Pig farm disease risk prediction method, system and platform based on small sample scene

Abstract

The application discloses a pig farm disease risk prediction method, system and platform based on a small sample scene, and the method is used for acquiring original record data corresponding to pig farm diseases in real time, and determining and generating pig farm disease record data corresponding to a small sample scene; a corresponding knowledge fusion model is constructed according to the pig farm disease original record data in the small sample scene, and early warning data corresponding to the pig farm diseases is generated in real time in combination with the knowledge fusion model; a corresponding risk point identification model is constructed, and pig farm disease risk prediction data corresponding to the early warning data is generated in real time according to the risk point identification model, which can assist front-line pig farm managers to achieve early discovery, accurate positioning, timely prevention and control and continuous tracking. Through risk early warning, intervention measures such as immunization, disinfection and purification are carried out on potential pig farms in time, so that the abortion rate and the stillbirth rate of the pig farms are reduced, and economic losses are avoided.

Description

Technical Field

[0001] This invention belongs to the field of pig farm disease risk prediction technology, specifically involving a method, system, and platform for predicting pig farm disease risk based on small sample scenarios. Background Technology

[0002] Currently, some disease information systems for pig farms on the market are mainly designed for the information management of a single disease, and are generally used for disease information management and improving the detection efficiency of related diseases.

[0003] However, existing pig farm disease information systems on the market still have many shortcomings, as follows:

[0004] Low coverage: The disease information systems for pig farms on the market generally only target a single disease, and have limited coverage of common major diseases in pig farms.

[0005] Lack of early warning capabilities: Pig farm disease management systems on the market only support disease detection functions or quantify the risk of sick pig farms, but fail to predict the risk of diseases and cannot achieve early prevention and control.

[0006] Lack of risk point identification capability: Some pig farm information systems only output the overall risk and fail to provide the ability to identify and quantify relevant risk points, making it difficult to achieve precise management.

[0007] High requirements for labeled data: Many intelligent systems on the market have high requirements for the amount of standard data, requiring at least hundreds of thousands or even millions of labeled data, which increases the difficulty of system construction and promotion.

[0008] Therefore, in order to address the above-mentioned technical problems and shortcomings, there is an urgent need to design and develop a method, system, and platform for predicting disease risks in pig farms based on small sample scenarios. Summary of the Invention

[0009] To overcome the shortcomings and difficulties of the existing technologies, the present invention aims to provide a method, system, platform, and storage medium for predicting pig farm disease risks based on small sample scenarios. This method can assist frontline pig farm managers in early detection, accurate location, timely prevention and control, and continuous tracking. Furthermore, through risk warnings, it enables timely intervention measures such as immunization, disinfection, and purification for potentially diseased pig farms, thereby reducing abortion rates, mortality rates during farrowing, and ultimately avoiding economic losses.

[0010] The primary objective of this invention is to provide a method for predicting disease risk in pig farms based on small sample scenarios;

[0011] The second objective of this invention is to provide a disease risk prediction system for pig farms based on small sample scenarios;

[0012] The third objective of this invention is to provide a disease risk prediction platform for pig farms based on small sample scenarios;

[0013] The fourth objective of this invention is to provide a computer-readable storage medium;

[0014] The first objective of this invention is achieved as follows: the method comprises the following steps:

[0015] Real-time acquisition of raw record data corresponding to pig farm diseases, and determination and generation of pig farm disease record data corresponding to small sample scenarios;

[0016] Based on the original disease records of pig farms in small sample scenarios, a corresponding knowledge fusion model is constructed, and in combination with the knowledge fusion model, early warning data corresponding to pig farm diseases is generated in real time.

[0017] Construct a corresponding risk point identification model, and based on the risk point identification model, generate pig farm disease risk prediction data corresponding to the early warning data in real time.

[0018] Furthermore, the real-time acquisition of raw record data corresponding to pig farm diseases, and the determination of whether the raw record data is raw record data of pig farm diseases in a small sample scenario, further includes:

[0019] Deep learning is used to process disease data in small sample scenarios, and unlabeled sample data corresponding to the original disease records of pig farms in small sample scenarios are labeled and processed.

[0020] Based on the disease data from the small sample scenario, an epidemic identification model is constructed in real time.

[0021] Furthermore, the deep learning-corresponding small-sample scenario disease data, and the annotation processing of unlabeled sample data corresponding to the original disease record data of pig farms in the small-sample scenario, also includes:

[0022] Preprocess disease characteristic data in small sample scenarios and perform pseudo-labeling on unlabeled samples from pig farms.

[0023] Furthermore, the step of constructing a corresponding knowledge fusion model based on the original disease records of pig farms in a small sample scenario, and generating real-time early warning data corresponding to pig farm diseases in conjunction with the knowledge fusion model, also includes:

[0024] By combining multi-strategy learning strategies, multiple disease prediction models are trained and constructed; the training process includes multi-task knowledge learning and knowledge fusion.

[0025] Furthermore, the step of constructing a corresponding risk point identification model and generating pig farm disease risk prediction data corresponding to the early warning data in real time based on the risk point identification model also includes:

[0026] Generate and acquire early warning risk data;

[0027] Train and construct a proxy explanation model, and based on the proxy explanation model, perform real-time attribution processing on the observation factors corresponding to the early warning risk data.

[0028] Furthermore, after constructing the corresponding risk point identification model and generating pig farm disease risk prediction data corresponding to the early warning data in real time based on the risk point identification model, the method further includes:

[0029] The system detects, investigates, and processes corresponding pig farm disease risk prediction data, and transmits the corresponding pig farm disease risk prediction data in real time.

[0030] The disease risk prediction data for the pig farm is visualized.

[0031] The second objective of the present invention is achieved as follows: the system comprises:

[0032] The acquisition and judgment unit is used to acquire raw record data corresponding to pig farm diseases in real time, and to judge and generate pig farm disease record data corresponding to small sample scenarios.

[0033] The first construction and generation unit is used to construct a corresponding knowledge fusion model based on the original record data of pig farm diseases in a small sample scenario, and to generate early warning data corresponding to pig farm diseases in real time by combining the knowledge fusion model.

[0034] The second construction and generation unit is used to construct a corresponding risk point identification model and, based on the risk point identification model, generate pig farm disease risk prediction data corresponding to the early warning data in real time.

[0035] Furthermore, the acquisition and determination unit further includes:

[0036] The annotation processing module is used for deep learning of disease data in corresponding small sample scenarios, and for annotating and processing unlabeled sample data corresponding to the original disease record data of pig farms in small sample scenarios.

[0037] The first construction module is used to build an epidemic identification model in real time based on the small sample disease data.

[0038] And / or, the annotation processing module further includes:

[0039] The preprocessing module is used to preprocess disease characteristic data in small sample scenarios and to perform pseudo-labeling on unlabeled samples from pig farms.

[0040] And / or, the first construction generation unit further includes:

[0041] The training module is used to train and build multiple disease prediction models by combining multi-strategy learning strategies; the training process includes: multi-task knowledge learning and knowledge fusion.

[0042] And / or, the second construction generation unit further includes:

[0043] The first generation module is used to generate and acquire early warning risk data;

[0044] The second construction module is used to train and build a proxy explanation model, and based on the proxy explanation model, to perform real-time attribution processing on the observation factors corresponding to the early warning risk data;

[0045] And / or, the system further includes:

[0046] The detection and screening module is used to detect, screen and process the corresponding pig farm disease risk prediction data, and transmit the corresponding pig farm disease risk prediction data in real time.

[0047] The visualization module is used to visually display the disease risk prediction data of the pig farm.

[0048] The third objective of this invention is achieved as follows: it includes a processor, a memory, and a control program for a pig farm disease risk prediction platform based on a small sample scenario; wherein the control program for the pig farm disease risk prediction platform based on a small sample scenario is executed on the processor, the control program for the pig farm disease risk prediction platform based on a small sample scenario is stored in the memory, and the control program for the pig farm disease risk prediction platform based on a small sample scenario implements the method for predicting pig farm disease risk based on a small sample scenario.

[0049] The fourth objective of this invention is achieved as follows: the computer-readable storage medium stores a control program for a pig farm disease risk prediction platform based on a small sample scenario, and the control program for the pig farm disease risk prediction platform based on a small sample scenario implements the pig farm disease risk prediction method based on a small sample scenario.

[0050] This invention acquires raw record data corresponding to swine farm diseases in real time, and identifies and generates swine farm disease record data corresponding to small sample scenarios. Based on the raw record data of swine farm diseases in the small sample scenarios, a corresponding knowledge fusion model is constructed, and early warning data corresponding to swine farm diseases is generated in real time using the knowledge fusion model. A corresponding risk point identification model is constructed, and swine farm disease risk prediction data corresponding to the early warning data is generated in real time based on the risk point identification model. The invention also includes a system, platform, and storage medium corresponding to the method. This invention can assist frontline swine farm managers in early detection, accurate location, timely prevention and control, and continuous tracking. Through risk early warning, timely intervention measures such as immunization, disinfection, and purification can be implemented in potentially diseased swine farms, thereby reducing abortion rates and mortality rates during farrowing, and thus avoiding economic losses. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a schematic diagram of the process of a pig farm disease risk prediction method based on a small sample scenario according to the present invention.

[0053] Figure 2 This is a schematic diagram of the overall architecture of an embodiment of a pig farm disease risk prediction method based on a small sample scenario according to the present invention;

[0054] Figure 3 This is a schematic diagram illustrating the application process of an embodiment of the pig farm disease risk prediction method based on a small sample scenario according to the present invention.

[0055] Figure 4 This is a schematic diagram of the system architecture of an embodiment of a pig farm disease risk prediction method based on a small sample scenario according to the present invention;

[0056] Figure 5 This is a schematic diagram of the epidemic identification model architecture of an embodiment of the disease risk prediction method for pig farms based on small sample scenarios according to the present invention;

[0057] Figure 6 This is a schematic diagram of the disease risk prediction model architecture of an embodiment of the pig farm disease risk prediction method based on a small sample scenario according to the present invention.

[0058] Figure 7 This is a schematic diagram of the attribution system architecture of an embodiment of a pig farm disease risk prediction method based on a small sample scenario according to the present invention;

[0059] Figure 8 This is a schematic diagram of the architecture of a pig farm disease risk prediction system based on a small sample scenario according to the present invention;

[0060] Figure 9 This is a schematic diagram of the architecture of a pig farm disease risk prediction platform based on a small sample scenario according to the present invention.

[0061] Figure 10 This is a schematic diagram of a computer-readable storage medium architecture in one embodiment of the present invention;

[0062] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0063] To facilitate a clearer understanding of the objectives, technical solutions, and advantages of this invention, the invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Those skilled in the art can easily understand other advantages and effects of this invention from the content disclosed in this specification.

[0064] This invention can also be implemented or applied through other different specific examples, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of this invention.

[0065] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0066] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Secondly, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0067] Preferably, the disease risk prediction method for pig farms based on small sample scenarios of the present invention is applied in one or more terminals or servers. The terminal is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions. Its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0068] The terminal can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal can interact with the customer via a keyboard, mouse, remote control, touchpad, or voice control device.

[0069] This invention provides a method, system, platform, and storage medium for predicting disease risks in pig farms based on small sample scenarios.

[0070] like Figure 1 The diagram shown is a flowchart of a pig farm disease risk prediction method based on a small sample scenario provided by an embodiment of the present invention.

[0071] In this embodiment, the method for predicting pig farm disease risk based on small sample scenarios can be applied to terminals or fixed terminals with display functions. The terminals are not limited to personal computers, smartphones, tablets, desktop computers or all-in-one computers with cameras, etc.

[0072] The pig farm disease risk prediction method based on small sample scenarios can also be applied to a hardware environment consisting of a terminal and a server connected to the terminal via a network. The network includes, but is not limited to, a wide area network (WAN), a metropolitan area network (MAN), or a local area network (LAN). The pig farm disease risk prediction method based on small sample scenarios in this embodiment of the invention can be executed by the server, by the terminal, or by both the server and the terminal.

[0073] For example, for terminals requiring small-sample scenario-based pig farm disease risk prediction, the small-sample scenario-based pig farm disease risk prediction function provided by the method of this invention can be directly integrated into the terminal, or a client for implementing the method of this invention can be installed. Alternatively, the method provided by this invention can also run on servers or other devices in the form of a Software Development Kit (SDK), providing an interface for the small-sample scenario-based pig farm disease risk prediction function. Terminals or other devices can then implement the small-sample scenario-based pig farm disease risk prediction function through the provided interface. The invention will be further described below with reference to the accompanying drawings.

[0074] like Figure 1-10 As shown, this invention provides a method for predicting disease risk in pig farms based on small sample scenarios. The method includes the following steps:

[0075] S1. Real-time acquisition of raw record data corresponding to pig farm diseases, and determination and generation of pig farm disease record data corresponding to small sample scenarios;

[0076] S2. Based on the original records of pig farm diseases in a small sample scenario, construct a corresponding knowledge fusion model, and combine the knowledge fusion model to generate early warning data corresponding to pig farm diseases in real time.

[0077] S3. Construct a corresponding risk point identification model, and based on the risk point identification model, generate pig farm disease risk prediction data corresponding to the early warning data in real time.

[0078] The real-time acquisition of raw record data corresponding to pig farm diseases, and the determination of whether the raw record data is raw record data of pig farm diseases in a small sample scenario, further includes:

[0079] S11. Deep learning corresponds to small sample disease data, and labeling and processing unlabeled sample data corresponding to the original disease record data of pig farms in small sample scenarios;

[0080] S12. Based on the small sample disease data, construct an epidemic identification model in real time.

[0081] The deep learning-corresponding small-sample scenario disease data, and the annotation and processing of unlabeled sample data corresponding to the original disease record data of pig farms in the small-sample scenario, also include:

[0082] S111. Preprocess disease characteristic data in small sample scenarios and perform pseudo-labeling on unlabeled samples from pig farms.

[0083] The step of constructing a corresponding knowledge fusion model based on the original disease records of pig farms in a small sample scenario, and generating real-time early warning data corresponding to pig farm diseases in combination with the knowledge fusion model, also includes:

[0084] S21. Combine multi-strategy learning strategies to train and build multiple disease prediction models; the training process includes multi-task knowledge learning and knowledge fusion.

[0085] The step of constructing a corresponding risk point identification model and generating pig farm disease risk prediction data corresponding to the early warning data in real time based on the risk point identification model also includes:

[0086] S31. Generate and acquire early warning risk data;

[0087] S32. Train and construct a proxy explanation model, and perform real-time attribution processing on the observation factors corresponding to the early warning risk data based on the proxy explanation model.

[0088] After constructing the corresponding risk point identification model and generating pig farm disease risk prediction data corresponding to the early warning data in real time based on the risk point identification model, the method further includes:

[0089] S33. Detect, investigate, and process the corresponding pig farm disease risk prediction data, and transmit the corresponding pig farm disease risk prediction data in real time.

[0090] S34. Visualize and display the disease risk prediction data of the pig farm.

[0091] Specifically, in this embodiment of the invention, the overall architecture is based on the existing pig farm information system and big data platform, and a disease early warning algorithm system is established. The overall architecture is as follows: Figure 2 As shown, the modules are described below:

[0092]

[0093] In this solution, specific applications include: Figure 3 As shown, the process includes model training, risk prediction, risk report generation and push, and disease detection and screening. The system performs model training updates and risk predictions weekly.

[0094] Through precise early warning push notifications, business units can achieve efficient screening. This proactive early warning approach reduces the pressure on frontline screening teams and effectively lowers prevention and control costs. Furthermore, the proactive push notification method allows the system to continuously and rapidly accumulate labeled data, helping it maintain accuracy and timeliness.

[0095] In the system architecture and process, the disease early warning algorithm system is the core of this project, consisting of three main modules: disease identification module, disease early warning module, and risk identification module. The overall architecture is as follows: Figure 4 As shown.

[0096] Data Notes: All data refers to processed sample data, which will not be further specified below.

[0097]

[0098] The goal of disease early warning algorithm systems is to achieve predictable and interpretable results under limited conditions. Therefore, the algorithm system overcomes the following problems:

[0099] 1. Scarce labeled data (small sample size): The labeled data for relevant diseases only reaches the level of thousands, so it is not easy to train a reliable prediction model simply and directly;

[0100] 2. Poor model interpretability: High-performance algorithm models generally cannot be interpreted.

[0101] Key technologies:

[0102] Multi-stage learning: Under the same conditions, epidemic identification is simpler than disease prediction. The system first establishes a reliable epidemic identification model, and then provides sufficient and reliable pseudo-labeled data for downstream prediction tasks.

[0103] Multi-strategy few-shot learning: The system integrates multiple learning strategies, enabling the algorithm to output high-accuracy recognition and prediction models even under limited sample conditions. Specific features are as follows:

[0104] a) Data augmentation: Two data augmentation methods were used: heuristic strategies based on disease characteristics and Mixup, to add pseudo-labeled data for model training.

[0105] b) Efficient reuse of labeled data: Based on the concept of transfer learning, various strategies such as joint multi-task learning, phased transfer learning, pairwise ranking, and imbalanced classification learning (Focal Loss) are adopted to improve the efficiency of labeled data utilization.

[0106] c) Make full use of a large amount of unlabeled data: Use the UDA (Unsupervised Data Augmentation) semi-supervised sample synthesis strategy to make unlabeled data available for model training.

[0107] d) Regularization techniques: Label smoothing regularization avoids overfitting under small sample conditions, making it possible to train complex models.

[0108] e) Robust Model Architectures: Two robust model architectures ensure effective epidemic identification and disease prediction. Epidemic Identification: DenseDeepNFM deep neural network framework.

[0109] Disease Prediction: Multi-GBDT+LR Model Framework

[0110] Based on model proxy interpretability technology: The TreeSHAP (Tree SHapley Additive exPlanation) model-independent interpretability technology is adopted. By performing reverse proxy modeling on the disease prediction model, the early warning model is made white-box to the maximum extent, and the risk point attribution is realized.

[0111] In the epidemic identification module, based on a small amount of labeled data (disease reporting, disease detection), epidemic identification is performed on a large number of pig farms with no historical disease records to determine whether a porcine reproductive and respiratory syndrome (PRRS) or viral diarrhea outbreak has occurred in the corresponding time period (weekly granularity).

[0112] Among them, the training algorithm is:

[0113] A. Heuristic Data Augmentation: Based on the characteristics of porcine reproductive and respiratory syndrome (PRRS) disease, pseudo-labels are applied to unlabeled samples from pig farms during the data preprocessing stage. This strategy is only applicable to PRRS.

[0114] False negative samples (negative samples): A false negative sample is defined as a pig farm sample that meets all of the following conditions;

[0115] This week's maternity mortality rate is below the 25th percentile.

[0116] The number of miscarriages this week is below the 25th percentile.

[0117] The ratio of abortions to sows in stock is less than 0.3%.

[0118] False positive sample (positive sample): A false positive sample is defined as a pig farm sample that meets all of the following conditions;

[0119] Antigen positivity rate > 5%;

[0120] A sow had an abortion this week.

[0121] B. Joint Multi-Task Learning: The training process for the deep model for epidemic identification is as follows:

[0122]

[0123] The epidemic identification model architecture uses a DenseDeepNFM deep network architecture. This deep network consists of three main modules: wide, deep, and cross, with an overall depth of 6 layers. The specific architecture is shown in the attached image. Figure 5 As shown, the parameters are shown in the table below:

[0124] Module parameter Feature layer Input 45-dimensional features to represent the current situation of the pig farm, such as the number of pigs in stock and the number of abortions. Wide Part The width module, in the form of a linear model, enables strong memory capabilities. Embedding Layer The input feature projection is a 16-dimensional feature representation, indicating the learning task, which is used for subsequent feature crossing. Bi-interaction <![CDATA[Feature embeddings are pairwise cross-represented as v pooling = ∑ i ∑ j (v i ⊙ v j )]]> MLP A multi-layer network is used to learn high-order feature correlations, employing the ReLU activation function, with an output dimension of 16->32->32. Logistics layer The output of the MLP is further projected and integrated with the Wide Part to produce the Loogistics prediction.

[0125] For the disease risk prediction module, with limited labeled data, an existing epidemic identification model is used to label unlabeled data. Multiple disease prediction models are trained using a multi-strategy learning approach. This module outputs a total of 6 risk prediction models: 3 for porcine reproductive and respiratory syndrome (PRRS) and 3 for viral diarrhea. Each model is used for early warning of the corresponding disease N weeks in advance.

[0126] Among them, the training algorithm has a strong correlation between the two tasks of predicting sow abortion and farrowing mortality and culling, since PRRS and diarrhea mainly affect two key indicators: sow abortion and farrowing mortality and culling. Therefore, there is transferable knowledge.

[0127] Given the characteristics of the aforementioned diseases, the system is trained using a multi-task approach, learning the corresponding knowledge in each task separately, and finally integrating the knowledge through a fusion model to achieve early warning for the corresponding diseases.

[0128] The training process is divided into two parts: multi-task knowledge learning and knowledge fusion. The specific process is as follows:

[0129]

[0130] Regarding the model structure, the model consists of two parts: multi-task learning and knowledge fusion. Multi-task learning covers three types of tasks and trains a total of 150 decision tree models. The fusion model uses the leaves of each tree as input and employs the Logistic Regression model for multi-task knowledge fusion, as shown in the attached diagram. Figure 6 As shown,

[0131] For the risk point identification submodule, after completing the disease prediction model, it is necessary to further attribute the risk points of the early warning risks to facilitate the business units to formulate targeted governance measures.

[0132] To achieve risk point attribution, the system utilizes TreeSHAP (Tree SHapley Additive exPlanation) model-agnostic interpretability technology to construct surrogate interpretive models for disease prediction, enabling risk attribution for each observed factor. This module outputs a total of 6 risk attribution models, with 3 each for porcine reproductive and respiratory syndrome (PRRS) and viral diarrhea.

[0133]

[0134] Because the risk point attribution model involves more than 360 risk points, it is not easy for business units to use directly.

[0135] To further simplify the use by business units, the attribution system categorizes risk points into 7 major themes, as detailed in the appendix. Figure 7As shown.

[0136] In specific embodiments, such as Figure 4 As shown, firstly, labeled data (such as inventory data, abortion data, etc.) and unlabeled data (such as inventory data, abortion data, etc.) corresponding to each week are collected and obtained respectively. Among them, labeled data is data with actual disease reporting records; unlabeled data is data without disease reporting records. Then, combined with pseudo-label data (data inferred from the actual situation of pig farms), a multi-task learning process is used to construct an epidemic identification model, and combined with unlabeled data, disease data is identified to obtain pig farm disease record data corresponding to small sample scenarios. In other words, the goal of epidemic identification is to train an "epidemic identification model" to achieve disease identification on a large number of pig farms (weeks) of "unlabeled data". Our solution for this step can achieve a usable "identification" accuracy in "small sample scenarios" (very few labeled data, thousands of records). The specific solution is: a "joint multi-task learning" algorithm based on a "deep learning model" can utilize a small amount of labeled data while combining a large amount of "unlabeled data" for multi-task joint learning. Once the "disease identification model" is obtained, the incidence rate can be determined using all "unlabeled data" (e.g., based on the model, the incidence rate is estimated to be 90% (pseudo-label); unknown (actual)).

[0137] Then, based on the identification data obtained above, and combined with a multi-strategy learning strategy, multiple disease prediction models are trained and constructed. These models predict the incidence data (i.e., labeled data) for future weeks. In other words, the goal of disease early warning is to train an "early warning model" that can estimate the probability of disease occurrence in "future weeks" based on the characteristics (indicators) of the pig farm in the current week. At this stage, there are already about 100,000 labeled and pseudo-labeled data, but this is only mid-level data and cannot directly train a reliable "early warning model." Therefore, a "disease risk prediction model" based on knowledge fusion is achieved through multi-task transfer learning. The specific solution is as follows: improve data efficiency through phased transfer learning; Phase 1: multi-task learning: pre-train three related prediction tasks: disease early warning, abortion task, and mortality task; Phase 2: knowledge fusion: perform "transfer learning" on the three prediction sub-models output from the previous stage to achieve "knowledge fusion" and obtain the final "risk prediction model" (for example, based on the data estimated by the above model, the incidence prediction for future weeks can be made).

[0138] Finally, by constructing a corresponding risk point identification model and training and constructing a surrogate explanation model, and by performing real-time attribution processing on the observation factors corresponding to the early warning risk data based on the surrogate explanation model, the pig farm disease risk prediction data corresponding to the early warning data is obtained.

[0139] To achieve the above objectives, the present invention also provides a disease risk prediction system for pig farms based on small sample scenarios, such as... Figure 8 As shown, the system specifically includes:

[0140] The acquisition and judgment unit is used to acquire raw record data corresponding to pig farm diseases in real time, and to judge and generate pig farm disease record data corresponding to small sample scenarios.

[0141] The first construction and generation unit is used to construct a corresponding knowledge fusion model based on the original record data of pig farm diseases in a small sample scenario, and to generate early warning data corresponding to pig farm diseases in real time by combining the knowledge fusion model.

[0142] The second construction and generation unit is used to construct a corresponding risk point identification model and, based on the risk point identification model, generate pig farm disease risk prediction data corresponding to the early warning data in real time.

[0143] The acquisition and determination unit further includes:

[0144] The annotation processing module is used for deep learning of disease data in corresponding small sample scenarios, and for annotating and processing unlabeled sample data corresponding to the original disease record data of pig farms in small sample scenarios.

[0145] The first construction module is used to build an epidemic identification model in real time based on the small sample disease data.

[0146] And / or, the annotation processing module further includes:

[0147] The preprocessing module is used to preprocess disease characteristic data in small sample scenarios and to perform pseudo-labeling on unlabeled samples from pig farms.

[0148] And / or, the first construction generation unit further includes:

[0149] The training module is used to train and build multiple disease prediction models by combining multi-strategy learning strategies; the training process includes: multi-task knowledge learning and knowledge fusion.

[0150] And / or, the second construction generation unit further includes:

[0151] The first generation module is used to generate and acquire early warning risk data;

[0152] The second construction module is used to train and build a proxy explanation model, and based on the proxy explanation model, to perform real-time attribution processing on the observation factors corresponding to the early warning risk data;

[0153] And / or, the system further includes:

[0154] The detection and screening module is used to detect, screen and process the corresponding pig farm disease risk prediction data, and transmit the corresponding pig farm disease risk prediction data in real time.

[0155] The visualization module is used to visually display the disease risk prediction data of the pig farm.

[0156] In the system solution embodiment of the present invention, the specific details of the method steps involved in the pig farm disease risk prediction based on a small sample scenario have been described above. That is to say, the functional modules in the system are used to implement the steps or sub-steps in the above method embodiment, which will not be repeated here.

[0157] To achieve the above objectives, the present invention also provides a pig farm disease risk prediction platform based on small sample scenarios, such as... Figure 9 As shown, it includes a processor, memory, and a control program for a pig farm disease risk prediction platform based on small sample scenarios;

[0158] The processor executes the control program for the pig farm disease risk prediction platform based on small sample scenarios. This control program is stored in the memory and implements the steps of the pig farm disease risk prediction method based on small sample scenarios. For example:

[0159] S1. Real-time acquisition of raw record data corresponding to pig farm diseases, and determination and generation of pig farm disease record data corresponding to small sample scenarios;

[0160] S2. Based on the original records of pig farm diseases in a small sample scenario, construct a corresponding knowledge fusion model, and combine the knowledge fusion model to generate early warning data corresponding to pig farm diseases in real time.

[0161] S3. Construct a corresponding risk point identification model, and based on the risk point identification model, generate pig farm disease risk prediction data corresponding to the early warning data in real time.

[0162] The specific details of the steps have been explained above and will not be repeated here.

[0163] In this embodiment of the invention, the built-in processor of the pig farm disease risk prediction platform based on small sample scenarios can be composed of integrated circuits. For example, it can be composed of a single packaged integrated circuit, or it can be composed of multiple integrated circuits packaged with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor connects to various components using various interfaces and lines, and executes programs or units stored in memory by running or executing them, and calling data stored in memory, to perform various functions of pig farm disease risk prediction based on small sample scenarios and process data.

[0164] The memory is used to store program code and various data. It is installed in the pig farm disease risk prediction platform based on small sample scenarios and enables high-speed and automatic access to programs or data during operation.

[0165] The memory includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0166] To achieve the above objectives, the present invention also provides a computer-readable storage medium, such as... Figure 10 As shown, the computer-readable storage medium stores a control program for a pig farm disease risk prediction platform based on small sample scenarios. This control program implements the steps of the pig farm disease risk prediction method based on small sample scenarios, for example:

[0167] S1. Real-time acquisition of raw record data corresponding to pig farm diseases, and determination and generation of pig farm disease record data corresponding to small sample scenarios;

[0168] S2. Based on the original records of pig farm diseases in a small sample scenario, construct a corresponding knowledge fusion model, and combine the knowledge fusion model to generate early warning data corresponding to pig farm diseases in real time.

[0169] S3. Construct a corresponding risk point identification model, and based on the risk point identification model, generate pig farm disease risk prediction data corresponding to the early warning data in real time.

[0170] The specific details of the steps have been explained above and will not be repeated here.

[0171] In the description of embodiments of the present invention, it should be noted that any process or method description in the flowcharts or otherwise described herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order according to the functions involved, as should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0172] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processing module, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, a “computer-readable medium” can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM).

[0173] Furthermore, the computer-readable medium can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0174] In this embodiment of the invention, to achieve the above objective, the invention also provides a chip system, the chip system including at least one processor, wherein when program instructions are executed in the at least one processor, the chip system performs the steps of the pig farm disease risk prediction method based on small sample scenarios, for example:

[0175] S1. Real-time acquisition of raw record data corresponding to pig farm diseases, and determination and generation of pig farm disease record data corresponding to small sample scenarios;

[0176] S2. Based on the original records of pig farm diseases in a small sample scenario, construct a corresponding knowledge fusion model, and combine the knowledge fusion model to generate early warning data corresponding to pig farm diseases in real time.

[0177] S3. Construct a corresponding risk point identification model, and based on the risk point identification model, generate pig farm disease risk prediction data corresponding to the early warning data in real time.

[0178] The specific details of the steps have been explained above and will not be repeated here.

[0179] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0180] This invention acquires raw record data corresponding to swine farm diseases in real time, and identifies and generates swine farm disease record data corresponding to small sample scenarios. Based on the raw record data of swine farm diseases in the small sample scenarios, a corresponding knowledge fusion model is constructed, and early warning data corresponding to swine farm diseases is generated in real time using the knowledge fusion model. A corresponding risk point identification model is constructed, and swine farm disease risk prediction data corresponding to the early warning data is generated in real time based on the risk point identification model. The invention also includes a system, platform, and storage medium corresponding to the method. This invention can assist frontline swine farm managers in early detection, accurate location, timely prevention and control, and continuous tracking. Through risk early warning, timely intervention measures such as immunization, disinfection, and purification can be implemented in potentially diseased swine farms, thereby reducing abortion rates and mortality rates during farrowing, and thus avoiding economic losses.

[0181] In other words, it possesses the following core capabilities: Disease risk warning and risk point identification: This system enables risk warning for two common major swine diseases, "PRRS" and "swine viral diarrhea," supporting prediction of disease incidence 1-3 weeks in advance. It also supports multi-factor diagnostics and risk point identification. Small sample learning capability: This system has developed a new multi-stage, multi-strategy, multi-model intelligent learning algorithm that can achieve sufficient learning and usable results with thousands of labeled samples combined with massive amounts of unlabeled data. Continuous learning capability: The system maintains a closed-loop linkage with frontline disease detection, rapidly accumulating labeled data through an active learning mechanism, enabling the system to continuously learn.

[0182] This system simultaneously targets both porcine reproductive and respiratory syndrome (PRRS) and viral diarrhea, achieving both early warning and risk point identification capabilities. It establishes a closed-loop system from online application to algorithm training, enabling continuous automatic learning. For disease early warning systems operating in limited-sample learning scenarios, a multi-stage, multi-strategy, and multi-model core algorithm system is constructed. While ensuring system effectiveness, the data requirement is reduced to the thousands, significantly lowering system construction costs and deployment difficulty.

[0183] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for predicting disease risk in pig farms based on small sample scenarios, characterized in that, The method includes the following steps: Real-time acquisition of raw record data corresponding to pig farm diseases, and determination and generation of pig farm disease record data corresponding to small sample scenarios; Based on the original disease records of pig farms in a small sample scenario, a corresponding knowledge fusion model is constructed. Combined with this knowledge fusion model, real-time early warning data corresponding to pig farm diseases is generated. The process of constructing the knowledge fusion model includes: An epidemic identification model was trained based on a small amount of labeled data; Based on the aforementioned epidemic identification model, unlabeled historical data is identified, and pseudo-labeled data is generated; By combining the original labeled data with the pseudo-labeled data, and using a multi-strategy learning strategy, a predictive model for disease early warning is trained through a training process involving multi-task transfer learning and knowledge fusion. Construct a corresponding risk point identification model, and based on the risk point identification model, generate pig farm disease risk prediction data corresponding to the early warning data in real time.

2. The method for predicting disease risk in pig farms based on small sample scenarios according to claim 1, characterized in that, The real-time acquisition of raw record data corresponding to pig farm diseases, and the determination and generation of pig farm disease record data corresponding to small sample scenarios, also includes: Deep learning is used to process disease data in small sample scenarios, and unlabeled sample data corresponding to the original disease records of pig farms in small sample scenarios are labeled and processed. Based on the disease data from the small sample scenario, an epidemic identification model is constructed in real time.

3. The method for predicting disease risk in pig farms based on small sample scenarios according to claim 2, characterized in that, The deep learning-corresponding small-sample scenario disease data, and the annotation and processing of unlabeled sample data corresponding to the original disease record data of pig farms in the small-sample scenario, also include: Preprocess disease characteristic data in small sample scenarios and perform pseudo-labeling on unlabeled samples from pig farms.

4. The method for predicting disease risk in pig farms based on small sample scenarios according to claim 1, characterized in that, The step of constructing a corresponding knowledge fusion model based on the original disease records of pig farms in a small sample scenario, and generating real-time early warning data corresponding to pig farm diseases in combination with the knowledge fusion model, also includes: By combining multi-strategy learning strategies, multiple disease prediction models are trained and constructed; the training process includes multi-task knowledge learning and knowledge fusion.

5. The method for predicting disease risk in pig farms based on small sample scenarios according to claim 1, characterized in that, The step of constructing a corresponding risk point identification model and generating pig farm disease risk prediction data corresponding to the early warning data in real time based on the risk point identification model also includes: Generate and acquire early warning risk data; Train and construct a proxy explanation model, and based on the proxy explanation model, perform real-time attribution processing on the observation factors corresponding to the early warning risk data.

6. A method for predicting disease risk in pig farms based on small sample scenarios, as described in claim 1 or 5, characterized in that, After constructing the corresponding risk point identification model and generating pig farm disease risk prediction data corresponding to the early warning data in real time based on the risk point identification model, the method further includes: The system detects, investigates, and processes corresponding pig farm disease risk prediction data, and transmits the corresponding pig farm disease risk prediction data in real time. The disease risk prediction data for the pig farm is visualized.

7. A disease risk prediction system for pig farms based on small sample scenarios, characterized in that, The system includes: The acquisition and judgment unit is used to acquire raw record data corresponding to pig farm diseases in real time, and to judge and generate pig farm disease record data corresponding to small sample scenarios. The first construction and generation unit is used to construct a corresponding knowledge fusion model based on the original disease record data of pig farms in a small sample scenario, and to generate early warning data corresponding to pig farm diseases in real time by combining the knowledge fusion model. The process of constructing the knowledge fusion model includes: An epidemic identification model was trained based on a small amount of labeled data; Based on the aforementioned epidemic identification model, unlabeled historical data is identified, and pseudo-labeled data is generated; By combining the original labeled data with the pseudo-labeled data, and using a multi-strategy learning strategy, a predictive model for disease early warning is trained through a training process involving multi-task transfer learning and knowledge fusion. The second construction and generation unit is used to construct a corresponding risk point identification model and, based on the risk point identification model, generate pig farm disease risk prediction data corresponding to the early warning data in real time.

8. A pig farm disease risk prediction system based on a small sample scenario according to claim 7, characterized in that, The acquisition and determination unit further includes: The annotation processing module is used for deep learning of disease data in corresponding small sample scenarios, and for annotating and processing unlabeled sample data corresponding to the original disease record data of pig farms in small sample scenarios. The first construction module is used to build an epidemic identification model in real time based on the small sample disease data. And / or, the annotation processing module further includes: The preprocessing module is used to preprocess disease characteristic data in small sample scenarios and to perform pseudo-labeling on unlabeled samples from pig farms. And / or, the first construction generation unit further includes: The training module is used to train and build multiple disease prediction models by combining multi-strategy learning strategies; the training process includes multi-task knowledge learning and knowledge fusion. And / or, the second construction generation unit further includes: The first generation module is used to generate and acquire early warning risk data; The second construction module is used to train and build a proxy explanation model, and based on the proxy explanation model, to perform real-time attribution processing on the observation factors corresponding to the early warning risk data; And / or, the system further includes: The detection and screening module is used to detect, screen and process the corresponding pig farm disease risk prediction data, and transmit the corresponding pig farm disease risk prediction data in real time. The visualization module is used to visually display the disease risk prediction data of the pig farm.

9. A disease risk prediction platform for pig farms based on small sample scenarios, characterized in that, This includes the processor, memory, and control program for a pig farm disease risk prediction platform based on small sample scenarios; The processor executes the control program for the pig farm disease risk prediction platform based on small sample scenarios, which is stored in the memory. The control program for the pig farm disease risk prediction platform based on small sample scenarios implements the pig farm disease risk prediction method based on small sample scenarios as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a control program for a pig farm disease risk prediction platform based on small sample scenarios. The control program for the pig farm disease risk prediction platform based on small sample scenarios implements the pig farm disease risk prediction method based on small sample scenarios as described in any one of claims 1 to 6.

Images