Livestock breeding management system and method based on machine learning algorithm
By constructing a production value prediction model based on machine learning algorithms, the impact of air pollution on livestock farm output value was addressed, efficient management strategies were optimized, and the production efficiency and quality of animal husbandry were improved.
Patent Information
- Application Number
- CN202511154114.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-18
- Publication Date
- 2025-11-21
AI Technical Summary
Air pollution in livestock farms has a negative impact on animal health and the environment, and existing technologies are insufficient to effectively manage and optimize livestock output.
A production value prediction model based on machine learning algorithms is constructed. By collecting and processing air environment data from livestock farms, the feature extraction strategy is optimized using Lasso regression and linear loss functions of support vector machines to predict livestock production value and output the optimal management strategy.
It improves the accuracy and reliability of air environment forecasting of livestock output, and can optimize the management strategies of livestock farms, thereby improving production efficiency and quality.
Smart Images
Figure CN120996973A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of livestock breeding management technology, specifically relating to a livestock breeding management system and management method based on machine learning algorithms. Background Technology
[0002] With the rapid development of my country's economy, animal husbandry has become an important component of the agricultural economy, providing residents with a variety of livestock products such as meat, milk, and eggs. In recent years, due to the continuous improvement of the living standards of Chinese residents, the demand for livestock and poultry meat products has increased rapidly. This huge market environment has prompted the transformation of the animal husbandry industry from traditional extensive farming methods to modern animal husbandry models such as large-scale, intensive, and factory-style farming.
[0003] However, with the intensive development of animal husbandry, the negative impact of air pollution on livestock health and production efficiency has become increasingly significant. Livestock farms contain large amounts of harmful gases, such as ammonia, hydrogen sulfide, and methane. These gases not only threaten animal health but also negatively impact the living environment of surrounding residents. Simultaneously, odor pollution, temperature and humidity fluctuations, and microbial activity from farms are constantly affecting air quality and ecological balance. Therefore, this application proposes a livestock management system and method that comprehensively considers air environment data within livestock farms, which is essential. Summary of the Invention
[0004] In view of this, in order to solve the problems mentioned in the background art, the purpose of this invention is to provide a livestock breeding management system and management method based on machine learning algorithms.
[0005] To achieve the above objectives, the present invention provides the following technical solution:
[0006] A livestock farming management method based on machine learning algorithms includes:
[0007] S1. Construct and train a production value prediction model based on machine learning algorithms;
[0008] S2. Collect current air environment data for the target livestock farm;
[0009] S3. Obtain the target attribute characteristics of the target livestock farm;
[0010] S4. Using the target's current air environment data and the target's attribute features as inputs to the output value prediction model, identify a target feature extraction strategy based on the target attribute features, extract target features from the target's current air environment data based on the target feature extraction strategy, and predict the target output value of the target livestock farm based on the target features.
[0011] S5. Based on the predicted output value of the target, select and output the optimal management strategy that can be applied to the target livestock farm.
[0012] According to an embodiment of the present invention, in step S1, training the output prediction model based on a machine learning algorithm includes:
[0013] S11. Construct an original dataset, which includes historical air environment data and historical output value of livestock farms with different attribute characteristics;
[0014] S12. Train the output prediction model using the original dataset.
[0015] According to an embodiment of the present invention, in step S11, constructing the original dataset includes: collecting and preprocessing historical air environment data and historical output value of livestock farms, wherein the preprocessing includes cleaning, deduplication, and missing value filling.
[0016] According to an embodiment of the present invention, step S12, training the output prediction model using the original dataset, includes:
[0017] Filter the first historical air environment data and first historical output value of livestock farms with the first attribute feature from the original dataset;
[0018] Using the first historical air environment data as the input of the output value prediction model, the first feature is extracted from the first historical air environment data according to the first feature extraction strategy, and the first predicted output value of the livestock farm is predicted based on the first feature.
[0019] The first feature extraction strategy is optimized by comparing the first predicted output value with the first historical output value.
[0020] According to one embodiment of the present invention, the first feature extraction strategy includes a correlation analysis strategy and a transformation processing strategy.
[0021] According to one embodiment of the present invention, the correlation analysis is performed using the L2 regularization penalty term of Lasso regression:
[0022] In the formula, This represents the regression coefficient.
[0023] According to one embodiment of the present invention, the transformation processing strategy includes one or more of the following: standardization processing, normalization processing, logarithmic transformation processing, polynomial processing, principal component analysis processing, linear discriminant analysis processing, and nonlinear dimensionality reduction processing.
[0024] According to one embodiment of the present invention, when comparing the first predicted output with the first historical output: minimize the loss function optimized with respect to the first predicted output and the first historical output.
[0025] According to an embodiment of the present invention, the loss function is expressed as:
[0026] In the formula, Indicates the penalty coefficient. This represents the first historical output value corresponding to the i-th first historical air environment data point. This represents the first predicted output value corresponding to the i-th historical air quality data point. This represents the threshold parameter.
[0027] As a general inventive concept, this invention also provides a livestock farming management system based on machine learning algorithms, comprising:
[0028] Model building module; used to build output prediction models based on machine learning algorithms;
[0029] Model training module; used to train output prediction models based on machine learning algorithms;
[0030] Data acquisition module; collects current air environment data and target attribute characteristics of the target livestock farm;
[0031] The model execution module is used to obtain the target predicted output value of the target livestock farm based on the target's current air environment data, the target's attribute characteristics, and the output value prediction model.
[0032] The filtering and output module filters and outputs the optimal management strategy that can be applied to the target livestock farm based on the target predicted output value.
[0033] Specifically, when the output prediction model is run through the model running module:
[0034] Using the target's current air environment data and the target's attribute features as inputs to the output prediction model, a target feature extraction strategy is identified based on the target attribute features. Target features are extracted from the target's current air environment data based on the target feature extraction strategy, and the target predicted output value of the target livestock farm is predicted based on the target features.
[0035] Compared with the prior art, the present invention has the following advantages:
[0036] This invention constructs a production value prediction model that can predict livestock output value using air environment data. The model is trained using multi-source data from livestock farms with different attribute characteristics, thus more comprehensively reflecting the impact of air environment on livestock output value and improving the accuracy and reliability of the model's predictions. Furthermore, in the specific analysis and prediction process, the production value prediction model incorporates machine learning algorithms such as Lasso regression and linear loss functions of support vector machines. This allows the production value prediction model to continuously optimize and learn during analysis and prediction, achieving efficient processing of complex and changing data characteristics. This further improves the explanatory power and predictive accuracy of the correlation between air pollution and livestock output value, and also enables further optimization and control of livestock farm management strategies based on the predicted output value. Attached Figure Description
[0037] Figure 1 This is a flowchart of the livestock breeding management method based on machine learning algorithms of the present invention;
[0038] Figure 2 This is a heatmap illustrating the correlation between the first historical air environment data and the first historical output value in the embodiment.
[0039] Figure 3 This is a comparison chart of the first predicted output value and the first historical output value in the embodiment. Detailed Implementation
[0040] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The structures, proportions, sizes, etc., depicted in the accompanying drawings are merely for illustrative purposes and to aid those skilled in the art, and are not intended to limit the implementation conditions of the invention. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to size, without affecting the effects and objectives of the invention, should still fall within the scope of the technical content disclosed in this invention. Furthermore, terms such as "upper," "lower," "left," "right," and "middle" used in this specification are merely for clarity and not intended to limit the scope of implementation. Changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention. It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein.
[0041] Example
[0042] like Figure 1As shown in the figure, the livestock breeding management method based on machine learning algorithm provided in this embodiment includes the following steps:
[0043] S1. Construct and train a production value prediction model based on machine learning algorithms.
[0044] (1) Constructing an output value prediction model
[0045] A production value prediction model is constructed based on the Lasso regression linear model. This production value prediction model includes a feature extraction layer, an encoding layer, an input layer, a decoding layer, and an output layer.
[0046] Regarding the feature extraction layer: Feature extraction based on correlation analysis is performed by introducing an L2 regularization penalty term from Lasso regression. The L2 regularization penalty term is the regression coefficient... The square of and the regularization parameter The product of, where The strength of regularization is controlled. Specifically, the L2 regularization penalty term is expressed as: The L2 regularization penalty term mathematically allows for the regression coefficient... It is 0, but it is more inclined to make the regression coefficient 0. By infinitely approaching 0, the L2 regularization penalty term can help control the complexity of the model and reduce the risk of overfitting.
[0047] Regarding the encoding layer: the features extracted by the feature extraction layer are encoded into data packets adapted to the output prediction model;
[0048] Regarding the input layer: the encoded data packet is input to the decoding layer;
[0049] Regarding the decoding layer: the predicted output value is obtained by processing the data packet, and the predicted output value is decoded;
[0050] Regarding the output layer: outputs the predicted output value after decoding.
[0051] (2) Training the output prediction model
[0052] S11. Collect historical air environment data and historical output values of livestock farms with different attribute characteristics, and perform preprocessing such as cleaning, deduplication, and missing value filling on the historical air environment data and historical output values; the historical air environment data includes air pollution monitoring data, meteorological data, etc.
[0053] Regarding meteorological data (such as temperature, humidity, wind speed, etc.): it comes from the meteorological bureau's official website, professional meteorological databases, or real-time records from meteorological monitoring stations, etc.
[0054] Regarding air pollution monitoring data (including but not limited to the concentrations of various pollutants such as nitrogen dioxide, sulfur dioxide, and PM2.5): data can be obtained from publicly available data platforms released by environmental protection departments and local environmental monitoring centers. Alternatively, data can be collected independently using specialized air quality monitoring equipment.
[0055] Regarding historical output value: it was collected through statistical yearbooks of livestock farms with different attributes, industry reports issued by the agricultural and rural affairs departments, and data products from relevant economic research institutions.
[0056] The preprocessed original dataset can be divided into a training set and a validation set in a 7:3 ratio.
[0057] S12. Train the output prediction model using the original dataset (training set):
[0058] ① Filter the first historical air environment data and first historical output value of livestock farms with the first attribute feature from the original dataset;
[0059] ② Using the first historical air environment data as the input of the output value prediction model, extract the first feature from the first historical air environment data according to the first feature extraction strategy, and predict the first output value of the livestock farm based on the first feature.
[0060] In this step, the first feature extraction strategy includes a correlation analysis strategy and a transformation processing strategy. Based on this:
[0061] The feature extraction layer extracts a first feature from the first historical air environment data according to the correlation analysis strategy (using the L2 regularization penalty term of Lasso regression for correlation analysis, and then extracting a first feature from the first historical air environment data whose correlation with the first historical output value exceeds a threshold). Taking a livestock farm in Hohhot City in 2022 as an example, the correlation analysis between the historical air environment data and the historical output value of the livestock farm is performed according to the correlation analysis strategy, and the results are as follows: Figure 2 The correlation analysis heatmap shown clearly reveals the corresponding characteristics in historical air environment data where the correlation with the historical output value of the livestock farm exceeds a threshold. This threshold can be preset manually or automatically calculated using algorithms such as averaging.
[0062] The encoding layer converts the first feature into a data packet adapted to the output prediction model according to the conversion processing strategy.
[0063] The input layer inputs the encoded data packets to the decoding layer.
[0064] The decoding layer obtains a first predicted output value by processing the data packet and then decodes the first predicted output value.
[0065] The output layer outputs the first predicted output value after decoding.
[0066] ③ Optimize the first feature extraction strategy by comparing the first predicted output value with the first historical output value.
[0067] Regarding the comparison between the first projected output value and the first historical output value: Taking livestock farms in Hohhot City in 2022 as an example, please refer to the comparison chart between the projected output value and the historical output value. Figure 3 As shown.
[0068] Regarding the optimization of the first feature extraction strategy: In the regression problem of the Lasso linear regression model, the goal is to find a function that fits the sample points in the dataset as closely as possible while maintaining sufficient smoothness. Specifically, we hope to find a function... This minimizes, for a given training set, the following loss function optimized with respect to the first predicted output and the first historical output:
[0069] In the formula:
[0070] This represents the penalty coefficient; it controls the model's tolerance for error or the severity of the penalty. It indirectly affects the model's behavior during training by scaling the value of the loss function: a larger value results in a smaller value. The model penalizes errors more severely, tending to reduce training error. Smaller... The model has a higher tolerance for error and allows for larger errors;
[0071] This represents the first historical output value corresponding to the i-th first historical air environment data point;
[0072] This represents the first predicted output value corresponding to the i-th first historical air environment data;
[0073] This represents the threshold parameter; it is often referred to as "tolerance" or "boundary".
[0074] when At that time, the model's prediction for the i-th historical air quality data is considered "good enough" and will not incur loss, i.e. ;
[0075] when Losses will only occur at that time. .
[0076] In summary, during the specific training process in this embodiment:
[0077] Regarding the correlation analysis strategy in the first feature extraction strategy: air environment data is used as the independent variable, and livestock output value is used as the dependent variable, by optimizing and adjusting the regularization parameter. The magnitude of the correlation was used to explore different correlation patterns and identify air pollution factors that have a significant impact on livestock output (features extracted by the feature extraction layer).
[0078] Regarding the transformation processing strategy in the first feature extraction strategy:
[0079] The transformation processing strategy includes one or more of the following: standardization processing, normalization processing, logarithmic transformation processing, polynomial processing, principal component analysis processing, linear discriminant analysis processing, and nonlinear dimensionality reduction processing.
[0080] Standardization: The features are transformed into a distribution with a mean of 0 and a standard deviation of 1 to eliminate dimensional differences between different features.
[0081] Normalization: The features are scaled to the range of [0, 1] or [-1, 1] to eliminate the influence of units and numerical range.
[0082] Logarithmic transformation: For skewed distribution features, logarithmic transformation can be used to make it closer to a normal distribution, thereby improving the predictive performance of the model.
[0083] Multinomial processing: By introducing higher-order terms or cross terms of features, the nonlinear expressive power of the model is increased.
[0084] Principal component analysis: The original features are projected into a low-dimensional space through linear transformation, preserving the maximum variance information.
[0085] Linear discriminant analysis: In supervised learning scenarios, the optimal projection direction is found by maximizing the ratio of between-class variance to within-class variance.
[0086] Nonlinear dimensionality reduction: suitable for visualizing high-dimensional data, and can preserve the local structure of the data.
[0087] In summary, the impact of air quality on livestock output varies for livestock farms with different attributes (e.g., different geographical locations, different types of attributes). Therefore, model training can establish a correspondence between different attributes of livestock farms and different feature extraction strategies. This facilitates targeted feature extraction from air quality data based on the unique attributes of livestock farms. Consequently, the trained output prediction model can accurately select features from air quality data that are more than a threshold in relation to livestock output, ensuring the accuracy and reliability of livestock output prediction.
[0088] S13. Use the original dataset (validation set) to evaluate and validate the prediction accuracy of the output prediction model, and define the output prediction model with a prediction accuracy of more than 90% as a qualified model.
[0089] S2. Collect current air environment data for the target livestock farm;
[0090] The target's current air environment data includes current air pollution monitoring data and current meteorological data. Regarding current meteorological data (such as temperature, humidity, and wind speed): these are sourced from the meteorological bureau's official website, professional meteorological databases, or real-time records from meteorological monitoring stations. Regarding current air pollution monitoring data (including but not limited to the concentrations of various pollutants such as nitrogen dioxide, sulfur dioxide, and PM2.5): these are obtained from publicly available data platforms released by environmental protection departments and local environmental monitoring centers. Data can also be collected independently using professional air quality monitoring equipment. Furthermore, the collected target's current air environment data undergoes preprocessing such as cleaning, deduplication, and missing value filling.
[0091] S3. Obtain the target attribute characteristics of the target livestock farm.
[0092] S4. Using the target's current air environment data and the target's attribute features as inputs to the output value prediction model, identify a target feature extraction strategy based on the target attribute features, extract target features from the target's current air environment data based on the target feature extraction strategy, and predict the target output value of the target livestock farm based on the target features.
[0093] In this step: the target feature extraction strategy includes a target correlation analysis strategy and a target transformation processing strategy. The feature extraction layer extracts target features from the current air environment data of the target according to the target correlation analysis strategy. The encoding layer converts the target features into a data packet adapted to the output prediction model according to the target transformation processing strategy. The input layer inputs the encoded data packet to the decoding layer. The decoding layer processes the data packet to obtain the target predicted output value and decodes the target predicted output value. The output layer outputs the decoded target predicted output value.
[0094] S5. Based on the predicted target output value, select and output the optimal management strategy applicable to the target livestock farm. The optimal management strategy includes a visual representation of the predicted target output value and management adjustment suggestions for livestock farming. For example, during periods of good air quality, appropriately increase livestock density to improve production efficiency; during periods of severe air pollution, take measures to reduce production to minimize losses.
[0095] The present invention also discloses an exemplary embodiment of a livestock farming management system based on machine learning algorithms, the system comprising:
[0096] Model building module; used to build output prediction models based on machine learning algorithms;
[0097] Model training module; used to train output prediction models based on machine learning algorithms;
[0098] Data acquisition module; collects current air environment data and target attribute characteristics of the target livestock farm;
[0099] The model execution module is used to obtain the target predicted output value of the target livestock farm based on the target's current air environment data, the target's attribute characteristics, and the output value prediction model.
[0100] The filtering and output module filters and outputs the optimal management strategy that can be applied to the target livestock farm based on the target predicted output value.
[0101] Specifically, when the output prediction model is run through the model running module:
[0102] Using the target's current air environment data and the target's attribute features as inputs to the output prediction model, a target feature extraction strategy is identified based on the target attribute features. Target features are extracted from the target's current air environment data based on the target feature extraction strategy, and the target predicted output value of the target livestock farm is predicted based on the target features.
[0103] Regarding the system described above, the specific methods by which each module performs its operations have been described in detail in the embodiments of the method, and will not be elaborated upon here.
[0104] In summary, the livestock farming management system and method based on machine learning algorithms provided in this embodiment are mainly built upon key technologies such as multi-source data fusion, machine learning model application, real-time data analysis, dynamic prediction, and strategy optimization. By integrating these technologies, it is possible to effectively achieve comprehensive analysis and accurate prediction of the impact of air pollution on livestock output, providing scientific basis and decision support for livestock farms, promoting intelligent and automated management of animal husbandry, and improving agricultural production efficiency and quality.
[0105] In another exemplary embodiment, an electronic device is also provided, the electronic device including a memory and a processor, and a program stored in the memory, which, when executed by the processor, implements one or more of the aforementioned steps of livestock farming management based on machine learning algorithms.
[0106] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the above-described livestock farming management method based on machine learning algorithms. For example, the computer-readable storage medium may be a first memory including program instructions, which may be executed by a first processor of an electronic device to complete the above-described livestock farming management method based on machine learning algorithms.
[0107] In another exemplary embodiment, a computer program product is also provided, comprising a computer program executable by a programmable device, the computer program having a code portion for performing the aforementioned livestock farming management method based on machine learning algorithms when executed by the programmable device. In some embodiments, part or all of the computer program may be loaded into and / or installed onto a device via ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of the aforementioned method may be performed. Alternatively, in other embodiments, the CPU may be configured by any other suitable means (e.g., by means of firmware) to perform one or more steps of the aforementioned livestock farming management method based on machine learning algorithms.
[0108] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.
Claims
1. A livestock farming management method based on machine learning algorithms, characterized in that, include: S1. Construct and train a production value prediction model based on machine learning algorithms; S2 acquires the target current air environment data for the target livestock farm; S3. Obtain the target attribute characteristics of the target livestock farm; S4. Using the target's current air environment data and the target's attribute features as inputs to the output value prediction model, identify a target feature extraction strategy based on the target attribute features, extract target features from the target's current air environment data based on the target feature extraction strategy, and predict the target output value of the target livestock farm based on the target features. S5. Based on the predicted output value of the target, select and output the optimal management strategy that can be applied to the target livestock farm.
2. The livestock breeding management method based on machine learning algorithm according to claim 1, characterized in that, In step S1, training the output prediction model based on machine learning algorithms includes: S11. Construct an original dataset, which includes historical air environment data and historical output value of livestock farms with different attribute characteristics; S12. Train the output prediction model using the original dataset.
3. The livestock breeding management method based on machine learning algorithm according to claim 2, characterized in that, In step S11, constructing the original dataset includes: Historical air environment data and historical output value of livestock farms are collected and preprocessed, including cleaning, deduplication, and missing value filling.
4. The livestock breeding management method based on machine learning algorithm according to claim 2, characterized in that, In step S12, training the output prediction model using the original dataset includes: Filter the first historical air environment data and first historical output value of livestock farms with the first attribute feature from the original dataset; Using the first historical air environment data as the input of the output value prediction model, the first feature is extracted from the first historical air environment data according to the first feature extraction strategy, and the first predicted output value of the livestock farm is predicted based on the first feature. The first feature extraction strategy is optimized by comparing the first predicted output value with the first historical output value.
5. The livestock breeding management method based on machine learning algorithm according to claim 4, characterized in that: The first feature extraction strategy includes a correlation analysis strategy and a transformation processing strategy.
6. The livestock breeding management method based on machine learning algorithm according to claim 5, characterized in that: The correlation analysis was performed using the L2 regularization penalty term of Lasso regression. In the formula, This represents the regression coefficient.
7. The livestock breeding management method based on machine learning algorithm according to claim 5, characterized in that: The transformation processing strategy includes one or more of the following: standardization processing, normalization processing, logarithmic transformation processing, polynomial processing, principal component analysis processing, linear discriminant analysis processing, and nonlinear dimensionality reduction processing.
8. The livestock breeding management method based on machine learning algorithm according to claim 4, characterized in that, When comparing the first predicted output with the first historical output: minimize the loss function optimized with respect to the first predicted output and the first historical output.
9. A livestock breeding management method based on machine learning algorithms according to claim 8, characterized in that, The loss function is expressed as: In the formula, Indicates the penalty coefficient. This represents the first historical output value corresponding to the i-th first historical air environment data point. This represents the first predicted output value corresponding to the i-th historical air quality data point. This represents the threshold parameter.
10. A livestock breeding management system based on machine learning algorithms, characterized in that, include: Model building module; Used to build output prediction models based on machine learning algorithms; Model training module; Used to train output prediction models based on machine learning algorithms; Data acquisition module; collects current air environment data and target attribute characteristics of the target livestock farm; The model execution module is used to obtain the target predicted output value of the target livestock farm based on the target's current air environment data, the target's attribute characteristics, and the output value prediction model. The filtering and output module filters and outputs the optimal management strategy that can be applied to the target livestock farm based on the target predicted output value. Specifically, when the output prediction model is run through the model running module: Using the target's current air environment data and the target's attribute features as inputs to the output prediction model, a target feature extraction strategy is identified based on the target attribute features. Target features are extracted from the target's current air environment data based on the target feature extraction strategy, and the target predicted output value of the target livestock farm is predicted based on the target features.