A method for predicting health risk of ruminants based on methionine metabolism

Abstract

The application provides a ruminant health risk prediction method based on methionine metabolism, comprising obtaining the methane concentration data of the target ruminant in the time sequence of exhaled gas, the soft expansion data of the ruminant food, the methionine content data and the lysine content data, and constructing a target prediction data set; calling a pre-trained metabolic correlation model, constructing a dynamic correlation relationship between the methionine metabolism characteristics and the health status of the ruminant, and outputting a dynamic correlation model representing the relationship between the metabolism characteristics and the health status; inputting the target prediction data set into the dynamic correlation model, and outputting a metabolic risk correlation characteristic set; dynamically adapting the metabolic risk correlation characteristic set to the risk mode in the health risk mode library, and outputting a ruminant health risk prediction result; and based on the metabolic risk correlation characteristic set and the ruminant health risk prediction result, generating health intervention guidance information for methionine metabolism regulation of the target ruminant.

Description

Technical Field

[0001] This disclosure relates to the field of computer processing technology, and in particular to a method for predicting health risks in ruminants based on methionine metabolism. Background Technology

[0002] In ruminant farming, accurately predicting health risks is crucial for ensuring profitability and animal welfare. The health of ruminants is influenced by a variety of factors, among which methionine metabolism plays a key role in maintaining normal physiological functions. As an essential amino acid, the synthesis, transformation, and accumulation of methionine metabolites are closely linked to the health of ruminants.

[0003] However, current methods for predicting health risks in ruminants often focus on single indicators or simple symptom observations, failing to comprehensively and accurately reflect the complex relationship between methionine metabolism and health status. Furthermore, the lack of effective methods to uncover key health-related information hidden within methionine metabolism data results in insufficient accuracy and timeliness in health risk prediction, hindering the provision of scientifically sound health intervention guidance for farmers and failing to meet the needs of refined management in ruminant farming. Summary of the Invention

[0004] In view of the aforementioned problems, this disclosure provides a method for predicting the health risks of ruminants based on methionine metabolism, aiming to improve the accuracy of predicting the health risks of ruminants, thereby improving the accuracy of guiding health intervention operations on the regulation of methionine metabolism in ruminants.

[0005] In conjunction with the first aspect of the present invention, an embodiment of the present invention provides a method for predicting the health risk of ruminants based on methionine metabolism, comprising: acquiring time-series data on methane concentration in exhaled gas, softness and expansion data of ruminant food, methionine content data, and lysine content data of a target ruminant, and constructing a target prediction data set; calling a pre-trained metabolic association model, inputting the target prediction data set into the metabolic association model, constructing a dynamic correlation between methionine metabolic characteristics and the health status of ruminants, and outputting a dynamic association model characterizing the relationship between metabolic characteristics and health status; inputting the target prediction data set into the dynamic association model, mining metabolic risk association features in the target prediction data set that are associated with the health status, and outputting a set of metabolic risk association features; loading a pre-stored health risk pattern library, dynamically adapting the set of metabolic risk association features to risk patterns in the health risk pattern library, and outputting a prediction result of the health risk of ruminants; and generating health intervention guidance information for regulating methionine metabolism in target ruminants based on the set of metabolic risk association features and the prediction result of the health risk of ruminants, wherein the health intervention guidance information is used to guide health intervention operations for regulating methionine metabolism in target ruminants.

[0006] In conjunction with a second aspect of the present invention, an embodiment of the present invention provides a ruminant health risk prediction system based on methionine metabolism, comprising: a memory having a computer program stored thereon; and a processor for executing the computer program in the memory to implement the steps of any of the methods in the first aspect.

[0007] Through the above-described technical solution, this disclosure can achieve at least the following effective effects:

[0008] By acquiring multi-dimensional data such as exhaled methane concentration and the softening and expansion of ruminant food over time from target ruminants, a target prediction dataset is constructed, which can comprehensively and accurately reflect the animal's physiological state. A pre-trained metabolic correlation model is used to construct a dynamic correlation between methionine metabolic characteristics and health status, outputting a dynamic correlation model that allows for in-depth analysis of the intrinsic relationship between the two. Inputting data into the dynamic correlation model to mine metabolic risk correlation features can accurately pinpoint key factors affecting health. Loading a health risk model library for dynamic adaptation outputs health risk prediction results, providing early warnings of potential health problems. Finally, based on the above results, health intervention guidance information on methionine metabolism regulation is generated, improving the accuracy of ruminant health risk prediction and thus enhancing the accuracy of health intervention operations for methionine metabolism regulation in ruminants. This provides farmers with scientific, precise, and personalized intervention plans, reducing farming risks.

[0009] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0010] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings:

[0011] Figure 1 This is a schematic diagram of the execution flow of the ruminant health risk prediction method based on methionine metabolism provided in this embodiment of the invention.

[0012] Figure 2 This is the implementation provided by the embodiments of the present invention. Figure 1 A schematic diagram of the execution flow of step S12.

[0013] Figure 3 This is the implementation provided by the embodiments of the present invention. Figure 1 A schematic diagram of the execution flow of step S14.

[0014] Figure 4 This is the implementation provided by the embodiments of the present invention. Figure 1 A schematic diagram of the execution flow of step S15.

[0015] Figure 5 This is a schematic diagram of exemplary hardware and software components of the ruminant health risk prediction system based on methionine metabolism provided in an embodiment of the present invention. Detailed Implementation

[0016] 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.

[0017] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0018] This invention provides a method for predicting health risks in ruminants based on methionine metabolism. (See also...) Figure 1 As shown, the method includes:

[0019] In step S11, data on methane concentration in the exhaled gas of the target ruminant, softness and expansion of ruminant food, methionine content, and lysine content are obtained over time to construct a target prediction dataset.

[0020] The target ruminant is a specific ruminant animal, such as a cow or sheep, used to study its health status and metabolic characteristics. Methane concentration data reflects the methane content in the exhaled breath of the target ruminant. Softness and expansion data describe the softness and expansion of ruminant food. Methionine content data reflects the content of the amino acid methionine in ruminant food. Lysine content data indicates the content of the amino acid lysine in ruminant food.

[0021] In this embodiment, firstly, methane concentration data can be collected in real time during ruminant exhalation using a gas detection device. Softness and swelling data can be obtained by observing the state of the food during feeding and measuring its physical properties using relevant instruments. Methionine and lysine content data can be determined from ruminant feed samples using chemical analysis methods. The different types of data are organized and categorized according to certain rules and formats, ultimately constructing a complete target prediction dataset to provide basic data support for subsequent analysis.

[0022] For example, when studying dairy cows on a farm, a gas detector is used to collect data on the methane concentration in the cows' exhaled breath. The feed is observed while the cows are eating, and data on the softness and swelling are obtained using measuring tools. Feed samples are collected and sent to the laboratory for analysis to obtain data on the methionine and lysine content. Then, these data are integrated into a target prediction dataset.

[0023] In step S12, the pre-trained metabolic association model is invoked, the target prediction dataset is input into the metabolic association model, a dynamic association relationship between methionine metabolic characteristics and the health status of ruminants is constructed, and a dynamic association model representing the relationship between metabolic characteristics and health status is output.

[0024] The dynamic correlation is used to reflect the relationship between methionine metabolic characteristics and the health status of ruminants over time or under different conditions. The dynamic correlation model is a mathematical model used to present the relationship between methionine metabolic characteristics and the health status of ruminants.

[0025] In this embodiment, a pre-trained metabolic association model is invoked. The target prediction dataset is input into the metabolic association model, which, based on its pre-defined algorithms and rules, mines and analyzes the information between methionine metabolic characteristics and the health status of ruminants. By comparing the characteristics and states of different data points, the potential relationship between methionine metabolic characteristics and the health status of ruminants is identified. Based on this potential relationship, a dynamic association model is constructed to reflect the dynamic changes in methionine metabolic characteristics and the health status of ruminants. Finally, a dynamic association model is output, intuitively demonstrating the relationship between the two.

[0026] For example, a target prediction dataset including various metabolic data of dairy cows is input into a metabolic association model. After analysis, the metabolic association model identifies the relationship between methionine metabolism and dairy cow health, and outputs a dynamic association model.

[0027] In step S13, the target prediction data set is input into the dynamic association model, metabolic risk association features that are associated with the health status in the target prediction data set are mined, and a metabolic risk association feature set is output.

[0028] Among them, metabolic risk association features are used to represent features in the target prediction dataset that are associated with the health status of ruminants and may cause metabolic risks.

[0029] In this embodiment, the target prediction dataset is input into a dynamic association model, which has already clearly defined the relationship between methionine metabolic characteristics and the health status of ruminants. The dynamic association model can scan and analyze each data feature in the target prediction dataset, and based on the previously established association relationships, determine the target features that are closely related to the health status of ruminants. In this way, metabolic risk association features with special significance are extracted from the target prediction dataset, and these features are summarized to form a metabolic risk association feature set.

[0030] In step S14, a pre-stored health risk pattern library is loaded, and the metabolic risk-related feature set is dynamically matched with the risk patterns in the health risk pattern library to output the health risk prediction results for ruminants.

[0031] The health risk model library pre-stores a set of various health risk models, including characteristics and manifestations under different risk models. Dynamic adaptation is a real-time, flexible matching process that matches the set of metabolic risk-related features with the risk models in the health risk model library. The health risk prediction result is a prediction of the health risk of ruminants obtained after dynamic adaptation.

[0032] In this embodiment, a pre-stored health risk pattern library is loaded. This library includes various known health risk patterns, each with corresponding characteristics and manifestations. Dynamically matching the set of metabolic risk-related features with the risk patterns in the health risk pattern library involves comparing and analyzing the metabolic risk-related features with the features of each risk pattern one by one. Based on the similarity and matching degree between features, the risk pattern that best matches the metabolic risk-related features is identified, thereby determining the potential health risks faced by ruminants, outputting health risk prediction results, and providing a basis for taking appropriate intervention measures.

[0033] In step S15, based on the metabolic risk association feature set and the ruminant health risk prediction results, health intervention guidance information for regulating methionine metabolism in the target ruminant is generated. The health intervention guidance information is used to guide the health intervention operation for regulating methionine metabolism in the target ruminant.

[0034] Among them, the health intervention guidance information is generated based on the metabolic risk association feature set and the health risk prediction results of ruminants, and is used to guide the health intervention operations on methionine metabolism regulation for target ruminants.

[0035] In this embodiment, based on a set of metabolic risk-associated features and ruminant health risk prediction results, the problems and risks associated with methionine metabolism in ruminants are analyzed. Combining professional knowledge and experience, corresponding health intervention measures and methods are developed to address these problems and risks. These intervention measures and methods are then organized and summarized to generate targeted and actionable health intervention guidance information. This information guides relevant personnel in implementing health interventions to regulate methionine metabolism in target ruminants, thereby improving their health status.

[0036] For example, based on the metabolic risk characteristics and health risk predictions of dairy cows, such as the possibility that abnormal methionine metabolism may lead to certain diseases, intervention measures such as adjusting the methionine content in feed can be developed to form health intervention guidance information.

[0037] The aforementioned technical solution constructs a target prediction dataset by acquiring multi-dimensional data such as the methane concentration in exhaled gases and the softening and expansion of ruminant food over time in target ruminants, comprehensively and accurately reflecting the animal's physiological state. A pre-trained metabolic correlation model is used to construct a dynamic correlation between methionine metabolic characteristics and health status, outputting a dynamic correlation model that allows for in-depth analysis of the intrinsic relationship between the two. Inputting data into the dynamic correlation model to mine metabolic risk correlation features can accurately pinpoint key factors affecting health. Loading a health risk model library for dynamic adaptation outputs health risk prediction results, providing early warnings of potential health problems. Finally, based on the above results, health intervention guidance information on methionine metabolism regulation is generated, improving the accuracy of ruminant health risk prediction and thus enhancing the accuracy of health intervention operations for methionine metabolism regulation in ruminants. This provides farmers with scientific, precise, and personalized intervention plans, reducing farming risks.

[0038] In a preferred embodiment, see Figure 2 As shown, in step S12, the pre-trained metabolic association model is invoked, the target prediction dataset is input into the metabolic association model, a dynamic association relationship between methionine metabolic characteristics and the health status of ruminants is constructed, and a dynamic association model characterizing the relationship between metabolic characteristics and health status is output, including:

[0039] In step S121, methionine metabolic features are extracted from the target prediction dataset. The methionine metabolic features include a first feature of methionine precursors, a second feature of methionine conversion intermediates, and a third feature of methionine metabolic end products.

[0040] Methionine metabolism characteristics are feature information that reflects the metabolic process of methionine in ruminants. The first characteristic of methionine precursors is the characteristics of substances related to methionine in the initial stage of metabolism, reflecting the initial state of metabolism. The second characteristic of methionine conversion intermediates is the characteristics of intermediate products during methionine metabolism, reflecting the intermediate process of metabolism. The third characteristic of methionine metabolic end products is the characteristics of the final products of methionine metabolism, reflecting the final result of metabolism.

[0041] In this embodiment, extracting methionine metabolic features from the target prediction dataset requires first clarifying the methionine metabolic process and related substances. Methionine precursors are the starting point of metabolism; by analyzing their manifestation and related parameters in the data, the first feature is extracted, such as their content range and frequency of occurrence. For methionine conversion intermediates, relevant data indicators, such as the formation rate of intermediates and their distribution in different tissues, are identified based on the metabolic pathway to extract the second feature. For methionine metabolic end products, data on their accumulation and excretion in the body are used to extract the third feature. The entire extraction process requires combining professional biological metabolic knowledge and data analysis methods to ensure that the extracted features accurately represent different stages of methionine metabolism.

[0042] For example, taking beef cattle research as an example, in the target prediction dataset, for methionine precursors, we analyze their initial content in feed and their absorption after entering the beef cattle's body, extracting the first characteristic, such as content changes. For intermediate products, we analyze their generation and transformation data in organs such as the rumen based on metabolic pathways, extracting the second characteristic, such as generation rate. For final products, we analyze their excretion in feces and urine, extracting the third characteristic, such as excretion volume.

[0043] In step S122, the health status information of ruminants corresponding to the target prediction data set is obtained, and the health status information of ruminants corresponds one-to-one with each sample data in the target prediction data set;

[0044] Among them, ruminant health status information is used to describe the physical condition of ruminants, whether they are sick, and other related information.

[0045] In this embodiment of the disclosure, obtaining the health status information of ruminants corresponding to the target prediction data set requires establishing a correspondence between the data and individual animals. This can be achieved by tackling ruminants with identification tags and recording the relevant data collection time and health check time for each animal. During the health check, professional veterinary examination methods, such as clinical examination and blood biochemical index testing, are used to comprehensively assess the health status of the ruminants, including whether they suffer from digestive system diseases, metabolic diseases, etc., and this health status information is then matched one-to-one with each sample data in the target prediction data set.

[0046] For example, in a beef cattle farm, each cattle is fitted with a numbered ear tag. When collecting target prediction data, the ear tag number and collection time are recorded. Afterward, veterinarians conduct regular checks on each cattle, such as measuring body temperature, examining fecal morphology, and testing relevant indicators in the blood, to determine their health status and correlate this information with the sample data corresponding to the ear tag number.

[0047] In step S123, the methionine metabolic characteristics are associated and integrated with the corresponding ruminant health status information to form a training data set for the metabolic association model.

[0048] The training dataset is the dataset used to train the model, enabling the model to learn the relationships between data.

[0049] In this embodiment, methionine metabolic characteristics are correlated and integrated with the corresponding health status information of ruminants to determine the correspondence rules between the two. Following the order of the sample data, the methionine metabolic characteristics of each sample are combined with its corresponding health status information. For example, the first characteristic of a sample's methionine precursor, the second characteristic of its intermediate product, and the third characteristic of its final product are combined with the health status information of the corresponding ruminant to form a data unit. All such data units from all samples are aggregated to form a training data set for the metabolic correlation model.

[0050] In step S124, the training dataset is input into the feature association layer of the pre-trained metabolic association model to perform hierarchical feature transformation on the methionine metabolism features and generate adaptive metabolic features that are suitable for health status association modeling.

[0051] Hierarchical feature transformation involves processing features according to a specific hierarchical structure to better suit modeling needs. Adapted metabolic features, after hierarchical feature transformation, are methionine metabolic features that are more suitable for health status association modeling.

[0052] In this embodiment, the training dataset is input into the feature association layer of a pre-trained metabolic association model. The feature association layer contains pre-defined algorithms and parameters. It first performs preliminary analysis and processing of methionine metabolic features, following different hierarchical structures. For example, it first normalizes precursor substance features to unify data from different ranges into a specific interval. Then, it performs similar transformation operations on intermediate and final product features, adjusting the feature representation based on their potential relationship with health status to generate adaptive metabolic features suitable for health status association modeling, thereby improving the adaptability of features to health status modeling.

[0053] In step S125, the dynamic association module of the metabolic association model is used to perform nonlinear association modeling on the adaptive metabolic features and ruminant health status information, construct a dynamic association relationship between methionine metabolic features and ruminant health status, and output a dynamic association model representing the relationship between metabolic features and health status based on the dynamic association relationship.

[0054] In this embodiment, a dynamic correlation module of the metabolic correlation model is used to perform nonlinear correlation modeling between adapted metabolic characteristics and ruminant health status information. The dynamic correlation module employs a complex algorithm to account for the complex nonlinear relationship between adapted metabolic characteristics and health status. For example, when certain metabolic characteristics change within a certain range, the health status may exhibit different trends. By analyzing this complex relationship, a correlation is constructed that accurately reflects the dynamic changes in methionine metabolic characteristics and ruminant health status. Based on this relationship, a dynamic correlation model characterizing the relationship between metabolic characteristics and health status is output, intuitively demonstrating the dynamic connection between the two.

[0055] The aforementioned technical solution's feature association layer effectively transforms and processes features to generate suitable metabolic features, thus improving data quality. The dynamic association module, through nonlinear association modeling, constructs a dynamic association model that accurately reflects the dynamic relationship between methionine metabolic features and the health status of ruminants. This model can precisely extract methionine metabolic features from complex data and accurately obtain corresponding ruminant health status information, forming a high-quality training dataset. This allows for a more realistic and comprehensive presentation of the relationship between the two, improving the accuracy of metabolic risk association features, health risk prediction, and health intervention guidance, thereby enhancing the level of ruminant health management.

[0056] In a preferred embodiment, in step S125, the dynamic correlation module of the metabolic correlation model performs nonlinear correlation modeling on the adapted metabolic characteristics and ruminant health status information to construct a dynamic correlation relationship between methionine metabolic characteristics and ruminant health status, including:

[0057] In step S1251, the adapted metabolic features are input into the feature encoding submodule of the dynamic association module, and the adapted metabolic features are subjected to dimension-unified encoding processing to generate encoded metabolic features with unified dimension representation.

[0058] The dimension-unified encoding process transforms adapted metabolic features of different dimensions and forms into an encoding representation with the same dimensions. The encoded metabolic features are methionine metabolic features with a unified dimensional representation after the dimension-unified encoding process.

[0059] In this embodiment, the feature encoding submodule first analyzes the various dimensional characteristics of the adapted metabolic features. Different adapted metabolic features may come from different data sources or measurement methods, and have different dimensional structures. For example, some features may be single numerical values, while others may be multi-dimensional vectors. To facilitate subsequent correlation analysis, a unified encoding rule needs to be designed based on the characteristics of these features. For single numerical features, they can be extended into a vector form of a specific dimension; for multi-dimensional vector features, they are transformed into a unified dimensional space according to certain mapping rules. In this way, all adapted metabolic features have the same dimensional representation, generating encoded metabolic features.

[0060] In step S1252, the health status information of the ruminant is input into the status identification submodule of the dynamic association module, and the health status information of the ruminant is standardized and identified to generate a standardized health status identification.

[0061] Standardized labeling is the process of labeling and classifying ruminant health status information according to unified standards. Standardized health status labels, after standardization, accurately and uniformly represent the health status of ruminants.

[0062] In this embodiment, the status identification submodule establishes a complete health status standard system. Based on common health problems and disease types in ruminants, and combined with professional veterinary knowledge, different health status levels and classification standards are determined. For example, health status is divided into different levels such as healthy, sub-healthy, and diseased, and detailed judgment indicators are formulated for each level. When processing ruminant health status information, each piece of health status information is judged and identified according to this standard system, accurately classifying it into the corresponding health status level, generating standardized health status identifiers, and improving the standardization of health status information.

[0063] In step S1253, the feature cross-fusion method is used to mine the association information between the encoded metabolic features and the standardized health status identifier to generate preliminary association features;

[0064] The feature cross-fusion method involves mining the association information between encoded metabolic features and standardized health status identifiers, extracting the potential relationship between the two through cross- and fusion operations. Preliminary associated features, mined using the feature cross-fusion method, are used to reflect the association between encoded metabolic features and standardized health status identifiers.

[0065] In this embodiment, the feature cross-fusion method involves crossing and combining encoded metabolic features with standardized health status identifiers across different dimensions. For example, various dimensions of the encoded metabolic features are mapped and associated with different categories of standardized health status identifiers. During this process, the interactions and influences between the two under different combinations are analyzed, extracting feature information that reflects the potential relationship between methionine metabolism and health status. In this way, the originally independent encoded metabolic features and standardized health status identifiers are linked, generating preliminary associated features.

[0066] For example, cross-analysis can be performed between different dimensions encoding metabolic features and standardized health status markers (healthy, sub-healthy, diseased). For instance, analyzing the changes in the concentration of a certain metabolic intermediate under different health states can extract preliminary correlation features that reflect the relationship between the two.

[0067] In step S1254, the preliminary association features are subjected to association signal enhancement processing to highlight the key association information between the encoded metabolic features and the standardized health status identifier, thereby generating enhanced association features;

[0068] In this embodiment, key features that play a crucial role in the correlation between methionine metabolism and health status are identified, such as those with large variations in feature values ​​and high correlation with changes in health status. These key features are then amplified or strengthened while weakly correlated features are suppressed. For example, a weighted processing method can be used to give higher weights to key features, thereby generating strengthened correlation features and suppressing other weakly correlated features, clearly demonstrating the key link between methionine metabolism features and health status.

[0069] In step S1255, a nonlinear correlation link between methionine metabolism characteristics and ruminant health status is constructed based on the enhanced correlation features, forming a dynamic correlation between methionine metabolism characteristics and ruminant health status.

[0070] Among them, the nonlinear association link is a connection path used to reflect the nonlinear relationship between encoded metabolic features and standardized health status identifiers.

[0071] In this embodiment of the disclosure, based on enhanced correlation features, the complex nonlinear relationship between encoded metabolic features and standardized health status identifiers is analyzed. By identifying the interaction patterns and variation laws among them, a nonlinear correlation link capable of connecting different features and health statuses is constructed.

[0072] These links are not simple linear connections, but rather reflect how changes in methionine metabolic characteristics under different conditions non-linearly affect the health status of ruminants. By constructing non-linear correlation links, a dynamic relationship between methionine metabolic characteristics and ruminant health status is formed, enhancing the realism of the complex connection between the two.

[0073] The aforementioned technical solutions, through feature cross-fusion and correlation signal enhancement processing, can deeply uncover and highlight the key correlation information between the two, making the correlation features more obvious and representative. The nonlinear correlation link constructed based on enhanced correlation features can accurately establish a dynamic correlation between methionine metabolic characteristics and ruminant health status, truly reflecting the complex nonlinear relationship between the two. This effectively standardizes the processing of adaptive metabolic characteristics and ruminant health status information, improving the accuracy of predicting health risks and developing targeted health interventions.

[0074] In a preferred embodiment, step S125, which involves outputting a dynamic correlation model characterizing the relationship between metabolic features and health status based on the dynamic correlation, includes:

[0075] In step S125a, the model parameters corresponding to the dynamic association relationship are extracted to form a model parameter set;

[0076] Model parameters are key values ​​or settings that affect the operation and results of the dynamic association model, used to adjust how the model processes data and the logic for establishing associations. The model parameter set is a collection that integrates all model parameters corresponding to the dynamic associations.

[0077] In this embodiment, the model corresponding to the dynamic association relationship contains numerous model parameters that influence its behavior during operation. These parameters may include feature weights, association thresholds, etc. Extracting these parameters requires in-depth analysis of the model's structure and operating mechanism. For example, in the association establishment part of the model, different features have different weights in the association calculation; these weights are important model parameters. By traversing the various functional modules of the model, all parameters affecting the model results are identified, and these parameters are organized according to certain rules and order to form a model parameter set.

[0078] Repeat the following steps: In step S125b, input the validation sample data from the training dataset into the model structure corresponding to the dynamic association relationship to generate the validation sample prediction result;

[0079] The validation sample data is a subset of the training dataset used to validate the model's performance. The validation sample prediction result is the model's output prediction of the health status of ruminants after the validation sample data is input into the model structure corresponding to the dynamic association relationship.

[0080] In this embodiment, a portion of the training dataset is selected as validation sample data according to a certain proportion and rules. This validation sample data must be representative, covering different methionine metabolic characteristics and health statuses. The selected validation sample data is input into the model structure corresponding to the dynamic correlation relationship. The model analyzes and processes the methionine metabolic characteristics in the validation sample data based on its internal algorithms and parameters, thereby generating validation sample prediction results.

[0081] In step S125c, model prediction bias information is generated based on the health status information corresponding to the verification sample prediction results and the verification sample data.

[0082] Among them, the model prediction bias information is used to reflect the degree of difference between the prediction results of the validation samples and the actual health status information corresponding to the validation sample data.

[0083] In this embodiment, the predicted results of the validation samples are compared one by one with the actual health status information corresponding to the validation sample data. The differences between the predicted and actual results in the judgment of health status are analyzed, such as whether a healthy ruminant is incorrectly predicted as diseased, or whether there is a prediction deviation in the severity of disease. Model prediction deviation information is generated by calculating the degree and proportion of this difference. Various methods can be used to measure the deviation, such as calculating prediction accuracy and misclassification rate, to comprehensively reflect the model's prediction deviation.

[0084] In step S125d, the model prediction deviation information is input into the parameter adjustment unit of the parameter optimization module to determine the target adjustment parameter and the corresponding adjustment direction in the model parameter set;

[0085] The parameter adjustment unit within the parameter optimization module is a functional unit specifically designed for adjusting model parameters. It determines the parameters requiring adjustment and the direction of adjustment based on model prediction deviation information. The target adjustment parameter is the model parameter that needs adjustment, as determined by the prediction deviation information. The adjustment direction indicates whether the target adjustment parameter should be increased or decreased.

[0086] In this embodiment, after receiving model prediction deviation information, the parameter adjustment unit can analyze the cause of the deviation and its relationship with the model parameters. Through preset algorithms and rules, it determines which model parameters have a greater impact on the deviation and identifies them as target adjustment parameters. Simultaneously, based on the specific circumstances of the deviation, it determines the adjustment direction for these target adjustment parameters. For example, if it is found that the model's prediction deviation for a certain health state is large and related to the weight parameter of a certain metabolic feature, this weight parameter may be identified as a target adjustment parameter, and the unit may decide whether to increase or decrease its weight value based on the deviation.

[0087] In step S125e, the target adjustment parameters are adjusted by gradient according to the adjustment direction to generate the adjusted model parameters;

[0088] Gradient adjustment is a method that gradually adjusts the target parameters according to a certain gradient magnitude to precisely optimize the model parameters. The adjusted model parameters are the new model parameter values ​​obtained after the gradient adjustment operation.

[0089] In this embodiment, after determining the target adjustment parameters and adjustment direction, the parameters are adjusted according to a certain gradient. The magnitude of the gradient needs to be set reasonably based on the characteristics and bias of the model. If the gradient is too large, it may lead to over-adjustment of the parameters and unstable model performance; if the gradient is too small, the adjustment efficiency will decrease. According to the selected gradient, the target adjustment parameters are gradually adjusted along the determined adjustment direction. The parameter value is recorded after each adjustment. Through multiple adjustments, the optimal parameter value that can reduce the prediction bias of the model is found, and the adjusted model parameters are generated.

[0090] In step S125f, the adjusted model parameters are updated to the model structure corresponding to the dynamic association relationship to form the updated dynamic association model;

[0091] In this embodiment, the adjusted model parameters are updated into the model structure corresponding to the dynamic correlation according to the model's structural requirements. This involves modifying and updating the internal parameter storage location of the model to ensure that the model can use the new parameters for calculation and analysis in subsequent runs. When processing methionine metabolic characteristics and establishing correlations with health status, the updated model will be adjusted based on the new parameters, thereby improving the model's predictive accuracy and forming an updated dynamic correlation model.

[0092] In step S125g, the test sample data from the training dataset is input into the updated dynamic association model to generate test sample prediction results. The model parameters are adjusted based on the test sample prediction results until the model prediction deviation information reaches the preset standard, and then the dynamic association model is output.

[0093] The test sample data is another set of data samples selected from the training dataset to test the performance of the updated model. The test sample prediction result is the prediction information about the health status of ruminants output by the model after the test sample data is input into the updated dynamic association model.

[0094] In this embodiment, a different portion of data from the training dataset, distinct from the validation sample data, is selected as the test sample data. This test sample data is input into the updated dynamic association model. The model analyzes and processes the test sample data according to its internal algorithm and new parameters, generating a test sample prediction result. Then, the test sample prediction result is compared with the actual health status information corresponding to the test sample data, and the model's prediction deviation is evaluated based on the comparison results. If the deviation does not reach a preset standard, the model parameters are optimized based on the deviation, and the above steps are repeated until the model's prediction deviation reaches the preset standard, at which point the final dynamic association model is output.

[0095] The above technical solution utilizes validation and test sample data to repeatedly validate and optimize the model. By analyzing model prediction deviation information, it accurately determines the target adjustment parameters and adjustment direction, and gradually optimizes the model parameters using a gradient adjustment method. Through multiple iterations of optimization, the dynamic correlation model can more accurately establish the dynamic correlation between methionine metabolic characteristics and the health status of ruminants, thereby improving the model's accuracy in predicting the health status of ruminants.

[0096] In a preferred embodiment, see Figure 3 As shown, in step S14, loading the pre-stored health risk pattern library, dynamically adapting the metabolic risk association feature set with the risk patterns in the health risk pattern library, and outputting the ruminant health risk prediction results include:

[0097] In step S141, a pre-stored health risk pattern library is loaded, and the correlation fit between each metabolic risk association feature in the metabolic risk association feature set and each risk pattern in the health risk pattern library is calculated.

[0098] Among them, the correlation fit is an indicator used to measure the degree of matching between metabolic risk correlation characteristics and each risk pattern in the health risk pattern library.

[0099] In this embodiment of the disclosure, after loading the health risk pattern library, for each metabolic risk-related feature in the metabolic risk-related feature set, it is compared with each risk pattern in the health risk pattern library. The various attributes of the metabolic risk-related feature, such as the feature's numerical range and trend, are analyzed, along with the manifestation of the corresponding feature in the risk pattern. The degree of matching between the two is quantified by calculating, for example, the similarity and overlap between the features. The quantified result of the matching degree is used as the association fit degree to assess the closeness of the association between the metabolic risk-related feature and each risk pattern.

[0100] In step S142, the target risk pattern that best matches each of the metabolic risk association features is selected based on the correlation fit.

[0101] Among them, the target risk pattern is the health risk pattern that best matches each metabolic risk associated feature, selected based on the calculation of the correlation fit with the metabolic risk associated features.

[0102] In this embodiment, based on the calculated correlation fit, each metabolic risk association feature is screened against risk patterns in the health risk pattern library. A reasonable fit threshold is set, or a ranking method is used to select the risk pattern with the highest correlation fit to each metabolic risk association feature from among numerous risk patterns. For example, all risk patterns can be ranked from high to low according to their correlation fit with a certain metabolic risk association feature, and the highest-ranked risk pattern is selected as the target risk pattern corresponding to that metabolic risk association feature, ensuring that the health risk pattern that best matches the performance of the metabolic risk association feature is found.

[0103] In step S143, health risk information corresponding to the target risk pattern is extracted, and a ruminant health risk prediction result is generated by combining the correlation fit. The ruminant health risk prediction result includes a health risk scenario description and the correlation fit.

[0104] The health risk information is a description of the health risks to ruminants corresponding to the target risk model, including the risk type and possible diseases. The health risk prediction result is generated by integrating the health risk information corresponding to the target risk model and the correlation fit, and includes the prediction of health risks to ruminants, covering the description of health risk scenarios and the correlation fit.

[0105] In this embodiment, after extracting the health risk information corresponding to the target risk pattern, it is integrated with the correlation fit. The health risk scenario description is a detailed description of the health risks that ruminants may face, written based on the characteristics and relevant knowledge of the target risk pattern. The health risk scenario description and the correlation fit are used together as the content of the health risk prediction result, so that the prediction result can not only clearly indicate the possible health risks, but also reflect the reliability of such risks through the correlation fit.

[0106] The aforementioned technical solution extracts health risk information from target risk patterns and combines it with correlation fit to generate health risk prediction results. This ensures that the prediction results include both detailed descriptions of health risk scenarios and reflect the reliability of the risks. It fully utilizes a pre-stored health risk pattern library to accurately calculate the correlation fit between metabolic risk association features and each risk pattern, thereby precisely identifying target risk patterns. This helps livestock farmers take targeted preventative measures in advance.

[0107] In a preferred embodiment, in step S141, calculating the correlation fit between each metabolic risk association feature in the metabolic risk association feature set and each risk pattern in the health risk pattern library includes:

[0108] In step S1411, each metabolic risk association feature in the metabolic risk association feature set is aligned with each risk pattern in the health risk pattern library to generate an aligned feature-pattern feature pair.

[0109] Feature dimension alignment involves adjusting the features of metabolic risk association features and risk patterns in the health risk pattern library to the same dimensional space, making them comparable during analysis. Feature-pattern feature pairs are combinations of metabolic risk association features and their corresponding risk pattern features that share the same dimension after feature dimension alignment.

[0110] In this embodiment, the metabolic risk association features and the risk patterns in the health risk pattern library may come from different data sources or have different feature representations, leading to inconsistencies in feature dimensions. Feature dimension alignment processing first analyzes the feature dimension structure of both to determine the dimensions that need adjustment. For metabolic risk association features, if their dimension is lower than that of the risk pattern features, the dimension is increased using methods such as interpolation; if it is higher, dimensionality reduction is performed, such as using principal component analysis to extract the main feature dimensions. A similar operation is performed on the risk pattern features to ensure that both are in the same dimensional space. Then, the processed metabolic risk association features are combined with the corresponding risk pattern features to form feature-pattern feature pairs.

[0111] In step S1412, the feature interaction unit performs feature interaction operations on the feature and pattern feature pairs to generate an interaction feature matrix;

[0112] Among them, the interaction feature matrix is ​​a matrix generated after the feature interaction unit operation, which can reflect the interaction relationship between feature and pattern feature pairs.

[0113] In this embodiment, the feature interaction unit employs a specific algorithm to perform interaction operations on feature and pattern feature pairs. For example, a convolutional neural network is used, with convolution kernels performing sliding convolution operations on feature and pattern feature pairs to extract interaction information between local features. The size and number of convolution kernels can be adjusted according to actual conditions to fully capture interaction relationships at different levels. Alternatively, a multilayer perceptron can be used, taking feature and pattern feature pairs as input and performing nonlinear transformations through multiple hidden layers to uncover complex nonlinear interaction relationships between them. After these operations, an interaction feature matrix is ​​generated, where each element reflects the degree of interaction between feature and pattern feature pairs in a specific dimension.

[0114] In step S1413, the interaction feature matrix is ​​input into the correlation calculation unit to perform element-level correlation analysis on the interaction feature matrix and extract the key elements reflecting the degree of correlation in the interaction feature matrix.

[0115] Among them, the key elements are those in the interaction feature matrix that can significantly reflect the degree of association between metabolic risk association features and health risk patterns.

[0116] In this embodiment, the correlation calculation unit first normalizes the interaction feature matrix, mapping the element values ​​to a specific range, such as [0,1], to facilitate subsequent analysis. Then, a correlation analysis method, such as Pearson correlation coefficient calculation, is used to analyze the correlation between each element and the overall correlation degree. For elements with high correlation, their changing trends under different conditions are further analyzed to determine which elements can stably reflect the association between metabolic risk association features and health risk patterns. By setting a threshold, elements with correlations higher than the threshold are selected as key elements. These key elements highlight the information in the interaction feature matrix that is most relevant to the correlation degree.

[0117] In step S1414, based on the distribution and feature strength of the key elements, a correlation metric is generated between each metabolic risk association feature and each risk pattern in the health risk pattern library.

[0118] The key element distribution refers to the positional and numerical distribution of key elements within the interaction feature matrix. Feature strength represents the contribution of the features represented by the key elements to the degree of association. The association metric is generated by combining the key element distribution and feature strength, and is used to measure the degree of association between metabolic risk association features and health risk patterns.

[0119] In this embodiment, the distribution of key elements in the interaction feature matrix is ​​analyzed. If key elements are concentrated in certain regions of the matrix, it indicates that the feature dimensions corresponding to these regions have a significant impact on the degree of association. Simultaneously, the feature strength of the key elements is considered, and its strength is determined by calculating factors such as the numerical value of the key elements and their proportion in the overall matrix. Combining the distribution and feature strength of the key elements, a weighted summation method is used to generate an association metric. For example, key elements with concentrated distribution and high feature strength are assigned larger weights, while key elements with dispersed distribution and low feature strength are assigned smaller weights. This calculation yields an association metric that accurately reflects the degree of association between metabolic risk association features and health risk patterns.

[0120] In step S1415, the correlation fit between each metabolic risk association feature and each risk pattern in the health risk pattern library is generated based on the correlation metric value between each metabolic risk association feature and each risk pattern in the health risk pattern library.

[0121] In this embodiment, based on the correlation metric values ​​between each metabolic risk association feature and each risk pattern in the health risk pattern library, a normalization process is used to map the correlation metric values ​​to the [0, 1] interval, obtaining the relative correlation fit. The normalization process can use the maximum-minimum normalization method, i.e., subtracting the minimum value from the correlation metric value and then dividing by the difference between the maximum and minimum values. After this processing, the correlation fit can intuitively reflect the degree of correlation matching between metabolic risk association features and health risk patterns; the closer the value is to 1, the higher the degree of correlation matching, and the closer it is to 0, the lower the degree of correlation matching.

[0122] The feature dimension alignment processing described above ensures comparability between the two during analysis. The feature interaction unit and correlation calculation unit delve into the potential relationships between them, extract key elements, and generate correlation metrics, ultimately obtaining an accurate correlation fit. This process enables a comprehensive and detailed analysis of the relationship between metabolic risk and health risk patterns. It achieves precise calculation of the correlation fit between metabolic risk features and health risk patterns, helping farmers to more accurately understand the health status of ruminants, take targeted preventative measures in advance, reduce disease incidence, and improve farming efficiency.

[0123] In a preferred embodiment, step S143, which involves extracting health risk information corresponding to the target risk pattern and generating a ruminant health risk prediction result based on the correlation fit, includes:

[0124] In step S1431, the structural information of the target risk pattern is parsed to generate the core elements in the target risk pattern used to describe the health risk scenario;

[0125] Structural information is used to represent the various components of the target risk model and their organizational structure and hierarchical relationships. The core element is the key component within the target risk model's structural information that directly describes the health risk scenario.

[0126] In this embodiment, the target risk pattern is analyzed in depth, starting from its stored data structure or knowledge representation. If the target risk pattern is stored in a tree structure, it is traversed layer by layer from the root node, analyzing the attributes and functions of each node to identify the nodes that play a key role in describing the health risk scenario; the content represented by these nodes is the core element. If the target risk pattern is a rule-based system, the conditions and conclusions of each rule are analyzed, and the conditions and conclusions directly related to the health risk scenario are extracted as the core elements. This accurately identifies the core elements in the target risk pattern used to describe the health risk scenario.

[0127] In step S1432, the core elements are organized according to logical relationships to form a standardized health risk scenario description;

[0128] Logical relationships refer to the rules and order of interrelationships such as causality, sequence, and parallelism among the core elements. Standardized health risk scenario descriptions are text or data structures that accurately and clearly describe health risk scenarios according to unified and standardized formats and requirements.

[0129] In this embodiment, the logical relationships between core elements are analyzed. For example, some core elements may be the causes of health risks, while others are the results, indicating a causal relationship. Some core elements may also appear simultaneously, exhibiting a parallel relationship. Based on these logical relationships, natural language processing techniques or specific data structure organization methods are used to organize the core elements in a reasonable order and format. For instance, causal elements are described first, followed by result elements. Parallel elements can be distinguished using specific conjunctions or symbols, forming a standardized description that conforms to unified specifications and accurately reflects the health risk scenario.

[0130] In step S1433, the correlation fit between the metabolic risk association features and the target risk pattern is converted into a risk confidence level expression;

[0131] Among them, the risk confidence level is expressed by using specific numerical values ​​or ranges to represent the reliability and probability of the association between metabolic risk association characteristics and target risk patterns.

[0132] In this embodiment, the association fit reflects the degree of association between metabolic risk association characteristics and target risk patterns. However, to more intuitively express the probability of risk occurrence, it needs to be converted into a risk confidence level. A probability transformation model is used to establish a mapping relationship between association fit and risk confidence level based on a large amount of historical data and experimental results. For example, a transformation function is determined by statistically analyzing the proportion of actual health risks occurring under different association fits. When the association fit reaches a certain value, the corresponding risk confidence level is calculated using this transformation function. Simultaneously, considering the distribution and uncertainty of the data, the calculated risk confidence level is corrected and optimized to more accurately reflect the actual situation.

[0133] In step S1434, the health risk scenario description and the risk confidence statement are associated and integrated, and sample identification information and prediction time information are added;

[0134] Among them, sample identification information is used to uniquely identify individual ruminants, such as animal ID number, ear tag number, etc. Prediction time information: the specific time point or time period for which health risk prediction is conducted.

[0135] In this embodiment, the description of health risk scenarios and the expression of risk confidence are linked and integrated to ensure accurate correspondence between the two. Simultaneously, to effectively manage and trace the prediction results, sample identification information is added, and the unique identifier of the ruminant is bound to the prediction result through database query or data association technology. Furthermore, prediction time information is recorded, which can be a system timestamp or a user-specified time format, and stored together with the prediction result.

[0136] In step S1435, the integrated information is processed in a structured manner to generate the ruminant health risk prediction result, which includes the health risk scenario description, the risk confidence statement, the sample identification information, and the prediction time information.

[0137] Structured processing involves organizing and storing the integrated information according to a predetermined data structure or format, giving it a clear hierarchy and relationships.

[0138] In this embodiment, based on a pre-designed data structure, such as JSON, XML format, or database table structure, the health risk scenario description, risk confidence level statement, sample identification information, and prediction time information are filled and organized according to their corresponding fields. During the structured processing, the integrity and accuracy of the data are ensured, and the data type and value range of each field are validated. For example, the risk confidence level statement should be a numeric type and within the range [0,1]. Through structured processing, the generated ruminant health risk prediction results can be easily stored, queried, and shared in different systems and applications.

[0139] The above technical solution analyzes the target risk pattern structure to obtain core elements and forms a standardized scenario description, clearly presenting the health risk status; it converts the correlation fit into a risk confidence expression, intuitively reflecting the probability of risk occurrence; it integrates and adds sample and prediction time information to ensure the traceability of results; finally, the structured processing makes the results easy to store and share, and can accurately generate health risk prediction results for ruminants.

[0140] In a preferred embodiment, see Figure 4 As shown, in step S15, the generation of health intervention guidance information for methionine metabolism regulation in target ruminants based on the metabolic risk association feature set and ruminant health risk prediction results includes:

[0141] In step S151, the health risk scenario description and correlation fit information in the ruminant health risk prediction results are analyzed to determine the current health risk scenario and risk correlation strength of the ruminant.

[0142] Among them, the correlation fit information is a data indicator used to reflect the degree of correlation between metabolic risk association characteristics and health risk patterns. The risk association strength is a numerical value used to quantify the degree of correlation between health risk scenarios and metabolic risk association characteristics.

[0143] In this embodiment, the health risk prediction results for ruminants are analyzed in depth. Natural language processing techniques (if the scenario description is in text form) or data parsing algorithms (if the data is structured) are used to extract key information from the health risk scenario description, clarifying the current health risk scenario of the ruminant, such as whether it suffers from a certain metabolic disease. Simultaneously, correlation fit information is obtained from the prediction results, which is typically presented in numerical form. To more intuitively measure the degree of risk correlation, the correlation fit is converted into risk correlation strength according to a preset conversion rule. For example, a threshold range for correlation fit is set, and based on the specific position of the correlation fit within this range, it is mapped to a risk correlation strength value between 0 and 1, with a larger value indicating a stronger correlation.

[0144] In step S152, core metabolic risk association features corresponding to the health risk scenario are extracted from the set of metabolic risk association features, and the metabolic impact path of the core metabolic risk association features is determined.

[0145] Among these, core metabolic risk-associated features are those most closely related to the current health risk scenario and have a significant impact on health risk within the set of metabolic risk-associated features. Metabolic impact pathways refer to the biochemical reaction pathways and action links that core metabolic risk-associated features undergo during metabolism in ruminants.

[0146] In this embodiment of the disclosure, a selection is made from a set of metabolic risk-related features based on the identified health risk scenarios. By calculating the correlation coefficient between each metabolic risk-related feature and the health risk scenario, such as the Pearson correlation coefficient, features with higher correlation coefficients are selected as core metabolic risk-related features. When determining metabolic impact pathways, a biological metabolic knowledge graph is used, which contains information on various metabolic reactions and material transformation relationships within ruminants.

[0147] Starting with core metabolic risk association features, searches and reasoning are performed within a knowledge graph to identify their transmission pathways and targets within the metabolic network. For example, if a core metabolic risk association feature is an abnormal level of a certain amino acid, the knowledge graph can trace the metabolic reaction chain involved by that amino acid, determining its impact pathways on other metabolites and physiological processes.

[0148] In step S153, based on the core metabolic risk association characteristics and the corresponding metabolic influence pathways, the corresponding regulatory direction, intervention method and implementation process are retrieved from the set of metabolic regulation intervention strategies to form an initial intervention strategy. The set of metabolic regulation intervention strategies includes regulatory directions, intervention methods and implementation processes corresponding to different metabolic risk association characteristics.

[0149] The metabolic regulation intervention strategy set is a collection of pre-collected and organized regulatory directions, intervention methods, and implementation procedures tailored to different metabolic risk association characteristics. The initial intervention strategy is a preliminary intervention plan retrieved from the metabolic regulation intervention strategy set based on core metabolic risk association characteristics and metabolic impact pathways.

[0150] In this embodiment, the identified core metabolic risk association features and metabolic impact pathways are used as indexes to perform precise matching retrieval within a set of metabolic regulation intervention strategies. The set of metabolic regulation intervention strategies is typically stored in a database, with each record containing information such as metabolic risk association features, metabolic impact pathways, regulatory direction, intervention methods, and implementation procedures. Using SQL queries or specific retrieval algorithms, records that perfectly match or are most similar to the current core metabolic risk association features and metabolic impact pathways are found. The corresponding regulatory direction (e.g., increasing or decreasing the intake of a certain substance), intervention methods (e.g., adjusting feed formulation, adding nutritional supplements), and implementation procedures (e.g., intervention timing, frequency, etc.) are extracted, and this information is combined to form an initial intervention strategy.

[0151] In step S154, the core metabolic risk association features, the regulation direction, and the implementation process are integrated to generate the health intervention guidance information. Each intervention item in the health intervention guidance information corresponds to the metabolic influence path of the core metabolic risk association features.

[0152] This approach organically integrates the identified core metabolic risk-related features, retrieved regulatory directions, and implementation procedures. Using metabolic pathways as a framework, it details the regulatory directions and implementation procedures corresponding to each core metabolic risk-related feature, ensuring that each intervention closely corresponds to a metabolic pathway. For example, for the core feature affecting methionine metabolism, it explains which intervention method (such as adding a specific amino acid) is used to regulate it at which stage of the metabolic pathway, along with the specific implementation steps and timeline. Furthermore, the integrated information is organized using clear and easily understandable language and format to generate the final health intervention guidance information.

[0153] The aforementioned technical solution analyzes health risk prediction results, clearly identifying risk scenarios and correlation strengths to provide accurate direction for subsequent interventions; extracts core metabolic risk correlation characteristics and determines metabolic impact pathways, pinpointing key intervention points; retrieves initial intervention strategies to ensure the scientific validity and effectiveness of the intervention; and integrates information to generate guidance information. It can accurately generate health intervention guidance information targeting methionine metabolism regulation in target ruminants. This enables farmers to clearly understand the intervention content and implementation methods. It helps to timely and effectively intervene in metabolic regulation in ruminants, reducing health risks, improving animal production performance and health levels, and minimizing farming losses.

[0154] In a preferred embodiment, step S13, which involves inputting the target prediction data set into the dynamic association model, mining metabolic risk association features in the target prediction data set that are associated with the health status, and outputting a metabolic risk association feature set, includes:

[0155] In step S131, the target prediction data set is classified and divided according to sample type to obtain methionine metabolism sample subsets corresponding to different sample types;

[0156] The methionine metabolism sample subset is a specific set of sample data related to methionine metabolism obtained after classifying the target prediction data set according to sample type.

[0157] In this embodiment, a comprehensive analysis of the target prediction dataset is performed to identify feature fields that can be used to distinguish sample types, such as age data and variety identification. Based on these feature fields, a data classification algorithm, such as a rule-based classification method, is used to split the target prediction dataset into multiple subsets according to preset classification rules (e.g., defining the dividing lines between different age groups, coding for different varieties, etc.). Each subset corresponds to a sample type and contains only data related to methionine metabolism under that sample type, such as methionine intake and metabolite concentration, thereby forming a methionine metabolism sample subset.

[0158] In step S132, each of the methionine metabolism sample subsets is input into the dynamic association model to perform in-depth mining of the metabolic features in each of the methionine metabolism sample subsets to obtain the corresponding target metabolic features.

[0159] In this embodiment, each subset of methionine metabolism samples is sequentially input into a dynamic association model. The model first preprocesses the input data, including data standardization and normalization, to ensure a uniform scale. Then, it utilizes its complex internal neural network structure (such as a multilayer perceptron or convolutional neural network) to process the data layer by layer. In each layer, neurons perform nonlinear transformations on the input data through activation functions, extracting abstract features from the data. As the number of layers increases, the model gradually uncovers deeper, target metabolic features related to methionine metabolism within the data. These target metabolic features can more accurately reflect the metabolic state of ruminants.

[0160] In step S133, candidate metabolic features that are potentially associated with health status are extracted from the target metabolic features corresponding to each subset of methionine metabolism samples.

[0161] Among them, candidate metabolic features are those screened from target metabolic features that may have a potential association with the health status of ruminants.

[0162] In this embodiment, a correlation analysis method is employed, such as calculating the Pearson correlation coefficient or Spearman rank correlation coefficient between the target metabolic feature and health status indicators (e.g., disease incidence, degree of abnormality in physiological indicators). By setting a correlation threshold, when the absolute value of the correlation coefficient between the target metabolic feature and the health status indicator is greater than the threshold, the target metabolic feature is determined to be a candidate metabolic feature. Furthermore, feature selection algorithms from machine learning, such as those based on information gain or the Gini index, can be used to evaluate the importance of each target metabolic feature for health status classification or prediction, selecting features with higher importance as candidate metabolic features.

[0163] In step S134, the candidate metabolic features corresponding to all the methionine metabolism sample subsets are summarized to form a total set of candidate metabolic features;

[0164] In this embodiment, an empty dataset is established as the overall set of candidate metabolic features. The candidate metabolic features corresponding to each subset of methionine metabolism samples are sequentially traversed, and these features are added to the overall set of candidate metabolic features according to a specific data structure (such as a list or array). During the addition process, it is ensured that no duplicate features appear; if duplicates exist, they are merged or deduplicated. Simultaneously, the source dataset information for each feature is recorded, ultimately forming a comprehensive set containing all candidate metabolic features.

[0165] In step S135, the candidate metabolic feature set is sorted by correlation strength, and metabolic features with higher correlation strength are retained to generate the metabolic risk correlation feature set.

[0166] Among them, the association strength ranking is based on the degree of association between candidate metabolic features and health status, and they are ranked from high to low.

[0167] In this embodiment, the calculated correlation coefficient or the importance score obtained by the feature selection algorithm is used as the association strength index for each candidate metabolic feature. A ranking algorithm, such as quicksort or mergesort, is used to sort all features in the candidate metabolic feature set in descending order according to the association strength index. Based on a preset feature quantity threshold or association strength threshold, the features ranked higher are selected. These features are more closely associated with health status and have a greater impact on health risk, forming a metabolic risk-associated feature set.

[0168] The above technical solution classifies and divides sample subsets according to sample type, fully considering the differences between different samples, making the mining more targeted; the dynamic association model deeply mines target metabolic features, improving the accuracy and comprehensiveness of feature extraction; candidate metabolic features are screened through correlation analysis and feature selection, irrelevant features are removed, and data noise is reduced; a total set of candidate metabolic features is formed and ranked by association strength, finally generating a set of metabolic risk association features, which can efficiently and accurately mine metabolic risk association features closely related to the health status of ruminants from the target prediction dataset.

[0169] This invention provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of any of the methods described in the foregoing embodiments.

[0170] This invention provides a ruminant health risk prediction system based on methionine metabolism, comprising: a memory storing a computer program thereon; and a processor for executing the computer program in the memory to implement the steps of any of the methods described in the foregoing embodiments.

[0171] Figure 5 The illustrated methionine metabolism-based ruminant health risk prediction system 100 includes a processor 1001 and a memory 1003. The processor 1001 and memory 1003 are connected, for example, via a bus 1002. Optionally, the methionine metabolism-based ruminant health risk prediction system 100 may further include a communication component 1004, which can be used for data interaction between the system 100 and other devices, such as data transmission and / or data reception. It should be noted that in actual operation, the communication component 1004 is not limited to one, and the structure of this methionine metabolism-based ruminant health risk prediction system 100 does not constitute a limitation on the embodiments of this application.

[0172] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0173] Bus 1002 may include a pathway for transmitting information between the aforementioned components. Bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 1002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0174] The memory 1003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing program code and capable of being read by a computer, without limitation herein.

[0175] The memory 1003 is used to store program code for executing embodiments of the present disclosure, and its execution is controlled by the processor 1001. The processor 1001 is used to execute the program code stored in the memory 1003 to implement the steps shown in the foregoing embodiments of the ruminant health risk prediction method based on methionine metabolism.

[0176] The preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present disclosure, various changes, modifications, substitutions and variations can be made to these embodiments, and all such changes, modifications, substitutions and variations fall within the protection scope of the present disclosure.

[0177] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction, and such combinations should also be considered as part of this disclosure. To avoid unnecessary repetition, this disclosure will not further describe the various possible combinations. The technical scope of this application is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for predicting health risks in ruminants based on methionine metabolism, characterized in that, include: Data on methane concentration in exhaled gases, softness and expansion of ruminant food, methionine content, and lysine content of target ruminants over time were obtained to construct a target prediction dataset. The pre-trained metabolic association model is invoked, the target prediction dataset is input into the metabolic association model, the dynamic association between methionine metabolic characteristics and the health status of ruminants is constructed, and the dynamic association model representing the relationship between metabolic characteristics and health status is output. Input the target prediction dataset into the dynamic association model, mine the metabolic risk association features in the target prediction dataset that are associated with the health status, and output the metabolic risk association feature set. Load the pre-stored health risk pattern library, dynamically adapt the metabolic risk-related feature set with the risk patterns in the health risk pattern library, and output the health risk prediction results of ruminants. Based on the metabolic risk association feature set and the health risk prediction results of ruminants, health intervention guidance information for regulating methionine metabolism in target ruminants is generated. The health intervention guidance information is used to guide the health intervention operations for regulating methionine metabolism in the target ruminants.

2. The method for predicting health risks in ruminants based on methionine metabolism according to claim 1, characterized in that, The process involves calling a pre-trained metabolic association model, inputting the target prediction dataset into the model, constructing a dynamic correlation between methionine metabolic characteristics and the health status of ruminants, and outputting a dynamic association model characterizing the relationship between metabolic characteristics and health status, including: Methionine metabolic features are extracted from the target prediction dataset. These features include a first feature of methionine precursors, a second feature of methionine conversion intermediates, and a third feature of methionine metabolic end products. Obtain the health status information of ruminants corresponding to the target prediction data set, wherein the health status information of ruminants corresponds one-to-one with each sample data in the target prediction data set; The methionine metabolism characteristics are associated and integrated with the corresponding health status information of ruminants to form a training data set for the metabolic association model. The training dataset is input into the feature association layer of the pre-trained metabolic association model to perform hierarchical feature transformation on the methionine metabolism features and generate adaptive metabolic features that are suitable for health status association modeling. The dynamic association module of the metabolic association model is used to perform nonlinear association modeling on the adapted metabolic features and ruminant health status information, construct a dynamic association relationship between methionine metabolic features and ruminant health status, and output a dynamic association model representing the relationship between metabolic features and health status based on the dynamic association relationship.

3. The method for predicting health risks in ruminants based on methionine metabolism according to claim 2, characterized in that, The dynamic correlation module of the metabolic correlation model performs nonlinear correlation modeling on the adapted metabolic characteristics and ruminant health status information, constructing a dynamic correlation relationship between methionine metabolic characteristics and ruminant health status, including: The adapted metabolic features are input into the feature encoding submodule of the dynamic association module, and the adapted metabolic features are subjected to dimension-unified encoding processing to generate encoded metabolic features with unified dimension representation. The health status information of the ruminant is input into the status identification submodule of the dynamic association module, and the health status information of the ruminant is standardized and identified to generate a standardized health status identification. A feature cross-fusion method is used to mine the association information between the encoded metabolic features and the standardized health status identifier to generate preliminary association features; The preliminary association features are subjected to association signal enhancement processing to highlight the key association information between the encoded metabolic features and the standardized health status identifier, thereby generating enhanced association features. Based on the enhanced correlation features, a nonlinear correlation link between methionine metabolic features and the health status of ruminants is constructed, forming a dynamic correlation between methionine metabolic features and the health status of ruminants.

4. The method for predicting health risks in ruminants based on methionine metabolism according to claim 2, characterized in that, The step of outputting a dynamic correlation model characterizing the relationship between metabolic features and health status based on the dynamic correlation includes: Extract the model parameters corresponding to the dynamic relationships to form a model parameter set; Repeat the following steps: input the validation sample data from the training dataset into the model structure corresponding to the dynamic association relationship, and generate the validation sample prediction results; Based on the health status information corresponding to the validation sample prediction results and validation sample data, model prediction bias information is generated. The model prediction deviation information is input into the parameter adjustment unit of the parameter optimization module to determine the target adjustment parameter and the corresponding adjustment direction in the model parameter set. The target adjustment parameters are adjusted using a gradient according to the adjustment direction to generate the adjusted model parameters. The adjusted model parameters are then updated into the model structure corresponding to the dynamic association relationship to form the updated dynamic association model. The test sample data from the training dataset is input into the updated dynamic association model to generate test sample prediction results. The model parameters are adjusted based on the test sample prediction results until the model prediction deviation information reaches the preset standard, and then the dynamic association model is output.

5. The method for predicting health risks in ruminants based on methionine metabolism according to claim 1, characterized in that, The pre-stored health risk pattern library is loaded, and the metabolic risk-related feature set is dynamically adapted to the risk patterns in the health risk pattern library to output ruminant health risk prediction results, including: Load the pre-stored health risk pattern library and calculate the correlation fit between each metabolic risk association feature in the metabolic risk association feature set and each risk pattern in the health risk pattern library; Based on the correlation fit, target risk patterns that best fit each of the metabolic risk correlation features are selected. Extract the health risk information corresponding to the target risk pattern, and generate ruminant health risk prediction results by combining the correlation fit. The ruminant health risk prediction results include a health risk scenario description and the correlation fit.

6. The method for predicting health risks in ruminants based on methionine metabolism according to claim 5, characterized in that, The calculation of the correlation fit between each metabolic risk association feature in the metabolic risk association feature set and each risk pattern in the health risk pattern library includes: Each metabolic risk association feature in the metabolic risk association feature set is aligned with each risk pattern in the health risk pattern library to generate an aligned feature-pattern feature pair. The feature interaction unit performs feature interaction operations on feature and pattern feature pairs to generate an interaction feature matrix. The interaction feature matrix is ​​input into the correlation calculation unit to perform element-level correlation analysis on the interaction feature matrix and extract the key elements in the interaction feature matrix that reflect the degree of correlation. Based on the distribution and feature strength of the key elements, a correlation metric is generated between each metabolic risk association feature and each risk pattern in the health risk pattern library. Based on the correlation metric between each metabolic risk association feature and each risk pattern in the health risk pattern library, the correlation fit between each metabolic risk association feature and each risk pattern in the health risk pattern library is generated.

7. The method for predicting health risks in ruminants based on methionine metabolism according to claim 5, characterized in that, The step of extracting health risk information corresponding to the target risk pattern and generating ruminant health risk prediction results by combining the correlation fit includes: The structural information of the target risk pattern is analyzed to generate the core elements in the target risk pattern used to describe health risk scenarios. The core elements are organized according to logical relationships to form a standardized description of health risk scenarios; The correlation fit between the metabolic risk association features and the target risk pattern is converted into a risk confidence level statement; The health risk scenario description and the risk confidence statement are linked and integrated, and sample identification information and prediction time information are added; The integrated information is structured to generate the ruminant health risk prediction result, which includes the health risk scenario description, the risk confidence level statement, the sample identification information, and the prediction time information.

8. The method for predicting health risks in ruminants based on methionine metabolism according to any one of claims 1-7, characterized in that, Based on the metabolic risk association feature set and ruminant health risk prediction results, health intervention guidance information for regulating methionine metabolism in target ruminants is generated, including: Analyze the health risk scenario descriptions and correlation fit information in the ruminant health risk prediction results to determine the current health risk scenarios and risk correlation strength of ruminants. Extract core metabolic risk association features corresponding to the health risk scenario from the set of metabolic risk association features, and determine the metabolic impact path of the core metabolic risk association features; Based on the core metabolic risk association characteristics and the corresponding metabolic impact pathways, the corresponding regulatory directions, intervention methods and implementation procedures are retrieved from the set of metabolic regulation intervention strategies to form an initial intervention strategy. The set of metabolic regulation intervention strategies includes regulatory directions, intervention methods and implementation procedures corresponding to different metabolic risk association characteristics. The core metabolic risk association features, the regulation direction, and the implementation process are integrated to generate the health intervention guidance information. Each intervention item in the health intervention guidance information corresponds to the metabolic impact pathway of the core metabolic risk association features.

9. The method for predicting health risks in ruminants based on methionine metabolism according to any one of claims 1-7, characterized in that, The step involves inputting the target prediction dataset into the dynamic association model, mining metabolic risk association features in the target prediction dataset that are associated with the health status, and outputting a set of metabolic risk association features, including: The target prediction dataset is classified and divided according to sample type to obtain methionine metabolism sample subsets corresponding to different sample types; Each of the methionine metabolism sample subsets is input into the dynamic association model to perform in-depth mining of the metabolic features in each of the methionine metabolism sample subsets, thereby obtaining the corresponding target metabolic features. From the target metabolic features corresponding to each subset of methionine metabolism samples, extract candidate metabolic features that are potentially associated with health status; The candidate metabolic features corresponding to all the methionine metabolism sample subsets are summarized to form a total set of candidate metabolic features; The candidate metabolic features are sorted by association strength, and the metabolic features with the highest association strength are retained to generate the metabolic risk association feature set.

10. A ruminant health risk prediction system based on methionine metabolism, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-9.

Images