for a uniform breeding strategy. Regions with higher quality dispersion require differentiated semen recommendation programs for sows of different quality levels to improve the targeting of the recommendation strategy.

[0044] In addition, it is necessary to construct a regional breed structure vector. This involves statistically analyzing the proportion of each breed of sow in the region to generate a regional breed structure vector, which characterizes the breed composition features of the region. For example, a regional breed structure vector can be represented as... ,in Let represent the proportion of sows of breed i in this region. The regional breed structure vector describes the composition ratio of sows of each breed in the region in vector form. It provides structured breed information for breed matching score calculation and regional matching filtering in semen recommendation, enabling the system to recommend suitable boar breed semen based on the regional sow breed composition.

[0045] Similar to the breed structure vector, a regional parity distribution vector also needs to be constructed. The proportion of sows of each parity in a region is statistically analyzed to generate a regional parity distribution vector, which characterizes the parity structure of the region. For example, a regional parity distribution vector can be represented as: ,in The percentage of sows of parity i in the region is used. In some embodiments, the parity range can be 1 to 6 parities; in other embodiments, it can be extended to 1 to 8 parities or divided into three categories: primiparous, multiparous, and advanced gestational age. Using a vectorized representation provides direct data input for subsequent breed and parity matching calculations.

[0046] The parity distribution vector of a region reflects the parity structure characteristics of a sow population in that region. Sows of different parities have different reproductive performance and mating requirements (for example, regions with a higher proportion of primiparous sows may need boars with higher semen quality stability). Incorporating parity distribution into the quality profile helps to improve the calculation accuracy of parity suitability score in semen recommendation.

[0047] Finally, a regional breeding index mean vector is constructed. This vector is built based on the population mean corresponding to each breeding index dimension of the regional sow population, and the population standard deviation corresponding to each breeding index dimension is recorded to characterize the breeding index distribution characteristics of the regional sow population. The regional breeding index mean vector can be represented, for example, as follows: ,in This represents the mean of the breeding index in the k-th dimension for the sow population in region r (the dimension can include paternal index, maternal index, meat quality index, growth index, reproductive index, etc.), and also records the population standard deviation σk corresponding to each breeding index dimension. The regional breeding index mean vector and its corresponding standard deviation describe the distribution characteristics of the regional sow population in each breeding index dimension. This provides mean and standard deviation parameters for subsequent calculation of the fit between the boar breeding index and the regional sow population index distribution using the Gaussian kernel function, enabling the calculation of the regional fit score of the breeding index to assess the degree of fit from the perspective of population distribution.

[0048] It should be noted that the regional breed structure vector, the regional parity distribution vector, and the regional breeding index distribution are not independent of each other: the former two reflect the compositional characteristics of the sow population, while the latter reflects the genetic performance distribution characteristics of the sow population. The three together constitute a regional sow quality profile, so that subsequent semen matching no longer depends on a single average value, but is based on the joint expression of population structure and distribution characteristics.

[0049] S104: Based on market data in the database, extract preference indicators for terminal markets in various regions and construct market user profile vectors.

[0050] Specifically, the market user profile vector is used to quantify the preference characteristics of hog products in the terminal market of each region, transforming qualitative market preferences into quantitative numerical vectors. This allows market demand information to directly participate in the calculation of comprehensive recommendation scores, realizing a market-oriented breeding decision-making mechanism and avoiding the problem of breeding direction being out of touch with terminal market demand.

[0051] In one embodiment, when constructing the market user profile vector, the following preferences are first extracted for the regional terminal market: lean meat percentage preference, meat quality grade preference, growth rate preference, body shape and appearance preference, and weight specification preference. Specifically, as shown in Table 1, the following preference indicators for each region are extracted from the market information system: Table 1 The lean meat percentage preference index is used to characterize the end market's requirement for lean meat percentage in live pigs. Lean meat percentage is one of the key indicators affecting the settlement price of live pig slaughter. Quantifying lean meat percentage preference into an index allows for prioritizing the matching of boars with genetic lean meat percentage performance consistent with the regional market's lean meat percentage preference when recommending semen, thereby enhancing the price competitiveness of commercial pigs in that regional market. The meat quality grade preference index is used to characterize the end market's price sensitivity to pork quality grade. Different regional markets have different price sensitivities to meat quality grade. Quantifying meat quality grade preference allows for the identification of regional markets sensitive to meat quality premiums and recommends semen from boars with excellent genetic performance in meat quality, thereby enhancing the meat quality grade premium revenue of commercial pigs. The growth rate preference index is used to characterize the end market's requirement for the age of live pigs at slaughter. Growth rate preference reflects the market's expectation of the slaughter cycle. Quantifying it allows the system to recommend semen from boars with outstanding genetic performance in growth rate for regions that prefer rapid slaughter, shortening the slaughter cycle of commercial pigs and improving capital turnover efficiency. The body shape and appearance preference index is used to characterize the end market's preference for pig body shape characteristics. Some regional markets have specific preferences for pig body shape (such as body length and height). By quantifying these preferences, the system can match boar semen with genetic traits consistent with market preferences, improving the market acceptance of commercial pigs' appearance. Weight specification preference indicators characterize the range of slaughter weights preferred by the end market. Different regional markets have different preferences for slaughter weight specifications. Quantifying these preferences allows for the recommendation of boar semen that helps commercial pigs reach their target slaughter weight, reducing price discounts due to mismatched weight specifications.

[0052] After extracting the aforementioned preference indicators, each market preference indicator is normalized to a uniform numerical range, for example, it can be normalized to... This interval forms the market user profile vector. The market user profile vector can be represented as follows: (4) in, Based on lean meat percentage preference, Based on meat quality preference, Based on growth rate preference, Based on body shape and appearance preferences, Weight and size preference scores are assigned. After normalizing each preference index to a unified numerical range, they are constructed into a vector form, eliminating differences in dimensions and numerical ranges between different preference indices. This allows the market user profile vector to be directly used for dimension-by-dimensional deviation calculation with the boar genetic performance score, providing a standardized input for calculating the market user matching score.

[0053] The scores for each preference indicator are calculated by weighting the price differences between different grades of products in the regional market. For example, the lean meat percentage preference score... The calculation can be expressed as follows: (5) in, This is the average price of pigs with a high lean meat percentage. This represents the average price of pigs with a low lean meat percentage. This represents the average price of live pigs in the region. The higher the price premium for pigs with a higher lean meat percentage, the better. The closer the score is to 1, the more sensitive the market in that region is to lean meat percentage. Preference scores are calculated based on price differences, directly transforming actual market price signals into quantitative indicators of preference strength. This ensures that preference scores accurately reflect the market's economic value assessment of different product characteristics, avoiding biases caused by subjective evaluations. Feature dimensions with larger price premiums receive higher preference scores, ensuring that semen recommendations prioritize matching product characteristics with the highest economic value in the market.

[0054] Alternatively, as another approach, the construction of market user profiles can also employ Natural Language Processing (NLP) methods instead of the price differential method. For example, by analyzing text data such as slaughterhouse feedback, consumer reviews, and industry reports, market preference features can be extracted to capture richer dimensions of market preferences (such as brand preference, safety preference, and origin preference). Using NLP methods can uncover market preference dimensions from unstructured text data that are difficult to directly reflect through price data, expanding the information coverage of market user profiles and making them suitable for application scenarios with more diversified market preference dimensions.

[0055] S105, based on the aquaculture input-output observation data in the database, performs multi-dimensional analysis on the profitability data of breeding systems in various regions, and outputs the optimal breeding system recommendation scheme for each region.

[0056] Specifically, the optimal breeding line recommendation scheme is obtained through multi-dimensional analysis of the profitability data of breeding lines in various regions based on the observation data of breeding input and output. A breeding line can be, for example, a system that produces commercial pigs through specific crossbreeding combinations, such as the Duroc × Landrace × Large White three-way crossbreeding system (referred to as "Duroc × Landrace × Large White"), the Duroc × two-way crossbreeding system, etc. The optimal breeding line recommendation scheme is determined based on actual profitability data rather than a single breeding index, which can identify truly high-efficiency breeding line combinations under specific regional conditions. This provides optimization targets for breeding line matching filtering and breeding line contribution score calculation in semen recommendation, ensuring that the recommended semen meets the role requirements of the breeding line.

[0057] Multi-dimensional analysis of profitability data is achieved through three levels: regional profitability scoring, group-level cross-regional weighted aggregate ranking, and comprehensive suitability scoring. This transforms the selection of supporting systems from relying on experience-based judgment to a systematic analysis based on actual profitability data. It allows for comparison of the profitability performance of different supporting systems within the same region at the regional level, and also for comparison of the comprehensive profitability of the same supporting system in different regions at the group level.

[0058] As a specific implementation method, at the regional level, for each supporting system c under each region r, the profitability score of the supporting system can be calculated by weighted summation of three dimensions: profitability level, profitability stability, and profitability growth trend.

[0059] The profitability score is based on the average gross profit per unit of the corresponding supporting system within the region, normalized by combining the maximum and minimum average gross profit per unit of all supporting systems in the region. For example, the profitability score can be calculated as follows: (6) in, For the region Supporting systems Gross profit per head and These represent the highest and lowest gross profit per head among all supporting systems in the region. Profitability scores are mapped to the highest and lowest gross profit per head through regional range normalization. The interval eliminates the impact of differences in absolute profitability between different regions, making the profitability of different supporting systems within the same region comparable, and ensuring that the supporting system with the highest current profitability receives the highest profitability score.

[0060] Profitability stability is calculated based on the coefficient of variation of gross profit per head across multiple batches. For example, the calculation of the profitability stability score can be expressed as follows: (7) Where, σ and These are the supporting systems in the region close The standard deviation and mean of gross profit per head for each batch (e.g., 12 batches). Profitability stability score measures the volatility of a product's profitability based on the coefficient of variation; a smaller coefficient of variation indicates more stable profitability, and a stability score closer to 1. Incorporating profitability stability into the scoring system helps avoid selecting products with drastic profit fluctuations, reducing the risk of significant profit declines due to market or environmental changes.

[0061] The profit growth trend is derived from a non-linear mapping of the slope of the average gross profit per head across multiple batches. For example, the calculation of the profit growth trend score can be expressed as follows: (8) in, For this supporting system in the region close The slope of the linear regression of gross profit per unit of batch (reflecting the profit trend). For adjustment coefficients (e.g.) The unit is / yuan). The slope is mapped to using the Sigmoid function. interval, For the correct time , When negative The profit growth trend score uses a sigmoid function to non-linearly map the slope of profit changes, assigning a trend score higher than 0.5 to portfolios showing an upward profit trend and a trend score lower than 0.5 to portfolios showing a downward profit trend. Incorporating profit trend into the scoring system identifies portfolios with continuously improving profits and increases their recommendation priority, while decreasing the recommendation priority of portfolios with continuously declining profits.

[0062] After obtaining the scores for the above three dimensions, the profitability score of the supporting system is obtained by weighted summation of the scores for profitability level, profitability stability, and profitability growth trend. For example, the calculation of the profitability score of the supporting system can be expressed as follows: (9) in, , , Each dimension has a weight, and the sum of the weights is 1. The profitability score, calculated by weighting the three dimensions, comprehensively considers the current profitability level, profitability volatility risk, and profitability development trend of the supporting system. This ensures that the score fully reflects the profitability of the supporting system and avoids making biased supporting system selection decisions based solely on a single dimension (such as the current profitability level). For example, the weights of the three dimensions can be set as follows: , , Among these factors, profitability has the highest weight. Setting profitability as the highest weight ensures that the most profitable supporting system dominates the scoring. Profitability stability and profit growth trend serve as auxiliary dimensions, respectively adjusting the scoring from the perspectives of risk control and development potential, so that the scoring results consider stability and sustainability while pursuing high profitability.

[0063] At the group level, the profitability scores of supporting systems in different regions are weighted and aggregated across regions to obtain the group's overall profitability score for that supporting system. The weight of each region participating in the weighted aggregation is determined based on the corresponding region's output share. Using output share as the regional weight ensures that regions with larger outputs have a greater weight in the group's overall profitability score, guaranteeing that the ranking of supporting systems at the group level prioritizes reflecting the profitability performance of major production areas, aligning with the decision-making orientation of maximizing overall group benefits. In other embodiments, the weight of each region can also be determined based on the corresponding region's profit contribution share. Using profit contribution share as the regional weight ensures that regions with larger profit contributions have a greater weight in the group's overall profitability score, directly using economic benefits as the guiding principle for weighted aggregation, suitable for scenarios where there are significant differences in profitability levels across different regions.

[0064] The calculation of the group's overall profitability score can be expressed as follows: (10) in, For the region The weights. The system is based on... The profitability ranking of breeding lines is obtained by sorting them from high to low. The overall profitability score of the group aggregates the profitability scores of breeding lines in each region according to the regional weight, so that the system can identify the breeding lines with the best overall profitability from the perspective of the group. This provides a quantitative ranking basis for breeding strategy decision-making and breeding line promotion at the group level.

[0065] Based on regional and group-level analyses, a comprehensive analysis of breeding system profitability, sow quality matching, and market-user matching is used to calculate the breeding system-regional fit score. The optimal and alternative breeding system solutions for each region are then ranked according to these scores. The calculation of the breeding system-regional fit score can be expressed as follows: (11) in, To score the profitability of the supporting system, For sow quality matching, For market user matching, , , Each dimension has a weight, and the sum of the weights is 1. The breeding system-regional suitability score integrates three dimensions—profitability, sow quality matching, and market demand matching—into the scoring framework. This ensures that the selection of the optimal breeding system not only considers profitability but also the degree of matching with the quality of the regional sow population and the degree of fit with the regional market demand, thus ensuring that the recommended breeding system can achieve optimal overall benefits under specific regional conditions.

[0066] sow quality matching The degree of matching between the quality of sows in a region and the requirements of their breeding systems can be calculated, for example, as follows: (12) in, For the region The quality index, For supporting systems The target sow quality index (based on breeding standards for breeding lines). For the entire group The range is considered. Sow quality matching degree quantifies the degree of matching between the regional sow population quality and the breeding line requirements by calculating the normalized deviation between the regional quality index and the target quality index of the breeding line. A higher matching degree is achieved when the regional quality index is closer to the breeding line target, ensuring that the recommended breeding line matches the actual quality level of the regional sow population and avoiding the failure to fully realize the benefits of the breeding line due to quality mismatch.

[0067] Market user matching The degree of matching between the characteristics of supporting products and the needs of the regional market can be calculated as follows: (13) in, For supporting systems In the Performance scores based on product characteristics (such as lean meat percentage score, meat quality score, growth rate score, etc.). For the region In the Market user preference segmentation on the dimensional platform The number of feature dimensions. Market user matching degree quantifies the degree of fit between the product features of the breeding system and market demand by calculating the average absolute deviation between the product feature scores of each product in the breeding system and the regional market preference scores. The higher the matching degree, the more consistent the product features of the breeding system are with market preferences, ensuring that the commercial pigs produced by the recommended breeding system can meet the preference requirements of the regional end market and enhance the market competitiveness and premium ability of the commercial pigs.

[0068] For example, the sum of the weights of each dimension in the matching system-regional adaptability score is 1, which can be set to... , , The system prioritizes profitability, ensuring that breeding stock selection is primarily based on actual profitability. Market user compatibility is weighted second, reflecting a market-oriented decision-making philosophy. Sow quality compatibility is weighted appropriately to ensure the suitability of breeding stock to the regional sow population. The system sorts breeding stock by compatibility score from highest to lowest and outputs the optimal and alternative breeding stock options for the corresponding region. Alternative breeding stock options may include, for example, the top three breeding stock options in terms of compatibility score. Outputting the optimal breeding stock and alternative options provides decision-makers with a primary and alternative solution. This allows for rapid switching to alternative options when the optimal breeding stock cannot be implemented due to inventory, production capacity, or other reasons, improving the flexibility and feasibility of decision-making.

[0069] In other embodiments, profit data analysis can also employ machine learning methods instead of statistical formulas. For example, machine learning models such as gradient boosting decision trees (e.g., XGBoost) or random forests can be used, taking product characteristics, regional characteristics, and market characteristics as inputs, and actual gross profit per head as the output label to train a profit prediction model. Machine learning methods can automatically capture the complex nonlinear relationships between product characteristics, regional characteristics, and market characteristics, making them suitable for scenarios where there are complex interactions between influencing factors, and can further improve the accuracy of profit prediction.

[0070] S106. Upon receiving a breeding request from the target area, based on the sow quality profile of the area, the market user profile vector, and the optimal breeding line recommendation scheme, candidate semen is screened, a comprehensive recommendation score is calculated for each candidate semen, and a semen recommendation scheme is output in descending order of comprehensive recommendation score.

[0071] Specifically, candidate semen screening occurs before the comprehensive score calculation, used to eliminate obviously unsuitable semen batches to reduce invalid calculations and non-compliant recommendations. Candidate semen screening may include one or more of the following: inventory status screening, inbreeding constraint filtering, breeding line matching filtering, and regional adaptation filtering. This multi-level candidate semen screening eliminates unusable, excessively risky, mismatched with breeding lines, or incompatible with the regional sow breed structure before calculating the comprehensive recommendation score, narrowing the candidate pool and improving the efficiency of subsequent score calculations and the validity of the recommendation results. In one embodiment: First, candidate semen is screened based on its availability, expiration date, and meeting quality standards. For example, semen can be retrieved from semen inventory data that is currently available, within its expiration date, and meets quality testing standards. Meeting quality standards generally means that at least one or more quality parameters meet usage criteria, specifically based on motility, concentration, morphological normality rate, contamination test results, or a comprehensive rating. This combined screening based on inventory status, expiration date, and quality ensures that the semen entering the subsequent scoring process has basic usability.

[0072] After screening based on inventory availability, inbreeding constraint filtering is performed. The inbreeding coefficient of the expected offspring between candidate boars and sows is calculated based on pedigree relationships. Candidate semen with an expected offspring inbreeding coefficient exceeding a preset threshold is filtered out. Pedigree relationships can be calculated, for example, based on pedigree data. Inbreeding constraint filtering, by calculating the inbreeding coefficient of the expected offspring and comparing it with a threshold, excludes candidate semen that may lead to excessive inbreeding in the offspring, controlling the inbreeding level of the herd and avoiding decreased offspring productivity and increased risk of genetic defects due to inbreeding decline, thus maintaining the long-term genetic health of the breeding herd. For example, the preset threshold can be set to 6.25%, filtering candidate semen with an expected offspring inbreeding coefficient exceeding 6.25%. A 6.25% inbreeding coefficient threshold is equivalent to the inbreeding level of offspring produced by cousin mating. Setting the threshold at this level effectively controls inbreeding risk while retaining a sufficient number of candidate semen for subsequent scoring and screening, achieving a balance between genetic safety and candidate range.

[0073] After inbreeding constraint filtering, a breeding line matching filter is performed. Boar semen that is compatible with the optimal breeding line in the target region is selected. For example, if the optimal breeding line in the target region is a Duroc × Landrace × Large White three-way cross, then Duroc boar semen is selected for the terminal sire role. Breeding line matching filtering ensures that the recommended boar semen breeds are consistent with the hybridization structure of the optimal breeding line in the target region, enabling mating results to produce commercial pigs that conform to the breeding line design, fully leveraging the hybridization advantages of the breeding line, and avoiding deviations from expected hybridization results due to breed mismatch.

[0074] After the breeding system matching filter, a regional adaptation filter is performed. This filters boar semen that is compatible with the sow breed structure of the target region. For example, boar semen compatible with the main sow breeds can be selected based on the proportion of each breed in the regional breed structure vector. The regional adaptation filter screens semen based on the actual composition of the sow breed structure in the region, ensuring that the recommended boar semen is compatible with the main sow breeds in the region at the breed level, avoiding the recommendation of semen that is incompatible with the regional sow breed structure, and improving the regional relevance and feasibility of the recommendation plan.

[0075] Furthermore, the calculation of the comprehensive recommendation score is explained. The comprehensive recommendation score is obtained by weighted summation of multiple sub-scores, including breeding index regional suitability score, semen quality comprehensive score, sow quality suitability score, market user matching score, breeding line contribution score, and semen freshness score. Through the weighted summation of these six sub-scores, a multi-dimensional comprehensive evaluation system is constructed, covering genetic quality, semen quality, regional suitability, market matching, breeding line role, and timeliness. This allows semen recommendation decisions to simultaneously consider multiple key dimensions, achieving a shift from single-index ranking to multi-dimensional comprehensive evaluation, and ensuring that recommendation results simultaneously consider genetic improvement, production efficiency, and economic benefits.

[0076] The regional fit score of the breeding index is calculated using a Gaussian kernel function to determine the degree of fit between each breeding index of the boar and the distribution of breeding indices of the sow population in the target region. The closer the boar index is to the distribution of the sow population index in the region, the higher the fit score. For example, the regional fit score of the breeding index is calculated using the following formula: (14) in, It traverses various breeding index dimensions (such as paternal index, maternal index, flesh quality index, growth index, and reproductive index). For the boar in the The value of the dimension index, For the region The sow herd was in the first The mean of the dimensional index, For the region The sow herd was in the first The standard deviation of the dimensional index For the first The weighting of the breeding index. The regional fit score based on the Gaussian kernel function assesses the fit of boars based on the regional sow population distribution. When the boar index falls near the regional sow population mean (e.g., ...), the fit is determined by the boar index. Within the specified range, the fit score is close to 1, and the fit score decreases exponentially with the deviation. Unlike traditional inter-individual index complementarity calculation methods, this method uses the regional population index distribution as a reference, which is more in line with the needs of regionalized precise recommendation and can identify boars that best match the genetic characteristics of the regional sow population.

[0077] The comprehensive semen quality score is calculated based on a weighted average of continuously measured parameters including sperm motility, sperm concentration, and the percentage of sperm with normal morphology. For example, the calculation can be expressed as follows: (15) in, Sperm motility was scored. Sperm concentration was scored. Sperm morphology is scored. The comprehensive semen quality score is calculated by weighting continuous parameters detected by the CASA system. Compared with traditional discrete grading (such as A / B / C grades), it can more finely distinguish the quality differences between different semen samples, avoid the information loss caused by discrete grading, and make the quality score more discriminative and accurate.

[0078] Sperm motility score The calculation can be expressed as, for example, as ,in The percentage of sperm that move forward (%) The baseline for a perfect sperm motility score is 80%. Sperm motility scoring is based on the proportion of progressively motile sperm. The score is calculated by comparing this proportion to the baseline and truncating to 1. This ensures that semen with motility levels equal to or exceeding the baseline receives a perfect score, while semen with motility below the baseline receives a proportional score. This approach guarantees an upper limit for high-motility semen while also differentiating between different motility levels. Sperm concentration scoring... The calculation can be expressed as, for example, as ,in Sperm concentration (billion / mL). The maximum score is based on a concentration benchmark (e.g., 300 million / mL). Sperm concentration scoring is calculated by comparing sperm concentration to the maximum score benchmark. Semen with a concentration meeting or exceeding the benchmark receives the maximum score, ensuring that the recommended semen has sufficient sperm concentration to guarantee fertilization success. Sperm morphology scoring... The calculation can be expressed as, for example, as ,in The percentage of normal morphology is %. The morphology score is based on a perfect score (e.g., 85%). The sperm morphology score is based on the percentage of normal morphology. The higher the percentage of normal morphology, the lower the proportion of sperm abnormalities and the stronger the fertilization ability. Including it in the quality score helps to exclude semen with a high percentage of abnormal morphology and improve the success rate of mating.

[0079] For example, the weights of each indicator in the comprehensive semen quality score can be set as follows: , , Setting sperm motility as the highest weight reflects the evaluation orientation that sperm motility is the most critical factor affecting fertilization success rate; sperm concentration and normal morphology rate are weighted equally, serving as auxiliary dimensions to jointly ensure the overall quality level of semen.

[0080] Sow quality matching is obtained by combining breed matching, parity matching, and sow quality complementarity matching, characterizing the degree of matching between candidate boars and the quality characteristics of the sow population in the target area. For example, its calculation can be expressed as follows: (16) in, For variety matching, For parity matching, To complement each other in terms of quality, , , The weights for each sub-item are as follows. The sow quality matching sub-evaluation comprehensively assesses the degree of matching between the candidate boar and the quality characteristics of the target area sow population from three sub-dimensions: breed matching, parity matching, and quality complementarity. This ensures that the recommended semen is not only matched with regional sows at the breed level, but also achieves refined matching at the parity structure and quality complementarity levels.

[0081] Variety matching For example, it can be based on the structural vector of boar breeds and regional sow breeds. The fit is calculated. For example, if the region is dominated by Landrace × Large White sows, then Duroc boars as the terminal sires have the highest breed fit score (e.g., Other breeds are scored based on their role in the breeding system. Breed compatibility scores are determined by the boar breed's role in the breeding system within the region's main sow breed composition, ensuring that the recommended boar breeds are most compatible with the regional sow breeds in crossbreeding combinations, fully leveraging hybrid vigor.

[0082] Parity matching For example, it can be based on the regional parity distribution vector. The compatibility score is calculated based on boar semen characteristics. For example, in areas with a high proportion of primiparous sows, boars with high semen quality stability can be prioritized. Parity compatibility scores incorporate the parity structure characteristics of sows in the region into the compatibility assessment. Sows of different parities differ in reproductive performance and mating needs. By calculating parity compatibility scores, semen recommendations can be differentiated based on the regional parity structure characteristics, improving mating success rates and reproductive outcomes.

[0083] Complementary quality For example, it can be represented as: (17) in, The overall quality score of boars is calculated by weighting the breeding index. This refers to the regional sow quality index. The overall quality score is extremely poor. The quality complementarity score measures the degree of complementarity between the boar and regional sow populations at the quality level by calculating the normalized deviation between the boar's quality score and the regional sow quality index. The highest quality complementarity score is achieved when the boar's quality and the regional sow's quality are moderately complementary, which helps improve the overall quality performance of offspring through crossbreeding complementarity effects.

[0084] For example, the weights of each sub-item in the sow quality fit score can be set as follows: , , Setting breed compatibility as the highest weight reflects the priority of breed matching as a fundamental condition for crossbreeding; quality complementarity is the next highest weight to ensure that boars and regional sows effectively complement each other in terms of quality; parity compatibility is maintained at an appropriate ratio to take into account the impact of regional parity structure on mating results.

[0085] The market user matching score is calculated as follows: Genetic performance scores for multiple product characteristic dimensions of candidate boars are obtained. These scores are mapped from the candidate boar's EBV or breeding index, and may include, for example, lean meat percentage genetic performance scores, meat quality genetic performance scores, and growth rate genetic performance scores. Mapping the candidate boar's EBV or breeding index to genetic performance scores for each product characteristic dimension allows for comparison of the boar's genetic potential with market user preferences on the same product characteristic dimensions, providing a comparable data basis for calculating the market user matching score. Market user preference scores for each product characteristic dimension corresponding to the target region are obtained. The deviation between the boar's genetic performance score and the regional market user preference score for each product characteristic dimension is calculated. The average deviation for all product characteristic dimensions is taken, and the market user matching score is determined based on the average deviation. The calculation of the market user matching score can be expressed as follows: (18) in, For the boar in the Genetic expression scores of product characteristics For the region In the Market user preference segmentation on the dimensional platform The number of feature dimensions is used. The market user matching score quantifies the degree of matching between boar genetic characteristics and market demand by calculating the average absolute deviation between the genetic performance scores of each product feature dimension of the boar and the regional market preference score. The more consistent the boar's genetic characteristics are with market preferences, the higher the matching score, ensuring that the product characteristics of the offspring produced by the recommended semen can meet the preferences and needs of the regional end market, and transforming market price signals into a decision-making basis for semen recommendations.

[0086] The contribution score of the breeding system is calculated as follows: The optimal breeding system corresponding to the target region is obtained. Based on the hybridization role of the candidate boar in the hybridization architecture of this optimal breeding system, a corresponding role score is assigned to obtain the contribution score. The more central the hybridization role of the candidate boar in the optimal breeding system, the higher the contribution score. For example, the calculation can be expressed as follows: (19) in, For the region The optimal matching system. The role of the boar in the breeding system is scored. For example, in the optimal breeding system of Duroc × Landrace × Large White three-way crossbreeding, the role score of the Duroc boar as the terminal sire is 1.0, the role score of the Landrace boar as the maternal sire is 0.8, the role score of the Large White boar as the maternal sire is 0.8, and the role score of other breeds is 0.2. The breeding system contribution score is scored according to the importance of the boar's hybridization role in the optimal breeding system. The more central the hybridization role of the boar, the higher the contribution score, ensuring that the recommended semen prioritizes the needs of the most critical role in the breeding system, so that the mating results can fully realize the hybrid vigor and profit potential of the breeding system.

[0087] The semen freshness score uses an exponential decay method, calculated based on the remaining effective days of the semen. The shorter the remaining effective period of the semen, the lower the freshness score. For example, the calculation can be expressed as follows: (20) in, This refers to the remaining effective days of semen. This refers to the total shelf life of semen in days (e.g., 5 days for fresh semen by default). The attenuation coefficient (e.g.) Semen freshness was scored using an exponential decay function, applied immediately after semen collection. The freshness score is 1. As the remaining shelf life decreases, the freshness declines exponentially, with the decline accelerating as the expiration date approaches. Compared to the linear decline method, exponential decay inflicts a greater penalty as the expiration date approaches, which better reflects the biological characteristic of non-linear decline in semen motility over time. This effectively guides the system to prioritize the consumption of semen that is about to expire, reducing semen waste.

[0088] After obtaining the scores for the above six sub-items, the calculation of the comprehensive recommendation score can be expressed as follows: (twenty one) in, to Assign weights to each sub-item score. For example, the weights for each sub-item score can be set as shown in Table 2 below: Table 2 It can be seen that, (Sow quality matching) and (Market user matching) accounted for 0.45% of the total, which is the highest priority combination, reflecting the core objective of precise recommendation based on regional sow quality and market user characteristics; (Breeding index regional adaptation) maintains a weight of 0.20 to ensure that breeding quality is not reduced; (Matching system contribution) Ensure that the recommended semen meets the matching system optimization goals; and As both a quality assurance and a timeliness dimension, these factors work together to ensure the actual usability of the recommended semen.

[0089] In other embodiments, the comprehensive recommendation score can also be calculated using multi-objective optimization algorithms instead of weighted summation. For example, multi-objective evolutionary algorithms such as the Non-dominated Sorting Genetic Algorithm II (NSGA-II) can be used to search for the Pareto optimal front, with multiple objectives including breeding index regional adaptation, sow quality adaptation, market user matching, breeding line contribution, and semen freshness utilization. Multi-objective optimization algorithms can optimize multiple objectives simultaneously without pre-setting the weights between objectives, avoiding the subjectivity of weight setting in weighted summation methods. They output multiple non-dominated solutions on the Pareto optimal front for decision-makers to choose from, making them suitable for application scenarios where there are strong conflicts between the objectives in different dimensions and it is difficult to determine the weights.

[0090] After sorting the semen samples by comprehensive recommendation score from highest to lowest, the system outputs recommended semen solutions. For example, it can output the top 5 candidate semen samples based on their comprehensive recommendation scores as recommended solutions. Outputting a predetermined number of candidate semen samples at the top of the ranking provides insemination personnel with a list of semen recommendations that can be directly implemented, while also retaining multiple alternative solutions to cope with inventory changes or other constraints in actual insemination.

[0091] In some embodiments, when outputting semen recommendation schemes, the system simultaneously outputs detailed sub-scores, inbreeding coefficients, and breeding line role information for each candidate semen, and saves a complete recommendation decision log. Detailed sub-scores may include, for example, specific values ​​for breeding index regional suitability score, comprehensive semen quality score, sow quality suitability score, market user matching score, breeding line contribution score, and semen freshness score. Outputting detailed sub-scores allows breeding personnel and breeding managers to understand the specific performance of each candidate semen across various evaluation dimensions, enhancing the transparency and interpretability of the recommendation results and facilitating manual review and adjustment of the recommendations by decision-makers based on actual circumstances. Inbreeding coefficients, for example, can be the inbreeding coefficients of the expected offspring of the candidate boar and the sow to be bred, allowing decision-makers to intuitively understand the inbreeding risk level of the offspring after mating with each candidate semen, providing data support for manual verification of genetic safety. The role information of the breeding system can be, for example, the hybridization role that a candidate boar plays in the optimal breeding system in the target region (such as terminal sire, maternal sire, etc.). This allows decision-makers to intuitively understand the positioning of each candidate semen sample in the breeding system's hybridization architecture, facilitating the verification of the consistency between the recommendation results and the breeding system optimization objectives. Maintaining a complete recommendation decision log can record, for example, the input parameters, screening process, scoring calculation process, and final recommendation results for each recommendation. A complete recommendation decision log provides a data foundation for post-event auditing and algorithm optimization, enabling the system to trace the entire process of each recommendation decision, supporting retrospective analysis of recommendation effects and continuous optimization of algorithm parameters.

[0092] To further optimize the accuracy of sow selection and culling decisions in the above embodiments, this application also provides a sow breeding scheme based on the goal of optimal breeding line. This scheme integrates four links: sow selection, boar breeding and semen recommendation, and breeding line optimization, forming a closed-loop breeding decision-making system. This ensures that the direction of genetic improvement is consistent with the needs of breeding lines and the market, thereby improving the efficiency of genetic progress.

[0093] Sow breeding decisions are based on the optimal breeding line in the target region as the optimization objective. This optimization objective is then transmitted to the sow breeding stage to ensure that selected sows can maximize their reproductive value within the optimal breeding line, avoiding the dispersion of genetic progress caused by a mismatch between sow breeding direction and breeding line requirements. A sow breeding priority score is calculated by comprehensively considering individual sow quality, breeding line role suitability, and offspring profitability potential. For example, the calculation can be expressed as follows: (twenty two) in, The overall quality score for each individual sow. To ensure compatibility with the character set, To share the profit potential of future generations , , The weights for each dimension are assigned. The sow breeding priority scoring system integrates three dimensions—individual quality, breeding line compatibility, and profit potential—into a comprehensive scoring framework. This ensures that breeding decisions not only focus on the sow's own quality but also consider her role compatibility within the breeding line and the economic potential of her offspring, thus achieving a shift from a single reproductive performance-oriented approach to a multi-dimensional comprehensive evaluation.

[0094] Matching system character suitability This measures the sow's suitability for the target breeding line. For example, if the target breeding line is a Duroc × Landrace × Large White three-way cross, then the Landrace × Large White two-way crossbred sow has the highest suitability as the maternal parent (e.g., ...). Purebred Large White sows have a high compatibility rate as maternal line material (e.g., 0.8), while other breeds have a lower compatibility rate. The compatibility of the breeding system is scored based on the importance of the sow breed's role in the target breeding system's hybridization structure, ensuring that the selected sows can assume the most suitable reproductive role in the breeding system and fully leverage the hybridization advantages of the breeding system.

[0095] Profit potential of offspring The calculation is based on the expected profit contribution of the sow's offspring in the target breeding line. For example, the calculation can be expressed as follows: (twenty three) in, The profitability score for the best supporting facilities in the region. The quality prediction score for the sow's expected offspring is calculated (based on the sow's quality score and the average genetic progress prediction of the breeding line). The offspring profitability potential score combines the breeding line profitability score with the sow's expected offspring quality prediction, quantifying the economic benefit potential that the sow contributes to the breeding line through reproduction. This enables breeding decisions to be economically oriented, prioritizing the selection of sows with the highest offspring profitability potential.

[0096] For example, the weights of each dimension in the sow breeding priority scoring can be set as follows: , , Setting the weights for individual quality and breeding line compatibility to the same maximum value reflects the dual requirement of sow breeding to simultaneously consider both individual quality and breeding line role compatibility; the weight for offspring profitability potential is slightly lower but maintained at a high proportion to ensure that economic benefits are fully reflected in breeding decisions.

[0097] Based on breeding priority scores, a list of recommendations for sow selection and culling is generated. For example, sows can be sorted from highest to lowest breeding priority score, with higher-priority sows prioritized for selection and insemination with high-quality semen, while lower-priority sows are included in the culling plan. This list of recommendations provides breeders with a directly implementable sow management plan, ensuring the orderly replacement of sow herds and continuously improving the overall quality and breeding stock compatibility of the regional sow population.

[0098] In addition to sow breeding decisions, embodiments of this application also include boar breeding planning. Boar breeding planning first predicts the semen requirements of different boar breeds based on the regional sow inventory, mating plan, and optimal breeding line. For example, the prediction of semen requirements can be expressed as follows: (twenty four) in, For the region The number of breeding sows For supporting systems medium-sized varieties The percentage of semen demand, This refers to the average annual mating frequency of sows (e.g., 2.2 times / year). This represents the average number of semen units produced per boar per year (e.g., 1500 units / year). Semen demand forecasting combines regional sow inventory, breeding plans, and the breed demand structure of the optimal breeding line to quantify the actual semen demand for each breed of boar. This provides demand-side data for boar breeding planning and avoids supply-demand imbalances caused by a mismatch between boar breeding direction and semen demand.

[0099] Based on the prediction of semen demand, a boar breeding priority score is calculated by comprehensively considering the boar breeding index, the role of the breeding line, the urgency of semen demand, and the contribution of population genetic diversity, and a boar breeding priority list is output. For example, the calculation can be expressed as follows: (25) in, Scoring the boar breeding index. Rate the supporting characters. Score the urgency of semen demand. Contributes to genetic diversity to The weights for each dimension are assigned. The boar breeding priority scoring system integrates four dimensions into the scoring framework: breeding quality, the role of the breeding line, the urgency of demand, and genetic diversity. This allows boar breeding decisions to pursue high breeding indices while also considering the needs of the breeding line, the balance between semen supply and demand, and the maintenance of population genetic diversity, thereby achieving optimal allocation of boar breeding resources.

[0100] Boar breeding index score For example, the boar's EBV or breeding index can be determined by its quantile ranking within the entire herd. Using the herd quantile to determine the breeding index score allows boars of different breeds and with different index dimensions to be compared on a unified scoring scale, ensuring that the boar with the best breeding quality receives the highest index score. (Matching line role scoring) The importance of the boar's role in the target breeding line is measured; for example, the role score of the terminal sire is higher than that of the maternal sire. The breeding line role score is differentiated based on the boar's importance in the hybridization structure, ensuring that breeding resources are prioritized for the most critical boars in the breeding line, thereby improving the overall genetic improvement efficiency of the breeding line. Semen demand urgency score. For example, the urgency of semen demand can be determined based on the ratio of predicted semen demand to the current stock of boars of that breed. The larger the demand gap, the higher the urgency. The semen demand urgency score quantifies the supply and demand relationship into a scoring index, so that breeding resources are prioritized for allocation to boars of breeds with the largest semen demand gap, timely supplementing the semen supply shortage and ensuring that the mating needs of various regions are met in a timely manner.

[0101] The contribution of genetic diversity to the population is calculated based on the kinship between the candidate boar and the existing breeding pig population. The more distant the kinship between the candidate boar and the existing breeding pig population, the higher the contribution of genetic diversity. For example, the calculation can be expressed as follows: (26) in, For candidate boars and the first The kinship coefficient of existing breeding pigs, The total number of existing breeding pigs. The genetic diversity contribution score measures the contribution of candidate boars to the genetic diversity of the existing breeding pig population by calculating the average kinship coefficient between the candidate boar and the existing breeding pig population. Boars with more distant kinship can introduce more new genetic variations into the population, which helps maintain the genetic diversity of the breeding pig population, reduces the risk of increased inbreeding levels, and ensures the long-term genetic health and breeding potential of the breeding pig population.

[0102] For example, the weights of each dimension in the boar breeding priority scoring can be set as follows: , , , The breeding index weight is set to the highest value to ensure that the bred boars have the best genetic quality; the roles of breeding lines and the urgency of demand are weighted equally to balance the needs of breeding lines and the supply and demand balance; the genetic diversity weight is kept at an appropriate proportion to maintain the genetic diversity of the population while pursuing genetic progress.

[0103] To enable the recommendation system to continuously improve based on actual production performance, this application also provides an effect tracking and iterative optimization scheme. By tracking the actual production performance and profit data after mating, the weight parameters of the recommendation algorithm are adaptively adjusted, and the regional sow quality profile is updated regularly, forming a closed-loop iterative system of recommendation, execution, tracking, feedback, and optimization. This system can continuously self-optimize based on actual results, adapt to market changes and breeding progress, and maintain long-term effectiveness.

[0104] Tracking production performance data after mating can include indicators such as conception rate, litter size, piglet birth weight, weaning weight, age at 100kg, backfat thickness, lean meat percentage, and meat quality grade. This data is then aggregated and statistically analyzed by breeding line, region, and individual boar. This multi-dimensional tracking of production performance data provides comprehensive data support for evaluating the actual effectiveness of recommended programs. Aggregating and statistically analyzing data by breeding line, region, and individual boar allows for analysis of the recommended program's effectiveness at different granularities, identifying which dimensions the program excels in or lacks.

[0105] The deviation between the actual profit and the expected profit of the supporting system can be calculated, for example, as follows: (27) in, To determine the actual gross profit per head of the supporting system, This represents the predicted profit score mapping value during recommendation. The calculation of profit deviation provides a quantitative indicator for evaluating the economic effectiveness of the recommendation scheme, enabling the system to objectively measure the gap between the actual profit performance and the expected profit after the implementation of the recommendation scheme, and providing feedback signals for subsequent weight adjustments and algorithm optimization.

[0106] When the actual profit of a particular product line in a corresponding region is consistently negative compared to its expected profit, the system automatically adjusts the profit score weight of that product line in that region, lowering its recommendation priority. This automatic weight adjustment mechanism based on profit deviation can promptly identify product lines whose actual profit is consistently lower than expected and automatically reduce their recommendation priority, preventing the system from continuously recommending inefficient product line combinations and shortening the identification and elimination cycle for inefficient product lines.

[0107] Building upon this foundation, and based on historical recommendation performance data, Bayesian optimization is used to adaptively adjust the weights of each component in the overall recommendation score. Bayesian optimization efficiently searches for the optimal weight combination by constructing surrogate models (such as Gaussian processes) and acquisition functions (such as desired improvement). Compared to grid search or random search, it can find near-optimal weight configurations with fewer evaluation iterations, making it suitable for scenarios with high evaluation costs, such as weight optimization.

[0108] The objective function is to maximize the average gross profit per head after the implementation of the recommendation scheme, and the values ​​of each weight are constrained to be within a certain range. Given an interval where the sum of all weights is 1, we search for the optimal weight combination. Using the average gross profit per head as the objective function, we directly optimize based on economic efficiency, ensuring that the direction of weight optimization aligns with improving actual profitability. The weight values ​​and summation constraints guarantee the mathematical rationality and interpretability of the optimization results.

[0109] In addition, the sow quality profiles for each region are recalculated and updated according to a preset cycle, such as monthly. Regularly updating the regional sow quality profiles can dynamically reflect the changing trends in regional sow quality (such as quality changes caused by selection and culling, new sows, etc.), ensuring that the recommended scheme is always based on the latest quality data and avoiding the disconnect between the recommended results and the actual situation in the region due to outdated quality profiles.

[0110] In some embodiments, adaptive weight optimization can be automatically performed once at a preset period (e.g., quarterly), and the optimization results will be pushed to the management terminal for approval before taking effect. Automatically performing weight optimization at a preset period and taking effect only after approval by the management terminal ensures the timeliness of optimization while introducing a manual review process. This avoids abnormal weight configurations that may result from automatic algorithm adjustments, ensuring the rationality and controllability of the weight optimization results.

[0111] Exemplary System This application provides an intelligent breeding system for pigs, such as... Figure 2 As shown, the system achieves a complete processing chain from multi-source data acquisition, data cleaning and integration, intelligent analysis and calculation to decision generation through a layered architecture design of data acquisition layer, data governance layer, intelligent analysis layer and decision engine layer.

[0112] The data acquisition layer is used to collect raw data from multiple systems, including breeding data, production data, semen inventory data, livestock input-output observation data, market data, and environmental data. These multiple systems may include, for example, breeding management systems, pig farm production management systems, semen management systems, ERP systems, market information systems, and environmental monitoring systems. By connecting to multiple business systems, the data acquisition layer achieves unified collection of raw data, providing complete data input for subsequent data governance and intelligent analysis, and avoiding the problem of incomplete analysis results due to incomplete data sources.

[0113] The data governance layer is used to clean, standardize, and integrate collected multi-source data. It includes modules for data cleaning, individual identifier association, region labeling, and time alignment. The data cleaning module, for example, can handle missing values, outliers, and duplicate records, and standardize data formats and coding standards. Through systematic data cleaning and standardization, the data governance layer eliminates noise and inconsistencies in multi-source data, providing high-quality data input for the intelligent analysis layer and ensuring the accuracy and reliability of the analysis results.

[0114] In some embodiments, the individual identification association module links breeding data, production data, semen data, and pedigree data to the individual level through ear tags, constructing a full life-cycle data profile for each pig. Using ear tags as unique identifiers links data scattered across different systems to the individual level, enabling the system to obtain the complete information chain for each pig from birth, growth, mating to production. This provides a unified data view for individual quality scoring, kinship calculation, and semen quality traceability, eliminating data silos. The regional labeling module labels each pig and each production record with regional, farm, and breeding type tags. Regional labeling assigns spatial and breeding-related attributes to the data, enabling the system to perform sow quality evaluation, profitability data analysis, and breeding compatibility calculations by region and breeding line, supporting cross-regional comparative analysis and differentiated recommendation decisions. The time alignment module aligns breeding events, production events, and financial events by monthly and batch dimensions, constructing a time-series dataset. The time alignment module eliminates the differences in time granularity and recording period between different data sources, enabling time-series analyses such as profit trend analysis and quality change tracking to be performed on a unified time dimension, ensuring the time consistency and comparability of analysis results.

[0115] The intelligent analysis layer comprises a regional sow quality evaluation module, a market user profiling module, a profitability data analysis module, and a boar semen precision recommendation engine. Through the collaborative work of multiple specialized modules, the intelligent analysis layer conducts intelligent analysis from four dimensions: regional sow quality, market demand preferences, breeding system profitability, and semen precision recommendation. This provides the decision engine layer with multi-dimensional analytical results, supporting the generation of comprehensive and accurate decision-making solutions.

[0116] Furthermore, in some embodiments, the intelligent analysis layer may also include a sow breeding decision-making module, a boar breeding planning module, and a breeding system optimization module. The sow breeding decision-making module, for example, can calculate the breeding priority of sows based on breeding system optimization goals, guiding sow retention and culling, aligning sow retention and culling decisions with breeding system needs, and ensuring that the renewal direction of the sow population is consistent with the breeding system optimization goals. The boar breeding planning module, for example, can formulate a boar breeding plan based on breeding system optimization goals and semen demand predictions, aligning boar breeding direction with semen demand and breeding system role requirements, avoiding waste of boar breeding resources and supply-demand imbalance. The breeding system optimization module, for example, can integrate profit analysis results, sow quality evaluation, and market user profiles to output optimal breeding system recommendations, providing optimal breeding system recommendations for each region, transforming breeding system selection from relying on experience-based judgment to a systematic decision based on multi-dimensional data analysis. By introducing the above three modules, the intelligent analysis layer can not only perform quality evaluation, market profiling, and semen recommendation, but also support the coordinated decision-making of sow selection, boar breeding, and breeding line optimization, forming a complete breeding decision analysis capability.

[0117] The decision engine layer transforms the computational results of the intelligent analysis layer into executable decision-making solutions, including solution generation, solution approval, solution issuance, and effect tracking functions. The decision engine layer converts intelligent analysis results into executable decision-making solutions, ensures the standardization of decisions through solution approval and issuance processes, and achieves closed-loop verification of decision effectiveness through effect tracking, supporting continuous optimization of solutions.

[0118] In some embodiments, the system may further include an application terminal layer, comprising a management terminal and a mobile terminal, for displaying regional sow quality profiles, breeding line profitability rankings, semen recommendation programs, sow breeding suggestions, and boar breeding plans. The management terminal may be, for example, a personal computer (PC) management terminal, for use by breeders and group decision-makers. The PC management terminal provides a large-screen, high-resolution data display environment, suitable for breeders and group decision-makers to conduct detailed data analysis, program approval, and strategic decision-making. The mobile terminal may be, for example, a mobile application (APP), for use by production managers and breeding personnel. The mobile terminal allows production managers and breeding personnel to view recommended programs and implementation guidance anytime on-site at the pig farm, improving the efficiency of decision information transmission and the convenience of on-site execution. The application terminal layer centrally displays the system's core analysis results and decision-making programs in a visual manner, enabling users in different roles to obtain the information they need according to their respective responsibilities.

[0119] Exemplary device This application also provides an electronic device, such as... Figure 3As shown, the electronic device includes at least one processor 301 and a memory 302 communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor. When executed by the at least one processor, the computer program implements all the steps of the intelligent breeding pig selection method described above, enabling the intelligent breeding pig selection method to operate automatically on general-purpose or dedicated electronic devices, improving the efficiency and consistency of breeding decisions. The at least one processor may be, for example, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), or other chips with data processing capabilities. The memory may be, for example, one or more of the following storage media: Random Access Memory (RAM), Read-Only Memory (ROM), Flash Memory, or Hard Disk Drive (HDD). The communication connection between the processor and the memory may be implemented, for example, via a bus 303.

[0120] Exemplary media and products This application also provides a computer-readable storage medium storing computer-executable instructions. When executed by a processor, these instructions implement all steps of the aforementioned intelligent breeding pig selection method, enabling the method to be stored, transmitted, and deployed as a software product, facilitating flexible application on different hardware platforms. This computer-readable storage medium can be, for example, a non-transitory computer-readable storage medium, including but not limited to disk storage, optical disk storage, flash memory, or solid-state drive (SSD).

[0121] This application also provides a computer program product, which includes a computer program that, when executed by a processor, implements all the steps of the aforementioned intelligent breeding pig selection method. This enables the intelligent breeding pig selection method to be packaged, distributed, and deployed in a standardized program product format, facilitating integration with different hardware platforms and software systems. The computer program product can be stored on a computer-readable storage medium or obtained through network download.

[0122] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for intelligent selection of breeding pigs, characterized in that, include: Collect raw data from multiple sources, including: breeding data, production data, semen inventory data, aquaculture input-output observation data, market data, and environmental data; The multi-source raw data is cleaned, individual identifiers are associated, regional labels are marked, and time alignment is performed to construct a unified database; Based on breeding and production data in the database, a multi-dimensional quality evaluation of sow populations in each region is conducted to construct a regional sow quality profile. Based on market data in the database, preference indicators for terminal markets in various regions are extracted to construct market user profile vectors. Based on the aquaculture input-output observation data in the database, the profitability data of breeding systems in various regions are analyzed from multiple dimensions, and the optimal breeding system recommendation scheme for each region is output. Upon receiving a breeding request from the target region, the system screens candidate semen based on the regional sow quality profile, the market user profile vector, and the optimal breeding line recommendation scheme. It then calculates a comprehensive recommendation score for each candidate semen and outputs a semen recommendation scheme in descending order of comprehensive recommendation score.

2. The method of claim 1, wherein, The multi-dimensional quality evaluation of sow populations in each region includes: Calculate the overall quality score of the sow: Perform weighted normalization on the reproductive performance index, growth performance index, body shape index, and health status index to obtain the reproductive performance score, growth performance score, body shape score, and health status score; multiply the reproductive performance score, growth performance score, body shape score, and health status score by their respective preset weights and then sum them to obtain the overall quality score of the sow. Calculate the regional quality index: Sum the individual quality scores of all sows in the region and divide by the total number of sows in the region to obtain the regional quality index, which is used to characterize the average quality level of the sow population in the region. Calculate the regional quality dispersion: Calculate the standard deviation of the quality scores of all individual sows in the region, divide the standard deviation by the regional quality index to obtain the regional quality dispersion, which is used to characterize the uniformity of the quality of the sow population in the region. Constructing a regional breed structure vector: Statistically determine the proportion of each breed of sow in the region to generate a regional breed structure vector, which is used to characterize the breed composition features of the region; Constructing a regional parity distribution vector: Statistically calculate the proportion of sows of each parity in the region to generate a regional parity distribution vector, which is used to characterize the parity structure of the region. Constructing a regional breeding index mean vector: This vector is constructed based on the population mean corresponding to each breeding index dimension of the regional sow population, and the population standard deviation corresponding to each breeding index dimension is recorded to characterize the breeding index distribution characteristics of the regional sow population.

3. The method according to claim 1, characterized in that, The constructed market user profile vector includes: Extract the lean meat percentage preference index, meat quality grade preference index, growth rate preference index, body shape and appearance preference index, and weight specification preference index of the regional terminal market; Each market preference index is normalized to a unified numerical range to form a market user profile vector; the score of each preference index is calculated by weighting the price differences of different levels of products in the regional market.

4. The method according to claim 1, characterized in that, The method involves multi-dimensional analysis of the profitability data of breeding systems in various regions based on aquaculture input-output observation data, and outputting optimal breeding system recommendations for each region, including: For each supporting system in each region, the profitability score of the supporting system is calculated by weighted summation of three dimensions: profitability level, profitability stability, and profitability growth trend. Among them, profitability level is based on the average gross profit per head of the corresponding supporting system in the region, and is obtained by normalization of the maximum and minimum average gross profit per head of all supporting systems in the region. Profitability stability is calculated based on the coefficient of variation of average gross profit per head of multiple batches. Profitability growth trend is obtained by nonlinear mapping of the slope of change of average gross profit per head of multiple batches. The profitability scores of the same supporting system in different regions are weighted and aggregated across regions to obtain the overall profitability score of the supporting system. The weight of each region participating in the weighted aggregation is determined based on the proportion of output or profit contribution of the corresponding region. The supporting systems are ranked from high to low according to the overall profitability score of the group to obtain the profitability ranking of the supporting systems at the group level. The system calculates the regional suitability score of the breeding system based on the profitability of the breeding system, the matching degree of sow quality, and the matching degree of market users. The system then sorts the breeding systems and outputs the optimal and alternative breeding system solutions for the corresponding regions based on the suitability scores.

5. The method of claim 1, wherein, The process of screening candidate semen: Screen candidate semen that are available in stock, within their expiration date, and meet the semen quality standards; Perform inbreeding constraint filtering: Based on pedigree calculations, filter candidate semen whose expected offspring inbreeding coefficient exceeds a preset threshold; Perform breeding line matching filter: screen boar semen that is compatible with the breed and strain and the optimal breeding line in the target area; Perform regional adaptation filtering: screen boar semen that is compatible with the sow breed structure of the target region.

6. The method of claim 1, wherein, The comprehensive recommendation score for each candidate semen sample is obtained by weighted summation of multiple sub-scores, including breeding index regional suitability score, semen quality comprehensive score, sow quality suitability score, market user matching score, breeding line contribution score, and semen freshness score; wherein: The breeding index regional fit score is calculated by using the Gaussian kernel function to determine the degree of fit between each breeding index of the boar and the breeding index distribution of the sow population in the target area. The closer the boar index is to the regional sow population index distribution, the higher the fit score. The comprehensive semen quality score is calculated by weighting parameters based on continuous monitoring of sperm motility, sperm concentration, and the rate of normal sperm morphology. The sow quality matching is obtained by combining breed matching, parity matching, and sow quality complementarity matching, which represents the degree of matching between the candidate boar and the quality characteristics of the sow population in the target area. The market user matching score is calculated as follows: the genetic performance scores of candidate boars for multiple product feature dimensions are obtained, and the genetic performance scores are mapped from the estimated breeding value or breeding index of the candidate boars; the market user preference scores for each product feature dimension corresponding to the target region are obtained; the deviation between the boar genetic performance score and the regional market user preference score under each product feature dimension is calculated, the average deviation of all product feature dimensions is calculated, and the market user matching score is determined based on the average deviation. The contribution score of the breeding system is calculated as follows: the optimal breeding system corresponding to the target area is obtained, and the corresponding role score is assigned according to the hybridization role played by the candidate boar in the hybridization architecture of the optimal breeding system to obtain the contribution score of the breeding system. The more central the hybridization role of the candidate boar in the optimal breeding line, the higher the contribution score of the breeding line. The semen freshness score is calculated using an exponential decay method, based on the remaining effective days of the semen. The shorter the remaining effective period of the semen, the lower the freshness score.

7. The method of claim 6, wherein, When outputting the semen recommendation scheme, the system outputs a predetermined number of candidate semen at the top of the ranking, along with detailed sub-scores, inbreeding coefficients, and mating system role information for each candidate semen, and saves a complete recommendation decision log.

8. The method according to claim 1, characterized in that, This also includes sow breeding decisions and boar breeding plans, among which: The sow breeding decision-making process includes: taking the best breeding line in the target area as the optimization target, calculating the sow breeding priority score by comprehensively considering the individual quality of the sow, the role suitability of the breeding line, and the profit potential of the offspring, and outputting a list of sow selection and culling recommendations based on the breeding priority score; The boar breeding plan includes: predicting the semen demand of different breeds of boars based on the regional sow inventory, mating plan, and optimal breeding line; calculating the boar breeding priority score by comprehensively considering the boar breeding index, breeding line role, semen demand urgency, and population genetic diversity contribution, and outputting a boar breeding priority list; the population genetic diversity contribution is calculated based on the kinship between candidate boars and the existing breeding pig population, and the more distant the kinship between the candidate boar and the existing breeding pig population, the higher the genetic diversity contribution.

9. The method according to any one of claims 1-8, characterized in that, Also includes: Performance tracking and iterative optimization: Track production performance data after mating and calculate the deviation between actual and expected profits of the breeding system; When the actual profit of a certain supporting system in the corresponding region is consistently negative compared to the expected profit, the profit score weight of that supporting system in that region will be automatically adjusted, and the recommendation priority of that supporting system will be reduced. Based on historical recommendation performance data, the weights of each item in the comprehensive recommendation score are adaptively adjusted using the Bayesian optimization method. The objective function is to maximize the average gross profit per head after the implementation of the recommendation scheme. Each weight is constrained to take values ​​in the range [0,1] and the sum of all weights is 1. The optimal weight combination is searched. The sow quality profiles for each region are recalculated and updated according to a preset cycle.

10. A system for intelligent selection of breeding pigs, characterized in that include: The data acquisition layer is used to collect raw data from multi-source systems, including breeding data, production data, semen inventory data, aquaculture input-output observation data, market data, and environmental data. The data governance layer is used to clean, standardize, and integrate the collected multi-source data, including a data cleaning module, an individual identifier association module, a region labeling module, and a time alignment module. The intelligent analysis layer includes a regional sow quality evaluation module, a market user profiling module, a profit data analysis module, and a boar semen precision recommendation engine; The decision engine layer is used to transform the calculation results of the intelligent analysis layer into executable decision solutions, including solution generation, solution approval, solution issuance, and effect tracking functions.

11. The system of claim 10, wherein, The individual identification association module of the data governance layer associates breeding data, production data, semen data and pedigree data to the individual level through ear tags, constructing a full life cycle data file for each pig; the regional label module labels each pig and each production record with regional, farm area and breeding line type labels; the time alignment module aligns breeding events, production events and financial events by monthly and batch dimensions, constructing a time series dataset.

12. The system of claim 10, wherein, The intelligent analysis layer also includes: a sow breeding decision-making module, a boar breeding planning module, and a breeding system optimization module.

13. The system according to any one of claims 10-12, characterized in that, Also includes: The application terminal layer, including the management terminal and mobile terminal, is used to display regional sow quality profiles, breeding line profitability rankings, semen recommendation programs, sow breeding suggestions, and boar breeding plans.

14. An electronic device, comprising: include: At least one processor; The system includes a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the at least one processor to perform the intelligent breeding method for breeding pigs as described in any one of claims 1 to 9.

15. A computer readable storage medium characterized by: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the intelligent breeding pig selection method according to any one of claims 1 to 9.

16. A computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent breeding method for breeding pigs as described in any one of claims 1 to 9.