Method, device, medium and equipment for predicting selenium addition amount by using pig blood biochemical indexes

Abstract

The present disclosure relates to a method, device, medium and equipment for predicting the amount of selenium addition using pig blood biochemical indicators, belonging to the field of data processing technology. The method comprises: extracting blood biomarker data related to the immune response mechanism by performing biochemical testing on collected pig blood samples; integrating the blood biomarker data, quantitative parameters of pig growth performance and dynamic ecological variables of the pig breeding environment into a multidimensional feature vector and inputting it into a selenium demand prediction model, obtaining the pig's physiological requirement for selenium output by the selenium demand prediction model, the selenium demand prediction model using the multidimensional feature vector of historical samples as input data and the corresponding reference selenium requirement as a supervision label, and being constructed through multi-layer neural network architecture training; calculating and adjusting the ratio of selenium addition in pig feed based on the pig's physiological requirement for selenium. In this way, the problem of excess or insufficient selenium caused by the traditional fixed addition method is avoided.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a method, device, medium and equipment for predicting the amount of selenium addition by using pig blood biochemical indicators. Background Art

[0002] Selenium has two key properties in pig nutrition. First, it is physiologically essential. As an indispensable trace element for pigs, selenium is deeply involved in the synthesis of antioxidant enzymes such as glutathione peroxidase (GPx), directly regulating core physiological processes such as antioxidant defense, immune regulation, and reproductive growth. Second, it is dose-sensitive, requiring strict control of selenium intake. A deficiency in GPx activity can lead to decreased immunity, growth retardation, and, in severe cases, metabolic diseases such as white muscle disease. Excessive intake can exceed the pig's tolerance threshold, leading to selenium toxicity, hair loss, neurological dysfunction, and even death. Blood immune indicators are invaluable in precisely regulating selenium supplementation. These indicators function as biomarkers. Blood GPx activity is positively correlated with blood selenium concentration, allowing for a quantitative assessment of selenium absorption efficiency and providing a precise indicator of selenium nutritional status in pigs. Immunoglobulin (IgG, IgA, etc.) levels directly reflect the degree of immune system activation and the strength of the immune response. By dynamically monitoring these indicators, we can break the limitations of traditional empirical addition methods, build a dynamic model, and achieve real-time matching of selenium supply with the physiological needs of pigs. Traditional selenium addition methods have obvious technical bottlenecks. First, static addition relies too much on fixed dosage standards, ignoring the differences in pig breeds, growth stage characteristics, and environmental pressures, resulting in significant differences in selenium intake levels within the group; second, there is a lag in regulation. Due to the lack of real-time monitoring methods, it is impossible to timely monitor the fluctuations in pigs' selenium demand under conditions such as disease invasion and stress response, which in turn leads to reduced pig production performance and increased morbidity; third, the management model is extensive. Excessive selenium addition will result in a waste of resources, while insufficient addition will lead to low immunity, increased morbidity, and even death in pigs, ultimately pushing up overall costs and reducing economic benefits. Therefore, it is of great significance to achieve dynamic and precise regulation of selenium addition in live pigs. Summary of the Invention

[0003] The purpose of the present disclosure is to provide a method, device, medium and equipment for predicting the amount of selenium addition using pig blood biochemical indicators to solve the above-mentioned related technical problems.

[0004] To achieve the above objectives, in a first aspect, the present disclosure provides a method for predicting the amount of selenium supplementation using pig blood biochemical indicators, comprising:

[0005] Through biochemical testing of the collected pig blood samples, blood biomarker data related to the immune response mechanism were extracted;

[0006] The blood biomarker data, quantitative parameters of pig growth performance, and dynamic ecological variables of the pig breeding environment are integrated into a multidimensional feature vector, which is input into a selenium requirement prediction model based on deep learning to obtain the pig's physiological requirement for selenium as output by the selenium requirement prediction model, wherein the selenium requirement prediction model uses the multidimensional feature vectors of historical samples as input data and the corresponding reference selenium requirement as a supervision label, and is constructed through multi-layer neural network architecture training;

[0007] The proportion of selenium added to the pig feed is calculated and adjusted based on the pig's physiological demand for selenium.

[0008] Optionally, the biochemical test of the collected pig blood samples is performed to extract blood biomarker data related to the immune response mechanism, specifically including:

[0009] The collected pig blood samples were analyzed for enzyme activity and protein content to obtain data on the catalytic activity of glutathione peroxidase (GSH-Px), the antioxidant activity of superoxide dismutase (SOD), the serum concentration of immunoglobulins (Ig), and the secretion levels of cytokines;

[0010] The blood biomarker data include the catalytic activity data of glutathione peroxidase, the antioxidant activity data of superoxide dismutase, the serum concentration data of immunoglobulins, and the cytokine secretion level data.

[0011] Optionally, the method further includes the steps of constructing a selenium demand prediction model and optimizing parameters, specifically including:

[0012] Constructing a structured training data set, the training data set comprising a sample multidimensional feature vector and a corresponding reference selenium requirement label, wherein the sample multidimensional feature vector includes blood biomarker data of historical samples, quantitative growth performance parameters, and dynamic ecological variables of the feeding environment;

[0013] Inputting the sample multidimensional feature vector into a prediction model to be trained based on a deep neural network to obtain a first predicted selenium requirement output by the prediction model to be trained;

[0014] Based on the deviation between the first predicted selenium requirement and the reference selenium requirement, constructing a loss function and calculating a loss value;

[0015] Utilizing the loss value, the weight parameters and bias parameters of the prediction model to be trained are iteratively updated through an adaptive algorithm combining stochastic gradient descent with momentum optimization to approach the global optimal solution.

[0016] Optionally, inputting the sample multidimensional feature vector into a prediction model to be trained based on a deep neural network to obtain a first predicted selenium requirement output by the prediction model to be trained specifically includes:

[0017] The training data set is divided into a predetermined proportion, wherein 70% of the data is used as a training subset for model parameter learning, and 30% of the data is used as a test subset for model performance evaluation;

[0018] Inputting the sample multidimensional feature vector in the training subset into the prediction model to be trained to obtain a first predicted selenium requirement output by the prediction model to be trained;

[0019] The step of constructing a loss function and calculating a loss value based on a deviation between the first predicted selenium requirement and the reference selenium requirement specifically includes:

[0020] Based on the numerical deviation between the first predicted selenium requirement and the reference selenium requirement in the training subset, a mean square error loss function is constructed and the corresponding loss value is calculated. The loss value is used to evaluate the model prediction error and guide the optimization of the weight parameters and bias parameters of the prediction model to be trained.

[0021] Optionally, the method comprises:

[0022] After the training iteration of the prediction model to be trained is completed, the multidimensional feature vector of the samples in the test subset is input into the trained prediction model to obtain a second predicted selenium requirement output by the trained prediction model;

[0023] Obtaining a reference selenium requirement in the test subset as a performance evaluation benchmark;

[0024] Calculating an absolute deviation or relative error between the second predicted selenium requirement and the reference selenium requirement;

[0025] When the absolute deviation value or the relative error value is less than or equal to a preset first threshold, it is determined that the output result of the trained prediction model meets the prediction accuracy requirement;

[0026] When the absolute deviation value or the relative error value is greater than a preset first threshold, it is determined that the output result of the trained prediction model does not reach the expected accuracy;

[0027] The prediction accuracy of the prediction model obtained by training on the test subset is counted. When the prediction accuracy is greater than a preset second threshold, the model training process is terminated; when the prediction accuracy is less than or equal to the preset second threshold, the model training process is continued.

[0028] Optionally, the method further comprises:

[0029] According to a preset time period, the embedded wearable blood sampling device deployed on the pig body is used to periodically perform automated collection of pig blood samples;

[0030] A multi-sensor network deployed in the pig breeding space collects environmental state parameters in real time, including temperature distribution data, humidity fluctuation data, and air quality comprehensive index data. The dynamic ecological variables of the breeding environment include the temperature distribution data, humidity fluctuation data, and air quality comprehensive index data.

[0031] The growth monitoring system dynamically records the quantitative parameters of pig growth performance, including daily average weight gain data, daily feed intake data and feed conversion efficiency data.

[0032] Optionally, the method comprises:

[0033] When the physiological requirement of selenium output by the selenium requirement prediction model exceeds a preset critical threshold, a notification signal is sent to the user terminal or the feeding management system through the communication interface. The notification signal is used to instruct the user or the feeding management system to adjust the proportion of selenium added in the pig feed according to the physiological requirement of selenium.

[0034] In a second aspect, a device for predicting the amount of selenium to be added using pig blood biochemical indicators is provided, comprising:

[0035] An extraction module is used to extract blood biomarker data related to the immune response mechanism by performing biochemical tests on the collected pig blood samples;

[0036] an acquisition module for integrating the blood biomarker data, quantitative parameters of pig growth performance, and dynamic ecological variables of the pig breeding environment into a multidimensional feature vector, inputting the vector into a selenium requirement prediction model based on deep learning, and obtaining the pig's physiological requirement for selenium as output by the selenium requirement prediction model, wherein the selenium requirement prediction model uses the multidimensional feature vector of historical samples as input data and the corresponding reference selenium requirement as a supervision label, and is constructed through multi-layer neural network architecture training;

[0037] The calculation module is used to calculate and adjust the proportion of selenium added in the pig feed according to the pig's physiological demand for selenium.

[0038] Optionally, the extraction module is used to:

[0039] The collected pig blood samples were analyzed for enzyme activity and protein content to obtain data on the catalytic activity of glutathione peroxidase (GSH-Px), the antioxidant activity of superoxide dismutase (SOD), the serum concentration of immunoglobulins (Ig), and the secretion levels of cytokines;

[0040] The blood biomarker data include the catalytic activity data of glutathione peroxidase, the antioxidant activity data of superoxide dismutase, the serum concentration data of immunoglobulins, and the cytokine secretion level data.

[0041] Optionally, the device includes a training module for:

[0042] Constructing a structured training data set, the training data set comprising a sample multidimensional feature vector and a corresponding reference selenium requirement label, wherein the sample multidimensional feature vector includes blood biomarker data of historical samples, quantitative growth performance parameters, and dynamic ecological variables of the feeding environment;

[0043] Inputting the sample multidimensional feature vector into a prediction model to be trained based on a deep neural network to obtain a first predicted selenium requirement output by the prediction model to be trained;

[0044] Based on the deviation between the first predicted selenium requirement and the reference selenium requirement, constructing a loss function and calculating a loss value;

[0045] Utilizing the loss value, the weight parameters and bias parameters of the prediction model to be trained are iteratively updated through an adaptive algorithm combining stochastic gradient descent with momentum optimization to approach the global optimal solution.

[0046] Optionally, the training module is used to:

[0047] The training data set is divided into a predetermined proportion, wherein 70% of the data is used as a training subset for model parameter learning, and 30% of the data is used as a test subset for model performance evaluation;

[0048] Inputting the sample multidimensional feature vector in the training subset into the prediction model to be trained to obtain a first predicted selenium requirement output by the prediction model to be trained;

[0049] The step of constructing a loss function and calculating a loss value based on a deviation between the first predicted selenium requirement and the reference selenium requirement specifically includes:

[0050] Based on the numerical deviation between the first predicted selenium requirement and the reference selenium requirement in the training subset, a mean square error loss function is constructed and the corresponding loss value is calculated. The loss value is used to evaluate the model prediction error and guide the optimization of the weight parameters and bias parameters of the prediction model to be trained.

[0051] Optionally, the device comprises:

[0052] A testing module is configured to input the multidimensional feature vector of the samples in the test subset into the trained prediction model after the training iteration of the prediction model to be trained is completed, so as to obtain a second predicted selenium requirement output by the trained prediction model;

[0053] A reference selenium element requirement acquisition module, configured to acquire the reference selenium element requirement in the test subset as a performance evaluation benchmark;

[0054] a processing module, configured to calculate an absolute deviation or a relative error between the second predicted selenium element requirement and the reference selenium element requirement;

[0055] A first judgment module is configured to determine that an output result of the trained prediction model meets a prediction accuracy requirement when the absolute deviation value or the relative error value is less than or equal to a preset first threshold value;

[0056] A second judgment module is configured to determine that an output result of the trained prediction model does not reach an expected accuracy when the absolute deviation value or the relative error value is greater than a preset first threshold;

[0057] The statistical module is used to count the prediction accuracy of the prediction model obtained by training on the test subset. When the prediction accuracy is greater than a preset second threshold, the model training process is terminated; when the prediction accuracy is less than or equal to the preset second threshold, the model training process is continued.

[0058] Optionally include:

[0059] The first collection module is used to periodically perform automated collection of pig blood samples according to a preset time period using an embedded wearable blood sampling device deployed on the pig's body;

[0060] A second acquisition module is configured to collect environmental state parameters in real time through a multi-sensor network deployed in the pig breeding space, including temperature distribution data, humidity fluctuation data, and air quality comprehensive index data, where the dynamic ecological variables of the breeding environment include the temperature distribution data, humidity fluctuation data, and air quality comprehensive index data;

[0061] The recording module is used to dynamically record the quantitative parameters of pig growth performance through the growth monitoring system, including daily average weight gain data, daily feed intake data and feed conversion efficiency data.

[0062] Optionally, the device comprises:

[0063] The sending module is used to send a notification signal to the user terminal or feeding management system through the communication interface when the physiological requirement of selenium output by the selenium requirement prediction model exceeds a preset critical threshold. The notification signal is used to instruct the user or feeding management system to adjust the proportion of selenium added in pig feed according to the physiological requirement of selenium.

[0064] According to a third aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0065] In a fourth aspect, a device is provided, comprising:

[0066] a memory having a computer program stored thereon;

[0067] A processor is used to execute the computer program in the memory to implement the steps of any one of the methods in the first aspect.

[0068] The above scheme achieves an accurate assessment of the selenium demand of pigs by establishing a model predictive relationship between blood biomarker data, quantitative parameters of pig growth performance, dynamic ecological variables of the pig breeding environment and selenium demand, avoiding the problem of excessive waste or insufficient selenium caused by the traditional fixed addition method, improving the utilization efficiency of selenium and reducing feed costs.

[0069] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following detailed description, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings:

[0071] Figure 1 It is a flow chart of a method for predicting the amount of selenium addition by using pig blood biochemical indicators, shown in an exemplary embodiment of the present disclosure.

[0072] Figure 2 It is a block diagram of an apparatus for predicting the amount of selenium addition using pig blood biochemical indicators, shown in an exemplary embodiment of the present disclosure.

[0073] Figure 3 It is a block diagram of an electronic device shown in an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0074] The following describes the specific embodiments of the present disclosure in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present disclosure and are not intended to limit the present disclosure.

[0075] Before introducing the method for predicting the amount of selenium addition by using pig blood biochemical indicators according to an embodiment of the present disclosure, the relevant scenarios of the embodiment of the present disclosure are first exemplified.

[0076] Selenium has two key properties in pig nutrition. First, it is physiologically essential. As an indispensable trace element for pigs, selenium is deeply involved in the synthesis of antioxidant enzymes such as glutathione peroxidase (GPx), directly regulating core physiological processes such as antioxidant defense, immune regulation, and reproductive growth. Second, it is dose-sensitive, requiring strict control of selenium intake. A deficiency in GPx activity can lead to decreased immunity, growth retardation, and, in severe cases, metabolic diseases such as white muscle disease. Excessive intake can exceed the pig's tolerance threshold, leading to selenium toxicity, hair loss, neurological dysfunction, and even death. Blood immune indicators are invaluable in precisely regulating selenium supplementation. These indicators function as biomarkers. Blood GPx activity is positively correlated with blood selenium concentration, allowing for a quantitative assessment of selenium absorption efficiency and providing a precise indicator of selenium nutritional status in pigs. Immunoglobulin (IgG, IgA, etc.) levels directly reflect the degree of immune system activation and the strength of the immune response. By dynamically monitoring these indicators, we can break the limitations of traditional empirical addition methods, build a dynamic model, and achieve real-time matching of selenium supply with the physiological needs of pigs. Traditional selenium addition methods have obvious technical bottlenecks. First, static addition relies too much on fixed dosage standards, ignoring the differences in pig breeds, growth stage characteristics, and environmental pressures, resulting in significant differences in selenium intake levels within the group; second, there is a lag in regulation. Due to the lack of real-time monitoring methods, it is impossible to timely monitor the fluctuations in pigs' selenium demand under conditions such as disease invasion and stress response, which in turn leads to reduced pig production performance and increased morbidity; third, the management model is extensive. Excessive selenium addition will result in a waste of resources, while insufficient addition will lead to low immunity, increased morbidity, and even death in pigs, ultimately pushing up overall costs and reducing economic benefits. Therefore, it is of great significance to achieve dynamic and precise regulation of selenium addition in live pigs.

[0077] To this end, the present disclosure provides a method for predicting the amount of selenium to be added using pig blood biochemical indicators. Figure 1 This is a flow chart of a method for predicting selenium addition using pig blood biochemical indicators according to an exemplary embodiment of the present disclosure. Figure 1 , the method comprising:

[0078] In step S11, blood biomarker data related to the immune response mechanism is extracted by performing biochemical tests on the collected pig blood samples.

[0079] In one embodiment, pig blood samples can be collected manually and periodically. As an example, pig blood samples can be collected manually using a blood sampling device at a set time period (e.g., 1 day, 3 days, or 7 days). In another possible embodiment, pig blood samples can also be collected automatically and periodically based on a set time period using an embedded wearable blood sampling device deployed on the pig.

[0080] Based on the collected pig blood, blood biomarker data related to immune response mechanisms can be extracted. For example, the value of each biomarker in multiple pig blood samples is determined, each pig blood sample is collected from a pig, and a raw data matrix is constructed based on the value of each biomarker in each pig blood sample, wherein the rows in the raw data matrix are the values of each biomarker for the same pig, and the columns are the values of the same biomarker for different pigs.

[0081] Then, the biomarker values of each pig blood sample are standardized according to the sample mean and standard deviation of the values of each biomarker in each pig blood sample to obtain a standardized data matrix Z;

[0082] Furthermore, the covariance matrix C is calculated based on the standardized data matrix Z. Calculation of the covariance matrix: For example,

[0083] Where C is a symmetric positive semidefinite matrix whose element c jk represents the covariance between the jth and kth biomarkers:

[0084] Then perform eigendecomposition on the covariance matrix C: Where: k is the kth largest eigenvalue (λ1≥λ2≥⋯≥λ p ≥0), v k ∈R p is the corresponding unit eigenvector (||v k ||2=1), different eigenvectors are mutually orthogonal: v j ⊤v k =0.

[0085] Then select the first m principal components (m≤p) and construct the projection matrix: Project the original data into the principal component space to obtain the reduced-dimensional data:

[0086] Among them, the i-th row of Y represents the coordinate of the i-th sample in the principal component space.

[0087] Furthermore, the proportion of variance explained by the kth principal component is: The cumulative variance contribution rate is the sum of the variance proportions explained by the first m principal components:

[0088] Then, based on the explained variance ratio and the cumulative variance contribution rate, the correlation coefficient between each biomarker and the Kth principal component is determined, and then based on the correlation coefficient corresponding to each biomarker, the blood biomarkers related to the immune response mechanism are determined.

[0089] In one possible embodiment, the biochemical test of the collected pig blood samples to extract blood biomarker data related to the immune response mechanism includes:

[0090] The collected pig blood samples were analyzed for enzyme activity and protein content to obtain data on the catalytic activity of glutathione peroxidase (GSH-Px), the antioxidant activity of superoxide dismutase (SOD), the serum concentration of immunoglobulins (Ig), and the secretion levels of cytokines;

[0091] The blood biomarker data include the catalytic activity data of glutathione peroxidase, the antioxidant activity data of superoxide dismutase, the serum concentration data of immunoglobulins, and the cytokine secretion level data;

[0092] Data on the catalytic activity of glutathione peroxidase, the antioxidant activity of superoxide dismutase, serum concentrations of immunoglobulins, and cytokine secretion levels can reflect the selenium nutritional status and immune status of animals. These indicators help provide a scientific basis for the supplementation of selenium.

[0093] In step S12, the blood biomarker data, the quantitative parameters of the pig's growth performance, and the dynamic ecological variables of the pig breeding environment are integrated into a multidimensional feature vector and input into a selenium requirement prediction model based on deep learning to obtain the pig's physiological requirement for selenium output by the selenium requirement prediction model. The selenium requirement prediction model uses the multidimensional feature vector of historical samples as input data and the corresponding reference selenium requirement as a supervision label, and is constructed through multi-layer neural network architecture training.

[0094] In one embodiment, a multi-sensor network deployed within the pig breeding space can collect environmental parameters in real time, including temperature distribution data, humidity fluctuation data, and a comprehensive air quality index. The dynamic ecological variables of the breeding environment include these temperature distribution data, humidity fluctuation data, and a comprehensive air quality index. Furthermore, a growth monitoring system can dynamically record quantitative parameters of pig growth performance, including average daily weight gain, daily feed intake, and feed conversion efficiency data.

[0095] To implement real-time collection of environmental parameters, a network of various sensors can be deployed within the pig house. For example, multiple temperature and humidity sensors can be installed in different areas of the pig house (such as resting areas, feeding areas, and at different heights) to capture differences in temperature distribution and real-time fluctuations in humidity. Furthermore, sensors for ammonia (NH3), carbon dioxide (CO2), hydrogen sulfide (H2S), and particulate matter can be deployed at key locations (such as near sewage outlets and areas with poor air circulation) to monitor air quality. These indicators can be combined or fused using specific algorithms to generate a comprehensive air quality index. These sensors are connected to data collection nodes or gateways via wireless (such as Zigbee and LoRa) or wired connections. They collect readings at a preset frequency (e.g., every minute or several minutes) and transmit the time-stamped data in real time to a central data processing system, forming a dynamic database of ecological variables in the piggery environment for use in predictive models.

[0096] To dynamically record quantitative parameters of pig growth performance, a growth monitoring system that integrates modern identification and weighing technologies can be used. For example, each pig is equipped with an electronic identification tag (such as an RFID ear tag). Automatic weighing platforms or channels are set up in pig house passageways or specific areas (such as watering points or feeding entrances). When pigs with ear tags pass through, the system automatically identifies them and records their current weight. This continuous recording allows for the calculation of average daily weight gain data. Simultaneously, intelligent feeders equipped with RFID identifiers and weighing sensors can identify each pig that comes to feed and accurately record their feed intake at each feeding, compiling daily feed intake data. Based on accumulated feed intake and weight gain data over a period of time, the system can automatically calculate feed conversion efficiency. All of these quantitative growth parameters are linked to individual pig identity and time, and are dynamically and automatically recorded and updated to a central system, ensuring that the model captures the most up-to-date and accurate individual growth performance information.

[0097] The following is an exemplary description of how to obtain the selenium demand prediction model. In one possible embodiment, the method includes the steps of constructing a selenium demand prediction model and optimizing parameters, specifically including:

[0098] Constructing a structured training data set, the training data set comprising a sample multidimensional feature vector and a corresponding reference selenium requirement label, wherein the sample multidimensional feature vector includes blood biomarker data of historical samples, quantitative growth performance parameters, and dynamic ecological variables of the feeding environment;

[0099] Inputting the sample multidimensional feature vector into a prediction model to be trained based on a deep neural network to obtain a first predicted selenium requirement output by the prediction model to be trained;

[0100] Based on the deviation between the first predicted selenium requirement and the reference selenium requirement, constructing a loss function and calculating a loss value;

[0101] Utilizing the loss value, the weight parameters and bias parameters of the prediction model to be trained are iteratively updated through an adaptive algorithm combining stochastic gradient descent with momentum optimization to approach the global optimal solution.

[0102] As an example implementation of a multi-layer neural network architecture, a multi-layer perceptron (MLP) neural network structure can be designed, consisting of an input layer, three hidden layers, and an output layer. The input layer has the same number of neurons as the multidimensional feature vector (approximately 20-30), and is used to receive input data. The three hidden layers have 64, 32, and 16 neurons, respectively, forming a pyramid structure. This facilitates feature extraction and dimensionality reduction. The output layer can consist of a single neuron, directly predicting the physiological requirement for selenium (e.g., mg / kg feed).

[0103] Each hidden layer can be followed by a batch normalization layer to accelerate training and improve stability. Reluctance linear unit (ReLU) can be used as the activation function to enhance the model's nonlinear expressiveness. In some embodiments, to prevent overfitting, a dropout layer can be added between hidden layers with a dropout rate set to 0.3 and an L2 regularization technique applied with a regularization coefficient set to 0.001.

[0104] In this way, model training can be performed. For example, the training subset can be divided into multiple batches. The batch size can be set to 64, for example. For each batch, an optimization algorithm (such as an adaptive algorithm, specifically implemented as mini-batch gradient descent) can be used. During each training cycle (epoch), the model can traverse the multidimensional feature vectors of the samples in the entire training subset and calculate a first predicted selenium requirement through forward propagation. Based on the numerical deviation between the first predicted selenium requirement and the reference selenium requirement in the training subset, a mean squared error (MSE) loss function is constructed and the corresponding loss value is calculated. Then, the gradient is calculated through the backpropagation algorithm, and the network weights and bias parameters are iteratively updated.

[0105] It should be understood that the training process can be repeated multiple times. For example, the model may require 100-200 training cycles to converge to a stable state.

[0106] In one possible implementation, inputting the sample multidimensional feature vector into a prediction model to be trained based on a deep neural network to obtain a first predicted selenium requirement output by the prediction model to be trained specifically includes:

[0107] Dividing the structured training dataset into a predetermined proportion (e.g., 70% of the data as a training subset for model parameter learning, and 30% of the data as a test subset for model performance evaluation);

[0108] Inputting the sample multidimensional feature vector in the training subset into the prediction model to be trained to obtain a first predicted selenium requirement output by the prediction model to be trained;

[0109] The step of constructing a loss function and calculating a loss value based on a deviation between the first predicted selenium requirement and the reference selenium requirement specifically includes:

[0110] Based on the numerical deviation between the first predicted selenium requirement and the reference selenium requirement in the training subset, a mean square error loss function is constructed and the corresponding loss value is calculated. The loss value is used to evaluate the model prediction error and guide the optimization of the weight parameters and bias parameters of the prediction model to be trained.

[0111] That is, 70% of the data set can be used as a training set, and 30% of the data set can be used as a test set. The model is thus trained using the training set. The specific training method can be found in the description of the above embodiment, and will not be detailed here for the sake of brevity.

[0112] In a possible implementation, the performance of the model can be evaluated based on the test subset. Exemplarily, the method can also include:

[0113] After the training iteration of the prediction model to be trained is completed, the multidimensional feature vector of the samples in the test subset is input into the trained prediction model to obtain a second predicted selenium requirement output by the trained prediction model. In addition, a reference selenium requirement in the test subset can be obtained as a performance evaluation benchmark.

[0114] Based on the acquired second predicted selenium element requirement and the reference selenium element requirement, an absolute deviation value or a relative error value between the second predicted selenium element requirement and the reference selenium element requirement may be calculated.

[0115] When the absolute deviation value or the relative error value is less than or equal to the preset first threshold, it is determined that the output result of the trained prediction model meets the prediction accuracy requirement; when the absolute deviation value or the relative error value is greater than the preset first threshold, it is determined that the output result of the trained prediction model does not reach the expected accuracy.

[0116] In one possible implementation, the prediction accuracy of the trained prediction model on the test subset can be calculated. For example, the data in the test subset is input into the model one by one, and the proportion of model predictions that meet the accuracy requirements is calculated to obtain the prediction accuracy. When the prediction accuracy is greater than a preset second threshold, the model training process can be terminated; if the prediction accuracy is less than or equal to the second threshold, the model training process can continue.

[0117] Among them, the above-mentioned first threshold and second threshold can be set based on needs, such as setting the first threshold to a specific error tolerance (for example, 0.05) and the second threshold to a target accuracy rate (for example, 95%). The embodiments of the present disclosure do not limit this.

[0118] Reference Figure 1 In step S13, the ratio of selenium added to the pig feed is calculated and adjusted according to the pig's physiological demand for selenium.

[0119] In one possible embodiment, when the physiological requirement of selenium output by the selenium requirement prediction model exceeds a preset critical threshold, a notification signal is sent to the user terminal or feeding management system through the communication interface, and the notification signal is used to instruct the user or feeding management system to adjust the proportion of selenium added in pig feed according to the physiological requirement of selenium.

[0120] For example, the notification signal can be transmitted as a digital signal over a network and presented as an alert, push notification, or other notification on a user terminal (e.g., a mobile app, computer software interface) or feeding management system. In this way, the user or automated system can be informed of the need to pay attention to or adjust selenium supplementation based on the notification signal. In this way, the user or system can calculate and adjust the amount of selenium added to pig feed based on the physiological requirement, thereby achieving precise selenium supplementation or adjustment.

[0121] The above scheme achieves an accurate assessment of the selenium demand of pigs by establishing a model predictive relationship between blood biomarker data, quantitative parameters of pig growth performance, dynamic ecological variables of the pig breeding environment and selenium demand, avoiding the problem of excessive waste or insufficient selenium caused by the traditional fixed addition method, improving the utilization efficiency of selenium and reducing feed costs.

[0122] Based on the same inventive concept, an embodiment of the present disclosure provides a device for predicting the amount of selenium to be added using pig blood biochemical indicators. Figure 2 This is a block diagram of a device that uses pig blood biochemical indicators to predict the amount of selenium added. Figure 2 The device for predicting the amount of selenium added by using pig blood biochemical indicators includes:

[0123] Extraction module 301, for extracting blood biomarker data related to the immune response mechanism by performing biochemical tests on the collected pig blood samples;

[0124] An acquisition module 302 is configured to integrate the blood biomarker data, the quantitative parameters of the pig's growth performance, and the dynamic ecological variables of the pig's breeding environment into a multidimensional feature vector, and input the vector into a selenium requirement prediction model based on deep learning to obtain the pig's physiological requirement for selenium as output by the selenium requirement prediction model, wherein the selenium requirement prediction model uses the multidimensional feature vector of historical samples as input data and the corresponding reference selenium requirement as a supervision label, and is constructed through multi-layer neural network architecture training;

[0125] The calculation module 303 is used to calculate and adjust the proportion of selenium added in the pig feed according to the physiological demand of the pig for selenium.

[0126] The above-mentioned device achieves an accurate assessment of the selenium needs of pigs by establishing a model predictive relationship between blood biomarker data, quantitative parameters of pig growth performance, dynamic ecological variables of the pig breeding environment and selenium requirements, avoiding the problem of excessive waste or insufficient selenium caused by the traditional fixed addition method, improving the utilization efficiency of selenium and reducing feed costs.

[0127] Optionally, the extraction module 301 is used to:

[0128] The collected pig blood samples were analyzed for enzyme activity and protein content to obtain data on the catalytic activity of glutathione peroxidase, the antioxidant activity of superoxide dismutase, the serum concentration of immunoglobulins, and the secretion level of cytokines;

[0129] The blood biomarker data include the catalytic activity data of glutathione peroxidase, the antioxidant activity data of superoxide dismutase, the serum concentration data of immunoglobulins, and the cytokine secretion level data.

[0130] Optionally, the device includes a training module for:

[0131] Constructing a structured training data set, the training data set comprising a sample multidimensional feature vector and a corresponding reference selenium requirement label, wherein the sample multidimensional feature vector includes blood biomarker data of historical samples, quantitative growth performance parameters, and dynamic ecological variables of the feeding environment;

[0132] Inputting the sample multidimensional feature vector into a prediction model to be trained based on a deep neural network to obtain a first predicted selenium requirement output by the prediction model to be trained;

[0133] Based on the deviation between the first predicted selenium requirement and the reference selenium requirement, constructing a loss function and calculating a loss value;

[0134] Utilizing the loss value, the weight parameters and bias parameters of the prediction model to be trained are iteratively updated through an adaptive algorithm combining stochastic gradient descent with momentum optimization to approach the global optimal solution.

[0135] Optionally, the training module is used to:

[0136] The training data set is divided into a predetermined proportion, wherein 70% of the data is used as a training subset for model parameter learning, and 30% of the data is used as a test subset for model performance evaluation;

[0137] Inputting the sample multidimensional feature vector in the training subset into the prediction model to be trained to obtain a first predicted selenium requirement output by the prediction model to be trained;

[0138] The step of constructing a loss function and calculating a loss value based on a deviation between the first predicted selenium requirement and the reference selenium requirement specifically includes:

[0139] Based on the numerical deviation between the first predicted selenium requirement and the reference selenium requirement in the training subset, a mean square error loss function is constructed and the corresponding loss value is calculated. The loss value is used to evaluate the model prediction error and guide the optimization of the weight parameters and bias parameters of the prediction model to be trained.

[0140] Optionally, the device comprises:

[0141] A testing module is configured to input the multidimensional feature vector of the samples in the test subset into the trained prediction model after the training iteration of the prediction model to be trained is completed, so as to obtain a second predicted selenium requirement output by the trained prediction model;

[0142] A reference selenium element requirement acquisition module, configured to acquire the reference selenium element requirement in the test subset as a performance evaluation benchmark;

[0143] a processing module, configured to calculate an absolute deviation or a relative error between the second predicted selenium element requirement and the reference selenium element requirement;

[0144] A first judgment module is configured to determine that an output result of the trained prediction model meets a prediction accuracy requirement when the absolute deviation value or the relative error value is less than or equal to a preset first threshold value;

[0145] A second judgment module is configured to determine that an output result of the trained prediction model does not reach an expected accuracy when the absolute deviation value or the relative error value is greater than a preset first threshold;

[0146] The statistical module is used to count the prediction accuracy of the prediction model obtained by training on the test subset. When the prediction accuracy is greater than a preset second threshold, the model training process is terminated; when the prediction accuracy is less than or equal to the preset second threshold, the model training process is continued.

[0147] Optionally include:

[0148] The first collection module is used to periodically perform automated collection of pig blood samples according to a preset time period using an embedded wearable blood sampling device deployed on the pig's body;

[0149] A second acquisition module is configured to collect environmental state parameters in real time through a multi-sensor network deployed in the pig breeding space, including temperature distribution data, humidity fluctuation data, and air quality comprehensive index data, where the dynamic ecological variables of the breeding environment include the temperature distribution data, humidity fluctuation data, and air quality comprehensive index data;

[0150] The recording module is used to dynamically record the quantitative parameters of pig growth performance through the growth monitoring system, including daily average weight gain data, daily feed intake data and feed conversion efficiency data.

[0151] Optionally, the device comprises:

[0152] The sending module is used to send a notification signal to the user terminal or feeding management system through the communication interface when the physiological requirement of selenium output by the selenium requirement prediction model exceeds a preset critical threshold. The notification signal is used to instruct the user or feeding management system to adjust the proportion of selenium added in pig feed according to the physiological requirement of selenium.

[0153] The present disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for predicting the amount of selenium to be added using pig blood biochemical indicators.

[0154] The present disclosure also provides a device, comprising:

[0155] a memory having a computer program stored thereon;

[0156] The processor is used to execute the computer program in the memory to implement the steps of the method for predicting the amount of selenium to be added by using pig blood biochemical indicators.

[0157] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method for predicting the amount of selenium addition using pig blood biochemical indicators, and will not be elaborated here.

[0158] Figure 3 FIG. 7 is a block diagram of an electronic device 700 according to an exemplary embodiment. Figure 3 As shown, the electronic device 700 may include: a processor 701 , a memory 702 , and may further include one or more of a multimedia component 703 , an input / output (I / O) interface 704 , and a communication component 705 .

[0159] The processor 701 is used to control the overall operation of the electronic device 700 to complete all or part of the steps in the method for predicting the amount of selenium to be added using pig blood biochemical indicators. The memory 702 is used to store various types of data to support the operation of the electronic device 700. This data may include, for example, instructions for any application or method operating on the electronic device 700, as well as application-related data, such as contact information, sent and received messages, pictures, audio, video, etc. The memory 702 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 703 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 702 or transmitted via the communication component 705. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 704 provides an interface between the processor 701 and other interface modules. The above-mentioned other interface modules may be a keyboard, a mouse, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 705 is used for wired or wireless communication between the electronic device 700 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more thereof, so the corresponding communication component 705 may include: a Wi-Fi module, a Bluetooth module, an NFC module.

[0160] In an exemplary embodiment, the electronic device 700 can be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the above-mentioned method for predicting the amount of selenium addition using pig blood biochemical indicators.

[0161] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, the program instructions implement the steps of the aforementioned method for predicting the amount of selenium to be added using porcine blood biochemical indicators. For example, the computer-readable storage medium may be the aforementioned memory 702 including the program instructions. The program instructions may be executed by the processor 701 of the electronic device 700 to implement the aforementioned method for predicting the amount of selenium to be added using porcine blood biochemical indicators.

[0162] In another exemplary embodiment, a computer program product is also provided. The computer program product includes a computer program that can be executed by a processor. When the computer program is executed by the processor, the steps of the method for predicting the amount of selenium addition using pig blood biochemical indicators are implemented.

[0163] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all fall within the scope of protection of the present disclosure.

[0164] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction. In order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0165] In addition, the various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.

Claims

1. A method for predicting the amount of selenium added using pig blood biochemical indicators, characterized in that: include: Through biochemical testing of the collected pig blood samples, blood biomarker data related to the immune response mechanism were extracted; The blood biomarker data, quantitative parameters of pig growth performance, and dynamic ecological variables of the pig breeding environment are integrated into a multidimensional feature vector, which is input into a selenium requirement prediction model based on deep learning to obtain the pig's physiological requirement for selenium as output by the selenium requirement prediction model, wherein the selenium requirement prediction model uses the multidimensional feature vectors of historical samples as input data and the corresponding reference selenium requirement as a supervision label, and is constructed through multi-layer neural network architecture training; Calculating and adjusting the proportion of selenium added to pig feed based on the pig's physiological demand for selenium; The biochemical test of the collected pig blood samples is performed to extract blood biomarker data related to the immune response mechanism, specifically including: The collected pig blood samples were analyzed for enzyme activity and protein content to obtain data on the catalytic activity of glutathione peroxidase, the antioxidant activity of superoxide dismutase, the serum concentration of immunoglobulins, and the secretion level of cytokines; The blood biomarker data include the catalytic activity data of glutathione peroxidase, the antioxidant activity data of superoxide dismutase, the serum concentration data of immunoglobulins, and the cytokine secretion level data.

2. The method according to claim 1, characterized in that The method further includes the steps of constructing a selenium demand prediction model and optimizing parameters, specifically including: Constructing a structured training data set, wherein the training data set includes a sample multidimensional feature vector and a corresponding reference selenium requirement label, wherein the sample multidimensional feature vector includes blood biomarker data of historical samples, quantitative growth performance parameters, and dynamic ecological variables of the breeding environment; inputting the sample multidimensional feature vector into a prediction model to be trained based on a deep neural network to obtain a first predicted selenium requirement output by the prediction model to be trained; constructing a loss function and calculating the loss value based on the deviation between the first predicted selenium requirement and the reference selenium requirement; Utilizing the loss value, the weight parameters and bias parameters of the prediction model to be trained are iteratively updated through an adaptive algorithm combining stochastic gradient descent with momentum optimization to approach the global optimal solution.

3. The method according to claim 2, characterized in that Inputting the sample multidimensional feature vector into a prediction model to be trained based on a deep neural network to obtain a first predicted selenium requirement output by the prediction model to be trained specifically includes: The training data set is divided into a predetermined proportion, wherein 70% of the data is used as a training subset for model parameter learning, and 30% of the data is used as a test subset for model performance evaluation; Inputting the multidimensional feature vector of the samples in the training subset into the prediction model to be trained to obtain a first predicted selenium requirement output by the prediction model to be trained; The step of constructing a loss function and calculating a loss value based on a deviation between the first predicted selenium requirement and the reference selenium requirement specifically includes: Based on the numerical deviation between the first predicted selenium requirement and the reference selenium requirement in the training subset, a mean square error loss function is constructed and the corresponding loss value is calculated. The loss value is used to evaluate the model prediction error and guide the optimization of the weight parameters and bias parameters of the prediction model to be trained.

4. The method according to claim 3, characterized in that The method comprises: After the training iteration of the prediction model to be trained is completed, the multidimensional feature vector of the samples in the test subset is input into the trained prediction model to obtain a second predicted selenium requirement output by the trained prediction model; Obtaining a reference selenium requirement in the test subset as a performance evaluation benchmark; Calculating an absolute deviation or relative error between the second predicted selenium requirement and the reference selenium requirement; When the absolute deviation value or the relative error value is less than or equal to a preset first threshold, it is determined that the output result of the trained prediction model meets the prediction accuracy requirement; When the absolute deviation value or the relative error value is greater than a preset first threshold, it is determined that the output result of the trained prediction model does not reach the expected accuracy; The prediction accuracy of the prediction model obtained by training on the test subset is counted. When the prediction accuracy is greater than a preset second threshold, the model training process is terminated; when the prediction accuracy is less than or equal to the preset second threshold, the model training process is continued.

5. The method according to any one of claims 1 to 4, characterized in that The method further comprises: According to a preset time period, the embedded wearable blood sampling device deployed on the pig body is used to periodically perform automated collection of pig blood samples; A multi-sensor network deployed in the pig breeding space collects environmental state parameters in real time, including temperature distribution data, humidity fluctuation data, and air quality comprehensive index data. The dynamic ecological variables of the breeding environment include the temperature distribution data, humidity fluctuation data, and air quality comprehensive index data. The growth monitoring system dynamically records the quantitative parameters of pig growth performance, including daily average weight gain data, daily feed intake data and feed conversion efficiency data.

6. The method according to any one of claims 1 to 4, characterized in that The method comprises: When the physiological requirement of selenium output by the selenium requirement prediction model exceeds a preset critical threshold, a notification signal is sent to the user terminal or the feeding management system through the communication interface. The notification signal is used to instruct the user or the feeding management system to adjust the proportion of selenium added in the pig feed according to the physiological requirement of selenium.

7. A device for predicting the amount of selenium added using pig blood biochemical indicators, characterized in that: include: An extraction module is used to extract blood biomarker data related to the immune response mechanism by performing biochemical tests on the collected pig blood samples; an acquisition module for integrating the blood biomarker data, quantitative parameters of pig growth performance, and dynamic ecological variables of the pig breeding environment into a multidimensional feature vector, inputting the vector into a selenium requirement prediction model based on deep learning, and obtaining the pig's physiological requirement for selenium as output by the selenium requirement prediction model, wherein the selenium requirement prediction model uses the multidimensional feature vector of historical samples as input data and the corresponding reference selenium requirement as a supervision label, and is constructed through multi-layer neural network architecture training; A calculation module is used to calculate and adjust the proportion of selenium added to the pig feed according to the pig's physiological demand for selenium; The extraction module is used to: The collected pig blood samples were analyzed for enzyme activity and protein content to obtain data on the catalytic activity of glutathione peroxidase, the antioxidant activity of superoxide dismutase, the serum concentration of immunoglobulins, and the secretion level of cytokines; The blood biomarker data include the catalytic activity data of glutathione peroxidase, the antioxidant activity data of superoxide dismutase, the serum concentration data of immunoglobulins, and the cytokine secretion level data.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

9. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 6.

Images