A method and system for analyzing poultry breeding records

Abstract

The application discloses a kind of poultry breeding record analysis method and system, and the application relates to poultry breeding technical field, comprising: multi-modal data dynamic fusion module, adaptive dynamic modeling module, multi-mechanism verification and screening module, system coordination interface module;The multi-modal data dynamic fusion module is used to construct and conflict cleaning dynamic knowledge graph to genotype data, environmental time series data and phenotype data.This poultry breeding record analysis method and system realize the deep correlation and real-time fusion of genotype, environment, phenotype multi-source data through dynamic knowledge graph and federal space-time embedding model, effectively break through the data island barrier of traditional system, can be independently balanced between long-term breeding target and short-term dynamic response, simultaneously, the dual verification mechanism of cause-effect reasoning and extreme confrontation test improves decision confidence, significantly reduces the resource waste and risk lag caused by experiential decision.

Description

Technical Field

[0001] This invention relates to the field of poultry farming technology, specifically to a method and system for analyzing poultry breeding records. Background Technology

[0002] As the poultry breeding industry moves towards intensification and precision, the technological bottlenecks of traditional breeding record analysis systems are becoming increasingly apparent. While current mainstream systems can achieve electronic storage and statistics of basic data, their underlying architecture, based on a single-dimensional structured database design, makes it difficult to integrate multi-source heterogeneous data. Data such as poultry genetic pedigrees, dynamic phenotypes, and environmental monitoring are scattered across different subsystems or manual ledgers, lacking unified standards and correlation models, resulting in severe "data silos." Breeding analysis remains at a single-point statistical level, unable to uncover potential patterns across dimensions. Moreover, existing systems often set breeding target parameters based on fixed thresholds or human experience. Faced with sudden epidemics, market fluctuations, or genotype iterations, static models cannot be adjusted and optimized in real time, leading to lagging breeding strategies. Algorithm modules lack dynamic learning capabilities, traditional statistical analysis methods are insufficient for nonlinear correlation mining, making it difficult to build iterative predictive models. When selecting disease-resistant strains, weight coefficients cannot be dynamically adjusted, and selection decisions rely on post-event retrospection. Furthermore, current breeding analysis systems have poor synergy with downstream industries. Changes in market demand for traits such as poultry meat quality cannot be quickly reflected in breeding models; production data from the farming sector also fails to achieve closed-loop optimization through genetic evaluation. This break in the "analysis-decision-application" chain leads to resource waste. Therefore, there is an urgent need for new analytical systems to overcome data barriers and achieve intelligent decision-making to address the complex challenges of modern poultry breeding. Summary of the Invention

[0003] (a) Technical problems to be solved

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for analyzing poultry breeding records. It solves the problem of how to construct a multidimensional correlation analysis system of environment-gene-phenotype through real-time fusion and dynamic modeling of multi-source heterogeneous data, and realize adaptive breeding strategy recommendation based on machine learning, thereby overcoming the technical bottlenecks of data silos, static analysis, and experience-based decision-making in traditional systems.

[0005] (II) Technical Solution

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method and system for analyzing poultry breeding records, comprising the following functional modules: a multimodal data dynamic fusion module, an adaptive dynamic modeling module, a multi-mechanism verification and screening module, and a system collaborative interface module;

[0007] The multimodal data dynamic fusion module is used to construct dynamic knowledge graphs and perform conflict cleanup on genotype data, environmental time-series data, and phenotypic data. It should be further explained that environmental time-series data, including temperature, humidity, and light intensity, as well as SNP sequences from genotype data and egg production rate and body weight from phenotypic data, are collected in real time through temperature and humidity sensors and gene sequencers in IoT devices. The data is then standardized into a unified format, JSON or Protobuf, and transmitted to the central server.

[0008] The adaptive dynamic modeling module implements multimodal spatiotemporal embedding modeling based on the federated learning framework and generates dynamic breeding strategies through a two-layer optimization engine. It should be further noted that, under the federated learning framework, local data from multiple farms are jointly trained on the model through gradient aggregation to avoid leakage of raw data. The local data is encrypted and stored on the local server.

[0009] The multi-mechanism verification and screening module uses causal reasoning and adversarial stress testing to verify the strategy. It should be further explained that after the strategy is generated, a causal reasoning engine, such as the DoWhy library, is called to construct counterfactual scenarios, and adversarial test scripts are launched to simulate extreme events.

[0010] The system's collaborative interface module enables protocol adaptation and visual decision-making interaction for multi-source heterogeneous devices. It should be further noted that, in the specific implementation process, an adapter middleware is developed to parse heterogeneous interfaces such as the Modbus protocol of temperature control devices and the VCF files of gene sequencers, and pushes them to the visualization interface in real time via WebSocket.

[0011] Preferably, the multimodal data dynamic fusion module includes:

[0012] The dynamic ontology modeling unit defines three-dimensional association rules for genetic lineage, environmental parameters, and phenotypic indicators based on a dedicated ontology library for poultry breeding.

[0013] The knowledge graph evolution unit uses the spatiotemporal sequence clustering algorithm ST-DBSCAN to identify the temporal coupling pattern between environmental parameters and phenotypic data, and encodes it as dynamic weighted edges;

[0014] The dual-channel data cleaning unit sets hard filtering thresholds for genotype and phenotype data through a rule engine, and performs soft corrections on sensor time-series data based on a generative adversarial network (GAN).

[0015] Preferably, the adaptive dynamic modeling module includes:

[0016] The heterogeneous federated spatiotemporal embedding model Hetero-FSTE maps genotype SNP sequences, phenotypic time series, and environmental tensors to a unified vector space, and employs the gated spatiotemporal attention mechanism GSTA to capture gene-environment interaction effects. Further explanation is needed regarding the specific implementation process:

[0017] Genotype embedding includes encoding SNP sequences into 128-dimensional vectors using a one-hot self-attention mechanism, where the SNP sequences include AA, AT, and TT.

[0018] Phenotypic embedding includes: extracting frequency domain features from the daily recorded egg production rate time series using wavelet transform, and then generating a 64-dimensional vector through a GRU network;

[0019] Environmental embedding includes: converting the temperature and humidity heat map divided by chicken coop area into a three-dimensional tensor, and outputting a 32-dimensional vector through a 3D convolutional layer;

[0020] Gated spatiotemporal attention (GSTA) includes: calculating the cross-attention score between gene vectors and environment vectors, and screening significant interaction terms, such as the weighted increase of the influence of SNP_A on egg production rate under high temperature conditions.

[0021] The vertical federated learning architecture allows local servers to retain the original data, uploading only encrypted model gradients to the central server for aggregation. This achieves cross-entity data privacy protection and format heterogeneity compatibility. It should be further noted that the specific implementation includes:

[0022] Data alignment: Based on differential privacy technology, individual IDs from multiple farms are aligned in an encrypted state, such as through SHA-256 hash encryption;

[0023] Gradient aggregation: Local model, i.e., farm A calculates the gradient. After encryption, it is uploaded to the central server and... Weighted average, update the global model.

[0024] Preferably, the adaptive dynamic modeling module further includes a two-layer optimization decision engine, which comprises a global deep Q-network (DQN) and a real-time proximal strategy optimization (PPO) algorithm. The global deep Q-network generates long-term breeding target genotype combinations based on historical data. The real-time proximal strategy optimization algorithm receives environmental monitoring data to dynamically adjust the objective function weights and verifies strategy compatibility through Monte Carlo Tree Search (MCTS).

[0025] Preferably, the multi-mechanism verification and screening module includes:

[0026] The counterfactual causality verification unit constructs counterfactual scenarios for a policy based on the potential outcome model (PO Model), compares actual data with counterfactual predictions, and quantifies the confidence level of the policy's causal effect. It should be further noted that in the specific implementation process:

[0027] Construct a counterfactual scenario: Assuming that the system-recommended "selection SNP_A" strategy was not adopted, predict the egg production rate of the unselected population based on the potential outcome model POModel;

[0028] Calculate the causal effect: the difference between the actual egg production rate Y1 and the counterfactual egg production rate Y0. The formula is: Δ = Y1 - Y0. If Δ > 5% and p < 0.05, the strategy is considered effective.

[0029] Extreme-AGAN, an extreme scenario adversarial generator, simulates genotype mutations, drastic environmental changes, and sudden market shifts. It uses Shapley value analysis to locate the vulnerability of the strategy and trigger model retraining.

[0030] It should be further explained that, in the specific implementation process:

[0031] Genotype mutation generation: Randomly mask 10% of SNP sites to simulate sequencing errors;

[0032] Dramatic environmental change: Injecting a 7-day period of high temperatures (+5°C) into temperature and humidity data;

[0033] Market Turnaround: Assuming a 30% drop in poultry demand, recalculate breeding target returns;

[0034] Shapley value analysis: Identifies the features that have the greatest impact on the stability of the strategy, such as a temperature control equipment failure causing a weight shift of >20%, triggering model retraining, thereby ensuring the scientific nature and risk resistance of the strategy.

[0035] Preferably, the counterfactual causality verification unit performs the following operations:

[0036] Construct counterfactual hypothetical scenarios for the system's recommendation strategy and calculate the contribution of the differences before and after the strategy implementation.

[0037] When the confidence level of the causal effect falls below a preset threshold, the strategy rollback and model iteration update are automatically triggered.

[0038] The difference contribution was calculated by combining Bayesian network and intervention factor decomposition.

[0039] Preferably, the system collaboration interface module includes:

[0040] The Intelligent Protocol Adapter (IPA) supports second-level protocol parsing of VCF file interfaces for gene sequencing platforms, Modbus protocols for intelligent environmental control equipment, and market data APIs. It should be further noted that the specific implementation process includes:

[0041] Gene sequencing platform interface: Parses SNP site information (such as chr1:123456A / T) in VCF files and converts it into a unified genotype code (0 / 1 / 2 corresponds to AA / AT / TT);

[0042] Modbus protocol parsing: Read the register address of the environmental control device (e.g., 0x0001 is the temperature value), convert it to a floating-point number and add a timestamp;

[0043] Market Data API: Poultry futures prices are obtained through OAuth2.0 authorization and integrated into the breeding revenue model on a daily basis.

[0044] The system provides a 3D breeding status map, dynamically visualizing genotype diffusion trends, environmental hotspot distribution, and strategy benefit heatmaps. It also offers an interactive interface for expert annotation and AI strategy tuning. Specifically, in implementation, WebGL is used to render the 3D model of the chicken house, with different colors marking genotype diffusion areas, red indicating high-incidence areas. Experts can manually adjust strategy parameters, such as changing the egg production rate weight from 0.6 to 0.8, and the system simulates the adjusted benefit heatmap in real time. Through the system's collaborative interface module, seamless access to multiple devices and visualized decision support are achieved.

[0045] Preferably, the dual-channel data cleaning unit includes:

[0046] An adversarial neural network discriminator performs secondary noise detection on the data that has been initially cleaned by the rule engine.

[0047] A dynamic weight allocator adjusts the cleaning intensity based on the credibility of the data source, where the data source credibility is ranked as follows: gene sequencing platform > manual input > edge sensor.

[0048] The data repair results are finally arbitrated by comparing the predictions of the Bi-LSTM (Bi-Long Short-Term Memory) network with the actual values.

[0049] Preferably, the Monte Carlo Tree Search (MCTS) verification process includes: simulating environmental fluctuations and market changes over the next 30 days before strategy deployment; generating a Markov chain model based on historical data to predict the probability of temperature and humidity shifts over the next 30 days (e.g., if the temperature on day t > 28°C, the probability of a temperature drop on day t+1 is 60%); evaluating the stability of the target traits and the resource consumption growth rate of the strategy under the simulated path; calculating the feed increment and immunization costs required for strategy execution; if the growth rate > 20%, i.e., a preset threshold, an early warning is triggered; when the resource consumption growth rate exceeds the preset threshold, a real-time layer strategy weight reallocation is automatically triggered, reducing the Q-value weight of high resource consumption strategies, such as from 0.8 to 0.6, and prioritizing strategy branches with a cost growth rate < 15%. Through the implementation of the Monte Carlo Tree Search (MCTS) verification process, the stability of the strategy under the simulated environment is verified.

[0050] Preferably, the Monte Carlo Tree Search (MCTS) verification process further includes:

[0051] A dynamic environmental path generator, based on historical environmental data and real-time weather forecast API, constructs a probability distribution model of temperature, humidity, and light intensity for the next 30 days.

[0052] The strategy robustness evaluation unit quantifies strategy stability through the following steps:

[0053] 1) Inject random disturbance events into the simulation path, such as temperature control failure due to equipment malfunction or feed supply interruption;

[0054] 2) Calculate the coefficient of variation of the target trait before and after the perturbation. If the coefficient of variation exceeds the preset threshold, the strategy is deemed to be insufficiently robust. The target traits include egg production rate and weight gain rate.

[0055] 3) Based on the evaluation results, automatically adjust the exploration in the real-time near-end policy optimization PPO algorithm by using the balance parameter to prioritize policy branches that are not sensitive to perturbations;

[0056] The dynamic adjustment process is achieved through joint optimization of gradient backpropagation and policy entropy, ensuring the model's adaptability in complex scenarios.

[0057] A method for analyzing poultry breeding records includes the following steps:

[0058] S1: Dynamic knowledge graph construction and conflict removal for genotype data, environmental time-series data, and phenotypic data;

[0059] S2: Multimodal spatiotemporal embedding modeling is implemented based on the federated learning framework, and dynamic breeding strategies are generated through a two-layer optimization engine;

[0060] S3: The strategy is validated by both causal reasoning and adversarial stress testing;

[0061] S4: Perform protocol adaptation and visual decision-making interaction for multi-source heterogeneous devices.

[0062] (III) Beneficial Effects

[0063] This invention provides a method and system for analyzing poultry breeding records. It has the following beneficial effects:

[0064] (I) This poultry breeding record analysis method and system, through dynamic knowledge graphs and federated spatiotemporal embedding models, achieves deep correlation and real-time fusion of multi-source data (genotype, environment, phenotype), effectively breaking down data silos in traditional systems. Based on the collaborative optimization of two-layer reinforcement learning and Monte Carlo tree search verification, the system can autonomously balance long-term breeding goals and short-term dynamic responses, improving the efficiency of identifying individuals with high genetic value and shortening the response time for strategy adjustments. Simultaneously, the dual verification mechanism of causal reasoning and extreme adversarial testing enhances decision confidence, significantly reducing resource waste and risk lag caused by experience-based decision-making.

[0065] (II) This poultry breeding record analysis method and system, through a dynamic environmental path generator and strategy robustness quantitative assessment, can simulate complex scenarios such as epidemics, extreme weather, and market fluctuations. It automatically optimizes strategy parameters based on the coefficient of variation threshold, ensuring improved trait stability under perturbation events. Furthermore, the intelligent protocol adapter and 3D visualization interactive design create a closed-loop system across the entire industry chain, integrating gene sequencing, environmental monitoring, and market data. This improves the alignment between breeding strategies and industry needs, optimizing resource utilization. This solution combines technological foresight with practical feasibility, providing core support for the intelligent upgrading of the poultry breeding industry. Attached Figure Description

[0066] Figure 1 This is a schematic diagram of the overall framework of the present invention. Detailed Implementation

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Please see Figure 1 The present invention provides a technical solution: a method and system for analyzing poultry breeding records, comprising the following functional modules: a multimodal data dynamic fusion module, an adaptive dynamic modeling module, a multi-mechanism verification and screening module, and a system collaborative interface module;

[0069] The multimodal data dynamic fusion module is used to dynamically construct knowledge graphs and perform conflict removal on genotype data, environmental time-series data, and phenotypic data.

[0070] The adaptive dynamic modeling module implements multimodal spatiotemporal embedding modeling based on the federated learning framework and generates dynamic breeding strategies through a two-layer optimization engine;

[0071] The multi-mechanism verification and screening module uses causal reasoning and adversarial stress testing to perform dual verification of the strategy;

[0072] The system collaboration interface module enables protocol adaptation and visual decision-making interaction for multi-source heterogeneous devices.

[0073] The multimodal data dynamic fusion module includes:

[0074] The dynamic ontology modeling unit defines three-dimensional association rules for genetic lineage, environmental parameters, and phenotypic indicators based on a dedicated ontology library for poultry breeding. It should be further explained that, in the specific implementation process, the three-dimensional association rules of the ontology library include genetic, environmental, and phenotypic dimensions. Specifically, in the genetic dimension, the association rules between SNP loci and dominant traits are defined, such as rs123 locus A → egg production rate +5%; in the environmental dimension, the influence function of temperature and humidity thresholds on growth rate is defined, such as weight gain rate decreasing when temperature >28℃; and in the phenotypic dimension, a nonlinear relationship model between egg production rate and feed intake is defined.

[0075] The knowledge graph evolution unit utilizes the spatiotemporal sequence clustering algorithm ST-DBSCAN to identify the temporal coupling patterns between environmental parameters and phenotypic data, and encodes them as dynamic weighted edges. It should be further noted that in the specific implementation, the ST-DBSCAN algorithm is used to process the environmental time-series data, including the input temperature sequence T = {t1, t2, ..., t...}. n}, humidity sequence H={h1, h2, …, h n Clustering identifies "high temperature and high humidity clusters" as periods of three consecutive days with temperature fluctuations >5℃ and humidity >80%; it also includes association, that is, associating the cluster with individuals whose egg production rate drops by more than 10% during the same period, generating dynamic edge weights, where weight = correlation coefficient × 0.8.

[0076] The dual-channel data cleaning unit sets hard filtering thresholds for genotype and phenotype data through a rule engine, and performs soft corrections on sensor time-series data based on a generative adversarial network (GAN). It should be further explained that the specific implementation includes rule engine cleaning and GAN soft correction. Rule engine cleaning includes hard filtering, namely: deleting SNP sites with sequencing depth <20X and abnormal records with a daily weight gain >15%; GAN soft correction includes a generator and a discriminator. The generator's LSTM network simulates normal temperature and humidity fluctuation patterns; the discriminator's convolutional neural network (CNN) distinguishes between real and generated data; after adversarial training, sensor abrupt data changes, such as a temperature instantaneously changing from 25℃→40℃→25℃, are smoothed and repaired.

[0077] The adaptive dynamic modeling module includes:

[0078] The heterogeneous federated spatiotemporal embedding model Hetero-FSTE maps genotype SNP sequences, phenotypic time series and environmental tensors to a unified vector space, and uses the gated spatiotemporal attention mechanism GSTA to capture gene-environment interaction effects.

[0079] The vertical federated learning architecture allows local servers to retain the original data and only upload encrypted model gradients to the central server for aggregation, thereby achieving cross-entity data privacy protection and heterogeneous format compatibility.

[0080] The adaptive dynamic modeling module also includes a two-layer optimization decision engine, which comprises a global deep Q-network (DQN) and a real-time proximal policy optimization (PPO) algorithm. The global deep Q-network generates long-term breeding target genotype combinations based on historical data. The real-time proximal policy optimization algorithm receives environmental monitoring data, dynamically adjusts the objective function weights, and verifies policy compatibility through Monte Carlo Tree Search (MCTS). It should be further noted that the specific implementation process includes:

[0081] Global Deep Q Network (DQN):

[0082] Input: Breeding data from the past 3 years (genotype, environment, phenotype);

[0083] Output: Q-value matrix, representing the long-term returns of different genotype combinations, such as the total egg production over 5 years, where different genotype combinations include SNP_A + SNP_B;

[0084] Training: Explore the optimal combination using an ε-greedy strategy, with the objective function being to maximize the cumulative reward from the discount.

[0085] Real-time Proximity Policy Optimization (PPO):

[0086] Input: Real-time environmental data, such as the latest epidemic report and market price fluctuations;

[0087] Policy update: Sample the latest data every 2 hours, calculate the policy gradient and limit the update magnitude. Clip threshold ε = 0.2 to prevent model oscillation;

[0088] Monte Carlo Tree Search (MCTS) Validation:

[0089] Simulation: Generate 1000 environmental paths for the next 30 days, with temperature and humidity normally distributed and perturbed by ±10%;

[0090] Evaluation: Calculate the variance of the target trait for each path. If the variance is greater than 15%, the strategy is considered unstable. The target trait is egg production rate.

[0091] Backtracking: Select the top 10% of paths with the smallest variance, backpropagate to the advantage function of the PPO algorithm, and adjust the policy weights.

[0092] The multi-mechanism verification and screening module includes:

[0093] The counterfactual causal verification unit constructs counterfactual scenarios of a strategy based on the potential outcome model (PO Model), compares actual data with counterfactual prediction results, and quantifies the confidence level of the causal effect of the strategy.

[0094] Extreme-AGAN, an extreme scenario adversarial generator, simulates genotype mutations, drastic environmental changes, and sudden market shifts. It uses Shapley value analysis to locate the vulnerability of the strategy and trigger model retraining.

[0095] The counterfactual causality verification unit performs the following operations:

[0096] Construct counterfactual hypothetical scenarios for the system's recommendation strategy and calculate the contribution of the differences before and after the strategy implementation.

[0097] When the confidence level of the causal effect falls below a preset threshold, the strategy rollback and model iteration update are automatically triggered.

[0098] The contribution of the difference was calculated by combining Bayesian network and intervention factor decomposition.

[0099] It should be further explained that the specific implementation process includes:

[0100] Calculation of difference contribution:

[0101] Bayesian network: Construct a causal graph of SNP sites, environmental parameters, and phenotypic indicators, calculate the posterior probability P, egg production rate ↑|SNP_A=1, temperature <25℃;

[0102] Intervention factor decomposition: Environmental disturbances are separated by do-calculus, such as the expected change in egg production rate at do (temperature = 25℃);

[0103] Strategy rollback mechanism:

[0104] If the confidence level of the causal effect is less than 90%, the system will automatically revert to the previous stable version of the strategy and mark the current strategy as "high risk".

[0105] Triggering model iteration: Increase the weight of adversarial training samples and prioritize learning historically successful policy patterns. The above operations, performed by the counterfactual causal verification unit, quantify causal effects and achieve dynamic policy iteration.

[0106] The system collaboration interface module includes:

[0107] The Intelligent Protocol Adapter (IPA) supports second-level protocol parsing for gene sequencing platform VCF file interfaces, intelligent environmental control equipment Modbus protocols, and market data APIs.

[0108] A 3D breeding trend map dynamically visualizes genotype diffusion trends, environmental hotspot distribution, and strategy benefit heatmaps, and provides an interactive optimization interface for human expert annotation and AI strategies.

[0109] The dual-channel data cleaning unit includes:

[0110] The adversarial neural network discriminator performs secondary noise detection on the data initially cleaned by the rule engine. The input includes temperature and humidity data after rule cleaning. During training, the generator produces synthetic data, and the discriminator distinguishes between real data and synthetic data. The output is to mark suspected noise points, such as a sudden drop in temperature of 3°C lasting for 10 minutes, with a probability labeling. If the probability is >70%, repair is triggered.

[0111] The dynamic weight allocator adjusts the cleaning intensity based on the data source credibility, where the data source credibility is ranked as follows: gene sequencing platform > manual input > edge sensor; the credibility rules give gene sequencing platform data a weight of 0.9, manual input a weight of 0.7, and edge sensor a weight of 0.5. The cleaning intensity includes GAN correction with an allowable correction range of ±5% for sensor data, while gene data is only allowed ±1% correction.

[0112] The data repair results are ultimately arbitrated by comparing the predictions from a bidirectional long short-term memory (Bi-LSTM) network with the actual values. This includes the difference sequences between the input rule-cleaned data and the GAN-repaired data; the Bi-LSTM predicts the reasonable value range for the next time point; if both sets of data exceed the range, they are marked as "awaiting manual review." The application of a dual-channel data cleaning unit improves both the accuracy and efficiency of data cleaning.

[0113] The Monte Carlo Tree Search (MCTS) verification process includes:

[0114] Simulate the path of environmental fluctuations and market changes over the next 30 days before deploying the strategy;

[0115] The stability of the target traits and the rate of increase in resource consumption of the evaluation strategy under the simulated path;

[0116] When the resource consumption growth rate exceeds a preset threshold, the real-time layer strategy weight reallocation is automatically triggered.

[0117] The Monte Carlo Tree Search (MCTS) validation process further includes:

[0118] The dynamic environment path generator constructs a probability distribution model of temperature, humidity, and light intensity for the next 30 days based on historical environmental data and real-time weather forecast API; it also accesses the weather API to obtain forecast data for the next 7 days and constructs an ARIMA model to predict the probability distribution of temperature and humidity for the following 23 days with a confidence interval of 95%.

[0119] The strategy robustness evaluation unit quantifies strategy stability through the following steps:

[0120] 1) Inject random disturbance events into the simulation path, including temperature control failure caused by equipment failure and feed supply interruption. Among them, equipment failure: randomly select 5% of temperature control nodes and simulate failure for 48 hours, with random temperature fluctuations of ±8℃; feed interruption: randomly insert an event of 50% decrease in feed supply for 3 days into the simulation path.

[0121] 2) Calculate the coefficient of variation of the target traits (egg production rate, weight gain rate) before and after the disturbance. If the coefficient of variation exceeds the preset threshold, the strategy is deemed to be insufficiently robust. The formula for calculating the coefficient of variation is: Coefficient of variation of egg production rate CV = standard deviation / mean. If CV > 0.25 threshold, the strategy is deemed to be ineffective.

[0122] 3) Based on the evaluation results, the exploration-utilization balance parameter in the real-time near-end policy optimization PPO algorithm is automatically adjusted to prioritize policy branches that are not sensitive to disturbances; the exploration rate ε = 0.3 of the original PPO algorithm is reduced to ε = 0.1 when the policy robustness is insufficient, and historically stable policies are prioritized.

[0123] The dynamic adjustment process is achieved through joint optimization of gradient backpropagation and policy entropy, ensuring the model's adaptability in complex scenarios. Specifically, it adds an entropy penalty term to the policy distribution to prevent over-reliance on a single policy branch. Furthermore, the Monte Carlo Tree Search (MCTS) validation process enhances the policy's robustness under extreme perturbations.

[0124] It should be further explained that, in the specific implementation process, this solution achieves deep association and real-time fusion of multi-source data (genotype, environment, phenotype) through dynamic knowledge graphs and federated spatiotemporal embedding models, effectively breaking down the data silos of traditional systems. Based on the collaborative optimization of two-layer reinforcement learning and Monte Carlo tree search verification, the system can autonomously balance long-term breeding goals and short-term dynamic responses, improving the efficiency of identifying individuals with high genetic value and shortening the response time for strategy adjustments. Simultaneously, the dual verification mechanism of causal reasoning and extreme adversarial testing enhances decision confidence and significantly reduces resource waste and risk lag caused by experience-based decision-making.

[0125] Through a dynamic environmental path generator and quantitative evaluation of strategy robustness, the system can simulate complex scenarios such as epidemics, extreme weather, and market fluctuations. Based on a coefficient of variation threshold, such as egg production rate CV < 0.25, it automatically optimizes strategy parameters to ensure improved trait stability under perturbation events. Furthermore, the intelligent protocol adapter and 3D visualization interactive design create a closed-loop system across the entire industry chain, integrating gene sequencing, environmental monitoring, and market data. This improves the alignment between breeding strategies and industry needs, and optimizes resource utilization. This solution combines technological foresight with practical feasibility, providing core support for the intelligent upgrading of the poultry breeding industry.

[0126] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0127] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A poultry breeding record analysis system, characterized in that, include: Multimodal data dynamic fusion module, adaptive dynamic modeling module, multi-mechanism verification and screening module, system collaboration interface module; The multimodal data dynamic fusion module is used to dynamically construct knowledge graphs and perform conflict removal on genotype data, environmental time-series data, and phenotypic data; the multimodal data dynamic fusion module includes: The dynamic ontology modeling unit defines three-dimensional association rules for genetic lineage, environmental parameters, and phenotypic indicators based on a dedicated ontology library for poultry breeding. The knowledge graph evolution unit uses the spatiotemporal sequence clustering algorithm ST-DBSCAN to identify the temporal coupling pattern between environmental parameters and phenotypic data, and encodes it as dynamic weighted edges; The dual-channel data cleaning unit sets hard filtering thresholds for genotype and phenotype data through a rule engine, and performs soft correction on sensor time-series data based on a generative adversarial network (GAN). The adaptive dynamic modeling module implements multimodal spatiotemporal embedding modeling based on a federated learning framework and generates dynamic breeding strategies through a two-layer optimization decision engine; the adaptive dynamic modeling module includes: The heterogeneous federated spatiotemporal embedding model Hetero-FSTE maps genotype SNP sequences, phenotypic time series and environmental tensors to a unified vector space, and uses the gated spatiotemporal attention mechanism GSTA to capture gene-environment interaction effects. The vertical federated learning architecture allows local servers to retain the original data and only upload encrypted model gradients to the central server for aggregation, which is used for cross-entity data privacy protection and format heterogeneity compatibility. The dual-layer optimization decision engine includes a global deep Q network (DQN) and a real-time proximal strategy optimization (PPO) algorithm. The global deep Q network generates long-term breeding target genotype combinations based on historical data. The real-time proximal strategy optimization algorithm receives environmental monitoring data to dynamically adjust the objective function weights and verifies strategy compatibility through Monte Carlo Tree Search (MCTS). The multi-mechanism verification and screening module uses causal reasoning and adversarial stress testing to perform dual verification of the strategy; The system collaboration interface module enables protocol adaptation and visual decision-making interaction for multi-source heterogeneous devices.

2. The poultry breeding record analysis system according to claim 1, characterized in that: The Monte Carlo Tree Search (MCTS) verification process includes: Simulate the path of environmental fluctuations and market changes over the next 30 days before deploying the strategy; The stability of the target traits and the rate of increase in resource consumption of the evaluation strategy under the simulated path; When the resource consumption growth rate exceeds the preset threshold, the real-time layer strategy weight reallocation is automatically triggered. The Monte Carlo Tree Search (MCTS) verification process further includes: A dynamic environmental path generator, based on historical environmental data and real-time weather forecast API, constructs a probability distribution model of temperature, humidity, and light intensity for the next 30 days. The strategy robustness evaluation unit quantifies strategy stability through the following steps: 1) Inject random disturbance events into the simulated path; 2) Calculate the coefficient of variation of the target trait before and after the perturbation. If the coefficient of variation exceeds the preset threshold, the strategy is deemed to be insufficiently robust. 3) Based on the evaluation results, automatically adjust the exploration-utilization balance parameter in the real-time near-end policy optimization PPO algorithm to prioritize policy branches that are not sensitive to perturbations.

3. The poultry breeding record analysis system according to claim 2, characterized in that: The multi-mechanism verification and screening module includes: The counterfactual causal verification unit constructs counterfactual scenarios of a strategy based on the potential outcome model (PO Model), compares actual data with counterfactual prediction results, and quantifies the confidence level of the causal effect of the strategy. Extreme-AGAN, an extreme scenario adversarial generator, simulates genotype mutations, drastic environmental changes, and sudden market shifts. It uses Shapley value analysis to locate the vulnerability of the strategy and trigger model retraining.

4. The poultry breeding record analysis system according to claim 3, characterized in that: The counterfactual causality verification unit performs the following operations: Construct counterfactual hypothetical scenarios for the system's recommendation strategy and calculate the contribution of the differences before and after the strategy implementation. When the confidence level of the causal effect falls below a preset threshold, the strategy rollback and model iteration update are automatically triggered. The difference contribution was calculated by combining Bayesian network and intervention factor decomposition.

5. The poultry breeding record analysis system according to claim 4, characterized in that: The system collaboration interface module includes: The Intelligent Protocol Adapter (IPA) supports second-level protocol parsing for gene sequencing platform VCF file interfaces, intelligent environmental control equipment Modbus protocols, and market data APIs. A 3D breeding trend map dynamically visualizes genotype diffusion trends, environmental hotspot distribution, and strategy benefit heatmaps, and provides an interactive optimization interface for human expert annotation and AI strategies.

6. The poultry breeding record analysis system according to claim 5, characterized in that: The dual-channel data cleaning unit includes: An adversarial neural network discriminator performs secondary noise detection on the data that has been initially cleaned by the rule engine. A dynamic weight allocator adjusts the cleaning intensity based on the credibility of the data source, where the data source credibility is ranked as follows: gene sequencing platform > manual input > edge sensor. The data repair results are finally arbitrated by comparing the predictions of the Bi-LSTM (Bi-Long Short-Term Memory) network with the actual values.

7. A method for analyzing poultry breeding records, characterized in that, Includes the following steps: S1: Dynamic knowledge graph construction and conflict cleanup are performed on genotype data, environmental time-series data, and phenotypic data; three-dimensional association rules for genetic lineage, environmental parameters, and phenotypic indicators are defined based on a dedicated ontology library for poultry breeding; the spatiotemporal sequence clustering algorithm ST-DBSCAN is used to identify the temporal coupling patterns between environmental parameters and phenotypic data and encode them as dynamic weighted edges; hard filtering thresholds for genotype and phenotypic data are set through a rule engine, and soft corrections are performed on sensor time-series data based on generative adversarial networks (GANs); S2: Multimodal spatiotemporal embedding modeling is realized based on the federated learning framework, and dynamic breeding strategies are generated through a two-layer optimization decision engine; genotype SNP sequences, phenotypic time series and environmental tensors are mapped to a unified vector space, and the gated spatiotemporal attention mechanism GSTA is used to capture gene-environment interaction effects; The vertical federated learning architecture allows local servers to retain the original data and only upload encrypted model gradients to the central server for aggregation; the two-layer optimization decision engine includes a global deep Q-network (DQN) and a real-time proximal policy optimization (PPO) algorithm; the global deep Q-network generates long-term breeding target genotype combinations based on historical data; the real-time proximal policy optimization algorithm receives environmental monitoring data to dynamically adjust the objective function weights and verifies policy compatibility through Monte Carlo Tree Search (MCTS). S3: The strategy is validated by both causal reasoning and adversarial stress testing; S4: Perform protocol adaptation and visual decision-making interaction for multi-source heterogeneous devices.

Images