Method and system for modeling the association between pasture environment parameters and beef cattle health

Abstract

The application discloses a pasture environment parameter and beef cattle health correlation modeling method and system, and particularly relates to the technical field of livestock feeding management decision, and is used for solving the problems of low health risk prediction accuracy and lagging management decision caused by the insufficient dynamic time sequence relationship description between environment parameters and physiological state, and the insufficient consideration of nonlinear interaction effects of multiple environmental factors; the delay effect mode of the influence of environment parameters on physiological indexes is identified through time sequence analysis, the behavior mode migration characteristics are extracted based on activity data, the nonlinear interaction effects and synergistic influence characteristics among environment parameters are quantified in combination with a response surface model, and a dynamic prediction model is constructed by fusing the above multidimensional characteristics, and finally, the health risk prediction results output by the model are used to generate pasture environment regulation and feeding management suggestions to realize accurate prediction and active intervention management of the health state of beef cattle.

Description

Technical Field

[0001] This invention relates to the field of livestock feeding management decision-making technology, and more specifically, to a method and system for modeling the correlation between pasture environmental parameters and beef cattle health. Background Technology

[0002] In the field of intensive livestock farming, accurate prediction and proactive intervention of disease risks in beef cattle, especially heat stress, are crucial for improving animal welfare and economic efficiency. Currently, real-time collection of pasture environmental parameters (such as temperature, humidity, ammonia concentration, and dust levels) and beef cattle physiological data (such as body temperature and activity levels) based on Internet of Things (IoT) technology, and the construction of early warning mechanisms accordingly, has become an important direction for technological development in this field. Existing typical technologies usually rely on setting fixed early warning thresholds for environmental or physiological indicators, or using statistical analysis to establish static correlation models between environmental and health data, aiming to achieve the identification and management of health status.

[0003] However, the impact of environmental parameters on the physiological state of beef cattle is a complex process that is dynamic, continuous, and exhibits significant time-delay effects. Furthermore, there are synergistic or antagonistic coupling effects among various environmental factors. Existing static or simple correlation models struggle to accurately characterize this dynamic, nonlinear temporal causal relationship, leading to discrepancies between disease risk predictions and the animals' actual stress responses. This results in insufficient accuracy and timeliness of early warning management measures, failing to provide a reliable basis for optimizing farm production management decisions. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for modeling the correlation between pasture environmental parameters and beef cattle health to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for modeling the correlation between pasture environmental parameters and beef cattle health includes the following steps:

[0007] S1. Obtain environmental parameter data of the pasture and physiological data of beef cattle within the target time period;

[0008] S2. Use time series analysis algorithms to process environmental parameter data and physiological data, and identify the delayed effect pattern of the influence of environmental parameters on physiological data;

[0009] S3. Identify the behavioral pattern migration characteristics of individual beef cattle based on activity data in physiological data;

[0010] S4. Based on environmental parameter data and physiological data, construct a response surface model of health status with respect to multiple environmental parameters. Quantify the nonlinear interaction effect between environmental parameters by analyzing the curvature characteristics of the response surface model, and determine the synergistic influence characteristics of the combination of environmental parameters on the health status of beef cattle by combining behavioral pattern transfer characteristics.

[0011] S5. Based on the delayed effect model, synergistic influence characteristics, and behavioral pattern transfer characteristics, a dynamic prediction model between environmental parameters and the health status of beef cattle is constructed.

[0012] S6. Utilize the dynamic prediction model to output the health risk prediction results of beef cattle and generate farm management decision-making suggestions.

[0013] Furthermore, environmental parameter data of the pasture and physiological data of beef cattle within the target time period are obtained, including:

[0014] Environmental parameter data is collected by IoT sensors deployed in the ranch environment. The environmental parameter data includes at least temperature and humidity, ammonia concentration and dust amount.

[0015] Physiological data are collected by wearing monitoring devices on individual beef cattle, including at least body temperature and activity level.

[0016] Furthermore, time-series analysis algorithms are used to process environmental parameter data and physiological data to identify delayed effect patterns of environmental parameters on physiological data, including:

[0017] Calculate the cross-correlation function between each environmental parameter data and each physiological index in the physiological data at multiple time offsets;

[0018] Find the peak point where the absolute value of the cross-correlation function is the largest;

[0019] The time offset corresponding to the peak point is determined as the delay time of the influence of the corresponding environmental parameter on the corresponding physiological index, thus forming a delay effect model.

[0020] Furthermore, the cross-correlation function of each environmental parameter data and each physiological indicator in the physiological data at multiple time offsets is calculated in the following way: After standardizing and preprocessing the time series of each environmental parameter and the time series of physiological indicators, the correlation coefficient of the two series at different lag time points is calculated within a preset time offset range; the time offset is gradually adjusted by a sliding window method, and the correlation coefficient value at each offset is recorded to form a cross-correlation function sequence; finally, the cross-correlation function reflecting the degree of correlation between environmental parameters and physiological indicators at different time delays is obtained.

[0021] Furthermore, based on activity data from physiological data, behavioral pattern transfer characteristics of individual beef cattle are identified, including:

[0022] Calculate the change in the statistical distribution characteristics of the current time period's activity level relative to the historical baseline activity level;

[0023] Identify anomalous changes in the behavioral state transition probability matrix based on time-series patterns in activity data;

[0024] By combining the changes in statistical distribution characteristics with the abnormal changes in the behavioral state transition probability matrix, behavioral pattern transfer characteristics are generated.

[0025] Furthermore, based on environmental parameter data and physiological data, a response surface model of health status with respect to multiple environmental parameters is constructed. The nonlinear interaction effects between environmental parameters are quantified by analyzing the curvature characteristics of the response surface model. Combined with behavioral pattern transfer characteristics, the synergistic influence characteristics of environmental parameter combinations on the health status of beef cattle are determined, including:

[0026] A response surface model was established by fitting the functional relationship between environmental parameter data and health status indicator parameters in physiological data using the multinomial regression method.

[0027] Calculate the Hessian matrix of the response surface model at each point in the environmental parameter value space, and determine the intensity of the nonlinear interaction effect between environmental parameters by analyzing the eigenvalue distribution of the Hessian matrix.

[0028] By weighted fusion of the intensity of nonlinear interaction effects and behavioral pattern transfer characteristics, the synergistic influence characteristics of environmental parameter combinations on the health status of beef cattle are generated.

[0029] Furthermore, the Hessian matrix of the response surface model at each point in the environmental parameter value space is calculated as follows: the second-order partial derivatives of the constructed response surface model are obtained to form the Hessian matrix; representative sampling points are selected within the range of environmental parameter values, and the Hessian matrix at each sampling point is calculated; the strength and direction of the interaction between environmental parameters are determined by analyzing the sign and magnitude of the eigenvalues ​​of the Hessian matrix; and the eigenvalue distribution characteristics are quantified into a nonlinear interaction effect intensity index.

[0030] Furthermore, based on delayed effect patterns, synergistic influence characteristics, and behavioral pattern transfer characteristics, a dynamic prediction model between environmental parameters and the health status of beef cattle is constructed, including:

[0031] The delayed effect pattern is used as the time series feature input, the synergistic influence feature is used as the interaction effect input, and the behavioral pattern transfer feature is used as the behavioral state input, and all of them are input into the time series machine learning model.

[0032] A multi-feature fusion training strategy was adopted to train the time series machine learning model, enabling the model to simultaneously learn the temporal delay patterns of environmental parameters, the interaction effects between parameters, and the migration patterns of animal behavior.

[0033] The output is a dynamic prediction model that can predict the health status of beef cattle based on environmental parameter data.

[0034] Furthermore, the dynamic prediction model is used to output the health risk prediction results of beef cattle, generating farm management decision-making recommendations, including:

[0035] The environmental parameter data collected in real time is input into the dynamic prediction model to obtain the predicted value of the health status of beef cattle;

[0036] The predicted health status of beef cattle is compared with a preset risk threshold to generate different levels of health risk prediction results.

[0037] Based on the level of health risk prediction results, a pre-set ranch management decision-making scheme library is matched, and corresponding environmental control measures and feeding management suggestions are output.

[0038] On the other hand, the present invention provides a modeling system for the correlation between pasture environmental parameters and beef cattle health, including the following modules:

[0039] The data acquisition module is used to acquire environmental parameter data of the pasture and physiological data of beef cattle within the target time period;

[0040] The time series analysis module is used to process environmental parameter data and physiological data using time series analysis algorithms to identify the delayed effect patterns of environmental parameters on physiological data.

[0041] The behavior recognition module is used to identify the behavioral pattern migration characteristics of individual beef cattle based on activity data in physiological data.

[0042] The surface analysis module is used to construct a response surface model of health status with respect to multiple environmental parameters based on environmental parameter data and physiological data. By analyzing the curvature characteristics of the response surface model, the nonlinear interaction effect between environmental parameters is quantified, and the synergistic influence characteristics of the combination of environmental parameters on the health status of beef cattle are determined by combining behavioral pattern transfer characteristics.

[0043] The model building module is used to construct a dynamic prediction model between environmental parameters and the health status of beef cattle based on delayed effect patterns, synergistic influence characteristics, and behavioral pattern transfer characteristics.

[0044] The decision output module is used to output the health risk prediction results of beef cattle using a dynamic prediction model, and generate farm management decision recommendations.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. By using time-series analysis algorithms to process environmental parameter data and physiological data, the delayed effect patterns of environmental parameters on physiological data can be accurately identified, effectively capturing the dynamic temporal relationship between environmental changes and cattle physiological responses. Identifying behavioral pattern migration characteristics of individual beef cattle based on activity data within physiological data allows for the keen detection of subtle changes in cattle behavior patterns, providing crucial information for health status assessment. Constructing a response surface model of health status with respect to multiple environmental parameters and analyzing its curvature characteristics allows for the quantification of nonlinear interaction effects between environmental parameters. Combining this with behavioral pattern migration characteristics determines the synergistic impact of combined environmental parameter effects on beef cattle health status, thus comprehensively understanding the complex mechanisms of the combined effects of multiple environmental factors.

[0047] 2. A dynamic prediction model based on delayed effect patterns, synergistic influence characteristics, and behavioral pattern migration characteristics can comprehensively consider multi-dimensional information such as the temporal delay characteristics of environmental factors, multi-parameter interaction effects, and individual behavioral changes, achieving accurate prediction of beef cattle health status. Using the health risk prediction results output by this dynamic prediction model, targeted farm management decision-making suggestions can be generated, helping farmers take timely preventative measures, significantly improving the accuracy and proactivity of farm health management, effectively reducing the risk of disease occurrence, and increasing farming efficiency. Attached Figure Description

[0048] Figure 1 This is a flowchart of the method for modeling the correlation between pasture environmental parameters and beef cattle health according to the present invention;

[0049] Figure 2 This is a schematic diagram of the structure of the pasture environmental parameters and beef cattle health correlation modeling system of the present invention. Detailed Implementation

[0050] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0051] Example 1: Figure 1 The present invention provides a method for modeling the correlation between pasture environmental parameters and beef cattle health, which includes the following steps:

[0052] S1. Obtain environmental parameter data of the pasture and physiological data of beef cattle within the target time period;

[0053] S2. Use time series analysis algorithms to process environmental parameter data and physiological data, and identify the delayed effect pattern of the influence of environmental parameters on physiological data;

[0054] S3. Identify the behavioral pattern migration characteristics of individual beef cattle based on activity data in physiological data;

[0055] S4. Based on environmental parameter data and physiological data, construct a response surface model of health status with respect to multiple environmental parameters. Quantify the nonlinear interaction effect between environmental parameters by analyzing the curvature characteristics of the response surface model, and determine the synergistic influence characteristics of the combination of environmental parameters on the health status of beef cattle by combining behavioral pattern transfer characteristics.

[0056] S5. Based on the delayed effect model, synergistic influence characteristics, and behavioral pattern transfer characteristics, a dynamic prediction model between environmental parameters and the health status of beef cattle is constructed.

[0057] S6. Utilize the dynamic prediction model to output the health risk prediction results of beef cattle and generate farm management decision-making suggestions.

[0058] S1. Obtain environmental parameter data of the pasture and physiological data of beef cattle within the target time period. The specific implementation is as follows:

[0059] When acquiring environmental parameter data of the pasture and physiological data of beef cattle within the target time period, environmental parameter data is first collected through IoT sensors deployed in the pasture environment. Specifically, multi-parameter environmental acquisition nodes are distributed and installed in the pasture and surrounding areas at a density of at least one monitoring point per 100 square meters. Each node integrates a temperature and humidity composite sensor, an electrochemical ammonia concentration sensor, and a laser scattering dust sensor. The temperature and humidity sensor measures temperatures from -10°C to 50°C and relative humidity from 0 to 100%, with a sampling frequency set to collect data every 5 minutes. The ammonia concentration sensor has a range of 0 to 100 ppm and a resolution of 0.1 ppm, employing a cyclic sampling mode every 10 minutes. The dust sensor monitors inhalable particulate matter with a particle size range of 0.3 micrometers to 10 micrometers, a concentration range of 0 to 1000 micrograms per cubic meter, and a sampling interval set at 15 minutes. All sensor nodes transmit the collected environmental parameter data to the central data processing server in real time via the LoRa wireless communication protocol. The data format adopts JSON structured data packets, which include fields such as sensor number, collection timestamp, temperature value, humidity value, ammonia concentration value, and dust concentration value.

[0060] When collecting physiological data through wearable monitoring devices on individual beef cattle, each participating cattle is equipped with a specially designed biomonitoring collar. This device incorporates a high-precision digital body temperature sensor and a triaxial accelerometer. The body temperature sensor probe is encapsulated in medical-grade stainless steel, fitting snugly against the skin surface of the cattle's neck, with a measurement accuracy of ±0.1 degrees Celsius and a sampling frequency of once per minute. The triaxial accelerometer has a range of ±16g and a sampling frequency of 10Hz, enabling continuous recording of the animal's activity acceleration data. The monitoring device's built-in microprocessor processes the raw acceleration data in real time, calculating the activity intensity index by three-dimensional vector modulus, and simultaneously employing a sliding window algorithm to identify activity behavior patterns. All physiological data is transmitted via Bluetooth 5.0 to the farm's base station gateway, and ultimately aggregated in the central data processing system. The data packet contains key information such as individual number, timestamp, body temperature measurement, and activity level indicators.

[0061] A rigorous quality control mechanism was established during data acquisition. Environmental sensors underwent weekly on-site calibration using a standard temperature and humidity calibration chamber, ammonia standard gas, and a dust concentration calibrator. Physiological monitoring equipment underwent accuracy verification every two weeks using a medical body temperature calibrator and a standard acceleration testing platform. All collected raw data underwent outlier filtering using an outlier removal algorithm based on the statistical 3σ principle. Time series interpolation was also used to supplement missing data, ensuring data integrity and reliability. Data storage employed a time-series database structure, with each data point bearing a precise timestamp and source identifier, providing a high-quality data foundation for subsequent analysis.

[0062] S2. Use time-series analysis algorithms to process environmental parameter data and physiological data, and identify the delayed effect patterns of environmental parameters on physiological data. Specifically, this is implemented as follows:

[0063] When using time series analysis algorithms to process environmental and physiological data, the first step is to standardize the collected time series data of environmental parameters and physiological indicators. The preprocessing includes two main steps: outlier removal and data normalization. Outlier removal employs a three-standard-deviation criterion based on a moving window. This involves calculating the mean and standard deviation of each time series within a 24-hour moving window, marking data points outside the range of the mean plus or minus three standard deviations as outliers, and replacing them using linear interpolation. Data normalization uses the Z-score standardization method, converting each time series into a standard normal distribution with a mean of zero and a standard deviation of one. Specifically, this involves subtracting the series mean from each data point and then dividing by the series standard deviation. The preprocessed time series eliminates the influence of dimensional differences and abnormal fluctuations, providing a high-quality data foundation for subsequent correlation analysis.

[0064] When calculating the cross-correlation function of each environmental parameter data and each physiological indicator in the physiological data at multiple time offsets, the upper and lower limits of the time offset range are first determined. Based on the physiological characteristics of beef cattle, the maximum time offset is set to ±24 hours to cover all possible delayed effects of environmental factors on physiological indicators. During the calculation, a sliding window method is used to progressively adjust the time offset, with the window size set to 1 hour, meaning the time offset increases or decreases by 1 hour each time. For each specific time offset, the environmental parameter time series and the physiological indicator time series are aligned according to that offset, and then the Pearson correlation coefficient between the two series is calculated. The correlation coefficient is calculated by dividing the covariance of the two series by the product of their respective standard deviations. The correlation coefficient values ​​calculated at each time offset are recorded, forming a complete cross-correlation function sequence. This sequence reflects the strength of the correlation between environmental parameters and physiological indicators at different time delays.

[0065] When searching for the peak point with the largest absolute value of the cross-correlation function, the calculated cross-correlation function sequence is first smoothed using a moving average filter with a window size of 3 to reduce the impact of random fluctuations. Then, the entire cross-correlation function sequence is traversed to find all local extrema, i.e., those points where the correlation coefficient value is greater than the values ​​of their immediate neighbors. Among these local extrema, the point with the largest absolute value is selected as a candidate peak point. A threshold for the absolute value of the correlation coefficient is set; only when the absolute value of the correlation coefficient of a candidate peak point is greater than 0.3 is it considered a valid peak point. This threshold is an empirical value derived from statistical analysis of a large amount of experimental data and can effectively filter out random correlations. If multiple peak points meet the criteria, the peak point with the smallest time offset is selected, because in reality, the influence of the environment on physiology usually has the shortest delay characteristic.

[0066] When determining the time offset corresponding to the peak point as the delay time of the influence of the corresponding environmental parameter on the corresponding physiological indicator, multiple factors need to be considered to ensure the reliability of the results. For each combination of environmental parameter and physiological indicator, the time offset corresponding to the peak point of its cross-correlation function is recorded. This offset indicates how long it takes for the environmental parameter to have its maximum impact on the physiological indicator after a change. The correlation coefficient value of this peak point is also recorded as an indicator of the influence intensity; the larger the absolute value of the correlation coefficient, the more significant the environmental influence. The delay time data of all combinations of environmental parameters and physiological indicators are organized into a delay effect pattern matrix. The rows of this matrix represent different types of environmental parameters, and the columns represent different physiological indicators. The matrix elements contain two data points: delay time value and influence intensity value. Finally, this delay effect pattern matrix is ​​validated. A cross-validation method is used to divide the dataset into a training set and a test set. The delay time patterns calculated on the training set need to be validated on the test set. The peak points of the cross-correlation function on the test set should appear at the same or similar time offset positions to ensure the reliability and stability of the delay effect pattern. The resulting delayed effect pattern will be used as an important feature input into the subsequent dynamic prediction model, providing temporal data support for accurately predicting the health status of beef cattle.

[0067] S3. Based on activity data from physiological data, identify the behavioral pattern transfer characteristics of individual beef cattle, specifically implemented as follows:

[0068] When identifying behavioral pattern migration characteristics in individual beef cattle based on activity data from physiological data, it is first necessary to establish historical baseline activity levels. These historical baseline activity levels are established by collecting historical activity data of the individual beef cattle in a healthy state, specifically selecting activity data from the same time period each day over the most recent 30 days as the baseline dataset. This historical activity data is then segmented by hour, and statistical characteristics such as the mean, standard deviation, maximum, minimum, and interquartile range are calculated for each hourly segment, forming a 24-hour periodic baseline statistical characteristic curve. The baseline statistical characteristic curve is established using a moving average method, calculating a weighted average of data from the same time period over several consecutive days. More recent data is assigned a higher weight, and more distant data a lower weight. The weighting coefficients are allocated according to a time decay function to ensure that the baseline value reflects the most recent normal behavioral patterns.

[0069] When calculating the change in the statistical distribution characteristics of activity data in the current time period relative to the historical baseline activity level, the current time period is first divided into the same time periods as the historical baseline, such as by hour. The same statistical characteristic indicators are calculated for the activity data within each time period, including distribution characteristic parameters such as mean, standard deviation, skewness, and kurtosis. These current statistical characteristic values ​​are compared with the corresponding statistical characteristic values ​​of the historical baseline time period to calculate the relative change. For example, the rate of change of the mean is calculated by subtracting the historical baseline mean from the current mean and then dividing by the historical baseline mean; the rate of change of the standard deviation is calculated using the same method. Simultaneously, changes in distribution pattern indicators are calculated, including the difference in skewness coefficient and the difference in kurtosis coefficient, which reflect changes in the distribution pattern of activity. Finally, a multi-dimensional vector of statistical distribution characteristic changes is obtained, containing changes in multiple dimensions such as the rate of change of the mean, the rate of change of the standard deviation, the difference in skewness, and the difference in kurtosis.

[0070] When identifying abnormal changes in the behavioral state transition probability matrix based on time-series activity data, it is first necessary to define the behavioral state categories of beef cattle. Based on statistical analysis of activity data, behavioral states are divided into four categories: resting, lightly active, moderately active, and vigorously active. The classification threshold is determined based on the statistical distribution of individual historical data; for example, activity levels can be divided into four intervals corresponding to the four states according to the quartiles. Then, a state transition probability matrix is ​​calculated based on the time-series data. The rows of this matrix represent the state at the previous moment, the columns represent the state at the current moment, and the matrix elements represent the probability of transitioning from one state to another. The state transition probability matrix for the current time period is calculated and compared with the state transition probability matrix for the historical baseline period, calculating the change in each transition probability. Emphasis is placed on monitoring changes in transition probabilities that are relatively stable in historical data; for example, the probability of transitioning from a resting state to a lightly active state is usually relatively stable, and abnormal changes in these probabilities may indicate abnormal behavioral patterns.

[0071] When generating behavioral pattern transfer features by combining changes in statistical distribution characteristics with anomalous changes in the behavioral state transition probability matrix, a feature fusion method is employed to organically combine the two types of features. First, the vector of changes in statistical distribution characteristics and the matrix of changes in behavioral state transition probability are standardized to eliminate dimensional differences. The standardization method uses min-max normalization to map each feature value to a range of 0 to 1. Then, a weight coefficient is assigned to each feature. The weight coefficient is determined based on the feature's importance assessment, which is obtained through historical data analysis. Features that show significant changes in historical anomalous events are given higher weights. For example, changes in the mean change rate and the probability of maintaining a resting state are usually highly important and therefore given larger weights. Finally, the standardized features are fused into a comprehensive behavioral pattern transfer feature index through weighted summation. The larger the index value, the greater the deviation of the behavioral pattern from the normal baseline. Simultaneously, each original feature component is retained, forming a multi-dimensional behavioral pattern transfer feature vector, providing detailed behavioral feature input for subsequent health status prediction. This behavioral pattern transfer feature can comprehensively reflect the short-term and long-term trend changes in beef cattle behavior patterns, providing important behavioral evidence for health status assessment.

[0072] S4. Based on environmental parameter data and physiological data, a response surface model of health status with respect to multiple environmental parameters is constructed. The nonlinear interaction effects between environmental parameters are quantified by analyzing the curvature characteristics of the response surface model. Furthermore, the synergistic influence characteristics of environmental parameter combinations on the health status of beef cattle are determined by combining behavioral pattern transfer characteristics. The specific implementation is as follows:

[0073] When constructing a response surface model of health status with respect to multiple environmental parameters based on environmental and physiological data, it is first necessary to determine the health status indicator parameters. These parameters are indicators extracted from physiological data that can comprehensively reflect the health status of beef cattle. Typically, the duration of abnormal body temperature and the magnitude of abnormal activity are selected as the basic data for the health status indicator parameters. A comprehensive health score is formed by weighted combination of these basic data. The weight coefficients are determined based on the sensitivity and specificity of each physiological indicator to the health status. Sensitivity is assessed by the magnitude of change of the indicator before the occurrence of abnormal health events in historical data, and specificity is assessed by the degree to which the indicator remains stable in a healthy state. When using polynomial regression to fit the functional relationship between environmental parameter data and health status indicator parameters, a quadratic polynomial is chosen as the basic form of the response surface model because it can better characterize the nonlinear relationship and interaction between variables. Specifically, the health status indicator parameter is equal to a linear combination of the constant term, linear term, quadratic term, and interaction term of the environmental parameter. The coefficients of each term are determined by least squares fitting. During the fitting process, a regularization method is used to prevent overfitting. The regularization coefficient is determined by cross-validation. The dataset is divided into a training set and a validation set. Models with different regularization coefficients are fitted on the training set, and the model prediction error is evaluated on the validation set. The regularization coefficient that minimizes the validation set error is selected.

[0074] When calculating the Hessian matrix of the response surface model at each point in the environmental parameter value space, it is first necessary to determine the range of environmental parameter values. This range is determined based on the minimum and maximum values ​​of each environmental parameter in historical monitoring data, and appropriately extended by a certain proportion to cover possible extreme cases. Representative sampling points are selected within the environmental parameter value space using the Latin hypercube sampling method. The number of sampling points is determined according to the dimensions of the environmental parameters, typically 20 to 50 sampling points per dimension, ensuring that the sampling points cover the entire parameter space and are evenly distributed. The Hessian matrix of the response surface model is calculated at each sampling point. The Hessian matrix is ​​a symmetric matrix whose elements are the second-order partial derivatives of the response surface model with respect to each environmental parameter. For quadratic polynomial models, these second-order partial derivatives are constants, therefore the Hessian matrix is ​​the same at all sampling points; for higher-order models, the second-order partial derivatives need to be calculated separately for each sampling point. The strength and direction of interactions between environmental parameters are determined by analyzing the sign and magnitude of the eigenvalues ​​of the Hessian matrix. The sign of the eigenvalue indicates the nature of the interaction: positive eigenvalues ​​indicate synergistic effects, and negative eigenvalues ​​indicate antagonistic effects. The absolute value of the eigenvalue indicates the strength of the interaction; the larger the absolute value, the stronger the interaction. When quantifying the eigenvalue distribution characteristics into an index of nonlinear interaction effect strength, the standard deviation of the eigenvalues ​​is used as the quantification index. The larger the standard deviation, the stronger the nonlinear interaction effect between environmental parameters. The ratio of positive to negative eigenvalues, as well as the sign and magnitude of the largest eigenvalue, are also recorded. This information together constitutes the index of nonlinear interaction effect strength.

[0075] When generating the synergistic impact characteristics of environmental parameter combinations on the health status of beef cattle by weighted fusion of the nonlinear interaction effect strength and behavioral pattern transfer characteristics, the two types of features are first standardized to have the same dimensions. The standardization method adopts Z-score standardization, that is, each feature value is subtracted from the mean of the feature over all samples and then divided by the standard deviation. The weights are determined based on the importance of the features in predicting health status. Importance is assessed by the correlation between the features and abnormal health status events in historical data. Features with stronger correlations are assigned higher weights. The specific weight coefficients are determined by combining expert experience and statistical data analysis. For example, experts in the field of animal husbandry can be invited to score the importance of each feature based on their professional knowledge. At the same time, the average change of each feature before the occurrence of abnormal health events in historical data can be analyzed. The expert scores and statistical data analysis results are combined and normalized to obtain the final weight coefficients. The weighted fusion adopts a linear weighted summation method, that is, the synergistic impact feature is equal to the weighted sum of the nonlinear interaction effect strength index and the behavioral pattern transfer feature, and the sum of the weight coefficients is 1. The final generated synergistic impact feature is a comprehensive index whose value reflects the combined influence of environmental parameter combinations on the health status of beef cattle through two pathways: nonlinear interaction effects and behavioral pattern changes. This feature will serve as an important input parameter for constructing a dynamic prediction model between environmental parameters and beef cattle health status. Throughout the process, it is necessary to ensure that the input and output data formats are consistent across all steps, and that feature naming is standardized to facilitate subsequent model construction and use. Intermediate calculation results should also be retained for model validation and result interpretation.

[0076] S5. Based on the delayed effect model, synergistic influence characteristics, and behavioral pattern transfer characteristics, a dynamic prediction model between environmental parameters and the health status of beef cattle is constructed. The specific implementation is as follows:

[0077] When constructing a dynamic prediction model between environmental parameters and the health status of beef cattle based on delayed effect patterns, synergistic influence features, and behavioral pattern transfer features, the input features first need to be preprocessed and formatted. When delayed effect patterns are used as time-series feature inputs, they need to be converted to a format suitable for time-series model inputs. Specifically, the delay time information of each combination of environmental parameter and physiological indicator is encoded into a time-delay feature vector. Each dimension of the vector corresponds to a specific time delay, and the dimension value represents the correlation strength between the environmental parameter and the physiological indicator at that delay time. When synergistic influence features are used as interaction effect inputs, they need to be organized into a multi-dimensional feature vector, where each dimension represents the synergistic influence strength of different combinations of environmental parameters. The dimension values ​​are filled using previously calculated synergistic influence feature indicators. When behavioral pattern transfer features are used as behavioral state inputs, the previously calculated multi-dimensional feature vector is directly used, containing feature values ​​from multiple dimensions such as changes in statistical distribution characteristics and changes in behavioral state transition probabilities.

[0078] When inputting three types of features into a time series machine learning model, a Long Short-Term Memory (LSTM) neural network is chosen as the basic model architecture. This model can effectively process time series data and capture long-term dependencies. The model input layer is designed with three independent feature input branches, corresponding to time series features, interaction effect features, and behavioral state features, respectively. Each branch contains a fully connected layer for feature preprocessing and dimensionality adjustment. The outputs of the three branches are fused in the middle layer of the model, concatenating feature vectors from different sources into a comprehensive feature vector, which is then input into the LTM neural network layer for time series modeling. The number of hidden units in the LTM neural network layer is determined based on the dimension of the input features, typically set to 2 to 3 times the dimension of the input features. The number of network layers is set to 2 to 3 to balance model complexity and expressive power. The model output layer uses a combination of fully connected layers and activation functions to ultimately output a predicted value of the beef cattle's health status. This predicted value is a continuous numerical value ranging from 0 to 1, representing the degree of health status.

[0079] When training a time-series machine learning model using a multi-feature fusion training strategy, the first step is to prepare a training dataset. This dataset includes historical environmental parameter data, corresponding feature data, and actual observed health status labels for beef cattle. The training process employs time-series cross-validation, dividing the dataset into multiple training and validation folds in chronological order to ensure the model's generalization ability on time-series data. Mean squared error loss is chosen as the loss function to measure the difference between the model's predicted values ​​and the actual health status. The optimization algorithm is adaptive moment estimation, with an initial learning rate of 0.001 and a learning rate decay strategy, dynamically adjusting the learning rate based on performance on the validation set during training. An early stopping strategy is used to prevent overfitting; training stops when the validation set loss no longer decreases for several consecutive training epochs. Model hyperparameters are determined using a grid search method, with the search scope including parameters such as the number of network layers, the number of hidden units, and the learning rate. The optimal hyperparameter combination performing best on the validation set is ultimately selected.

[0080] When outputting a dynamic prediction model capable of forecasting beef cattle health status based on environmental parameter data, the trained model parameters and structure need to be persistently stored. This includes pre-processing parameters such as the mean and standard deviation for feature standardization and the mapping relationships for feature encoding. The model output format adopts an open model exchange format to ensure deployment and use in different computing environments. A model inference interface is also provided, which receives real-time environmental parameter data as input. After feature extraction and preprocessing, this data is fed into the trained model to output predicted beef cattle health status. The model update mechanism employs periodic incremental training, fine-tuning the model at regular intervals using the latest monitoring data to adapt to changes in environmental conditions and the physiological state of beef cattle. The final dynamic prediction model comprehensively considers the time-series delay effects of environmental parameters, interaction effects between parameters, and changes in animal behavior patterns, achieving accurate predictions of beef cattle health status and providing reliable technical support for ranch management decisions. The entire model construction process requires detailed recording of experimental parameters and training logs to ensure model repeatability and interpretability.

[0081] S6. Utilize the dynamic prediction model to output the health risk prediction results for beef cattle, and generate farm management decision-making recommendations. The specific implementation is as follows:

[0082] When using a dynamic prediction model to output health risk prediction results for beef cattle, the first step is to input real-time collected environmental parameter data into the trained model. Real-time environmental parameter data is continuously collected via an IoT sensor network deployed within the ranch, including parameters such as temperature, humidity, ammonia concentration, and dust levels. This data is transmitted to the data processing center at fixed time intervals (e.g., every 5 minutes). Before inputting the data into the model, the real-time environmental parameter data undergoes the same preprocessing operations as during the training phase, including data cleaning, outlier handling, and standardization, ensuring that the input data maintains the same data format and distribution characteristics as used during model training. The preprocessed environmental parameter data is organized according to the input format required by the model, containing parameter values ​​from the current time and multiple previous time points to reflect the temporal variation characteristics of the environmental parameters. After inputting the organized data into the dynamic prediction model, the model outputs a predicted health status value for the beef cattle between 0 and 1. The closer the value is to 1, the better the health status; the closer it is to 0, the worse the health status.

[0083] When comparing the predicted health status of beef cattle with preset risk thresholds, it is necessary to first determine the method for setting the risk thresholds. Risk thresholds are determined by analyzing the correspondence between predicted and actual health status values ​​in a large amount of historical data. Receiver operating characteristic (ROC) curve analysis is used to find the optimal threshold point, which balances the false positive and false negative rates. Typically, three risk thresholds are set to classify health risk into four levels: a predicted value greater than 0.8 indicates low risk (good health); a predicted value between 0.6 and 0.8 indicates moderate risk (requiring attention); a predicted value between 0.4 and 0.6 indicates medium risk (potential health risks); and a predicted value less than 0.4 indicates high risk (poor health). In the actual comparison process, the predicted health status values ​​output by the model are compared step-by-step with these preset thresholds to determine the corresponding risk level and generate a corresponding health risk prediction report, which includes the risk level identifier and the specific predicted value.

[0084] When matching the pre-set farm management decision-making scheme library with the health risk prediction results, a comprehensive decision-making scheme library must first be established. This library contains detailed management measures for different risk levels, developed by livestock experts based on long-term practical experience and validated through actual effectiveness. For low-risk levels, the library stores routine maintenance recommendations, such as maintaining existing environmental parameter levels and daily feeding management; for general risk levels, it provides preventative measures, such as appropriately increasing ventilation or adjusting feed ratios; for medium-risk levels, it offers intervention measures, such as adjusting environmental control system parameters or conducting health checks; and for high-risk levels, it provides emergency treatment plans, such as immediate adjustment of environmental parameters and veterinary intervention. The matching process uses a rule-based mapping method, associating risk levels with corresponding measures to output specific and actionable environmental control measures and feeding management recommendations, such as specific operational instructions like adjusting temperature to a specific range, increasing ventilation frequency, and adjusting feed amounts. These recommendations are directly pushed to managers through the farm management system, providing decision support for timely response measures.

[0085] Example 2: Figure 2 A schematic diagram of the modeling system for the correlation between pasture environmental parameters and beef cattle health according to the present invention is provided. The modeling system for the correlation between pasture environmental parameters and beef cattle health includes the following modules:

[0086] The data acquisition module is used to acquire environmental parameter data of the pasture and physiological data of beef cattle within the target time period;

[0087] The time series analysis module is used to process environmental parameter data and physiological data using time series analysis algorithms to identify the delayed effect patterns of environmental parameters on physiological data.

[0088] The behavior recognition module is used to identify the behavioral pattern migration characteristics of individual beef cattle based on activity data in physiological data.

[0089] The surface analysis module is used to construct a response surface model of health status with respect to multiple environmental parameters based on environmental parameter data and physiological data. By analyzing the curvature characteristics of the response surface model, the nonlinear interaction effect between environmental parameters is quantified, and the synergistic influence characteristics of the combination of environmental parameters on the health status of beef cattle are determined by combining behavioral pattern transfer characteristics.

[0090] The model building module is used to construct a dynamic prediction model between environmental parameters and the health status of beef cattle based on delayed effect patterns, synergistic influence characteristics, and behavioral pattern transfer characteristics.

[0091] The decision output module is used to output the health risk prediction results of beef cattle using a dynamic prediction model, and generate farm management decision recommendations.

[0092] The calculations involved in the embodiments are all dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0093] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0094] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0095] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0096] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0097] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0098] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for modeling the correlation between pasture environmental parameters and beef cattle health, characterized in that, Includes the following steps: S1. Obtain environmental parameter data of the pasture and physiological data of beef cattle within the target time period; S2. Use time-series analysis algorithms to process environmental parameter data and physiological data, and identify delayed effect patterns of environmental parameters on physiological data, including: Calculate the cross-correlation function between each environmental parameter data and each physiological index in the physiological data at multiple time offsets; Find the peak point where the absolute value of the cross-correlation function is the largest; The time offset corresponding to the peak point is determined as the delay time of the influence of the corresponding environmental parameter on the corresponding physiological index, forming a delay effect model; S3. Based on activity data from physiological data, identify behavioral pattern transfer characteristics of individual beef cattle, including: Calculate the change in the statistical distribution characteristics of the current time period's activity level relative to the historical baseline activity level; Identify anomalous changes in the behavioral state transition probability matrix based on time-series patterns in activity data; By combining the changes in statistical distribution characteristics with the abnormal changes in the behavioral state transition probability matrix, behavioral pattern transfer characteristics are generated. S4. Based on environmental parameter data and physiological data, construct a response surface model of health status with respect to multiple environmental parameters. Quantify the nonlinear interaction effects between environmental parameters by analyzing the curvature characteristics of the response surface model. Combine this with behavioral pattern transfer characteristics to determine the synergistic effects of environmental parameter combinations on the health status of beef cattle, including: A response surface model was established by fitting the functional relationship between environmental parameter data and health status indicator parameters in physiological data using the multinomial regression method. Calculate the Hessian matrix of the response surface model at each point in the environmental parameter value space, and determine the intensity of the nonlinear interaction effect between environmental parameters by analyzing the eigenvalue distribution of the Hessian matrix. By weighted fusion of the intensity of nonlinear interaction effects and behavioral pattern transfer characteristics, the synergistic influence characteristics of environmental parameter combinations on the health status of beef cattle are generated. S5. Based on the delayed effect model, synergistic influence characteristics, and behavioral pattern transfer characteristics, a dynamic prediction model between environmental parameters and the health status of beef cattle is constructed. S6. Utilize the dynamic prediction model to output the health risk prediction results of beef cattle and generate farm management decision-making suggestions.

2. The method for modeling the correlation between pasture environmental parameters and beef cattle health according to claim 1, characterized in that, Obtain environmental parameter data of the pasture and physiological data of beef cattle within the target time period, including: Environmental parameter data is collected by IoT sensors deployed in the ranch environment. The environmental parameter data includes at least temperature and humidity, ammonia concentration and dust amount. Physiological data are collected by wearing monitoring devices on individual beef cattle, including at least body temperature and activity level.

3. The method for modeling the correlation between pasture environmental parameters and beef cattle health according to claim 1, characterized in that, The cross-correlation function of each environmental parameter data and each physiological indicator in the physiological data at multiple time offsets is calculated as follows: After standardizing and preprocessing the time series of each environmental parameter and the time series of physiological indicators, the correlation coefficient of the two series at different lag time points is calculated within a preset time offset range; the time offset is gradually adjusted by a sliding window method, and the correlation coefficient value at each offset is recorded to form a cross-correlation function sequence; finally, the cross-correlation function reflecting the degree of correlation between environmental parameters and physiological indicators at different time delays is obtained.

4. The method for modeling the correlation between pasture environmental parameters and beef cattle health according to claim 1, characterized in that, The Hessian matrix of the response surface model at each point in the environmental parameter value space is calculated as follows: the second-order partial derivatives of the constructed response surface model are obtained to form the Hessian matrix; representative sampling points are selected within the range of environmental parameter values, and the Hessian matrix at each sampling point is calculated; by analyzing the sign and magnitude of the eigenvalues ​​of the Hessian matrix, the strength and direction of the interaction between environmental parameters are determined. The eigenvalue distribution characteristics are quantified into an index of the intensity of nonlinear interaction effects.

5. The method for modeling the correlation between pasture environmental parameters and beef cattle health according to claim 1, characterized in that, Based on the delayed effect model, synergistic influence characteristics, and behavioral pattern transfer characteristics, a dynamic prediction model between environmental parameters and the health status of beef cattle is constructed, including: The delayed effect pattern is used as the time series feature input, the synergistic influence feature is used as the interaction effect input, and the behavioral pattern transfer feature is used as the behavioral state input, and all of them are input into the time series machine learning model. A multi-feature fusion training strategy was adopted to train the time series machine learning model, enabling the model to simultaneously learn the temporal delay patterns of environmental parameters, the interaction effects between parameters, and the migration patterns of animal behavior. The output is a dynamic prediction model that can predict the health status of beef cattle based on environmental parameter data.

6. The method for modeling the correlation between pasture environmental parameters and beef cattle health according to claim 1, characterized in that, The dynamic prediction model outputs the health risk prediction results for beef cattle, generating farm management decision-making recommendations, including: The environmental parameter data collected in real time is input into the dynamic prediction model to obtain the predicted value of the health status of beef cattle; The predicted health status of beef cattle is compared with a preset risk threshold to generate different levels of health risk prediction results. Based on the level of health risk prediction results, a pre-set ranch management decision-making scheme library is matched, and corresponding environmental control measures and feeding management suggestions are output.

7. A system for modeling the correlation between pasture environmental parameters and beef cattle health, used to implement the method for modeling the correlation between pasture environmental parameters and beef cattle health as described in any one of claims 1-6, characterized in that, Includes the following modules: The data acquisition module is used to acquire environmental parameter data of the pasture and physiological data of beef cattle within the target time period; The time series analysis module is used to process environmental parameter data and physiological data using time series analysis algorithms to identify the delayed effect patterns of environmental parameters on physiological data. The behavior recognition module is used to identify the behavioral pattern migration characteristics of individual beef cattle based on activity data in physiological data. The surface analysis module is used to construct a response surface model of health status with respect to multiple environmental parameters based on environmental parameter data and physiological data. By analyzing the curvature characteristics of the response surface model, the nonlinear interaction effect between environmental parameters is quantified, and the synergistic influence characteristics of the combination of environmental parameters on the health status of beef cattle are determined by combining behavioral pattern transfer characteristics. The model building module is used to construct a dynamic prediction model between environmental parameters and the health status of beef cattle based on delayed effect patterns, synergistic influence characteristics, and behavioral pattern transfer characteristics. The decision output module is used to output the health risk prediction results of beef cattle using a dynamic prediction model, and generate farm management decision recommendations.

Images