CN-116525131-B - Pig farm disease risk prediction method, system and platform based on small sample scene

CN116525131BCN 116525131 BCN116525131 BCN 116525131BCN-116525131-B

Abstract

The invention discloses a pig farm disease risk prediction method, a pig farm disease risk prediction system and a pig farm disease risk prediction platform based on a small sample scene, wherein the method is used for acquiring original record data corresponding to pig farm diseases in real time, and judging and generating pig farm disease record data corresponding to the small sample scene; the method comprises the steps of constructing a corresponding knowledge fusion model according to pig farm disease original record data in a small sample scene, combining the knowledge fusion model to generate early warning data corresponding to pig farm diseases in real time, constructing a corresponding risk point identification model, and generating pig farm disease risk prediction data corresponding to the early warning data in real time according to the risk point identification model, so that a first-line pig farm manager can be assisted to find in advance, accurately locate, timely prevent and control and continuously track. Intervention measures such as immunization, disinfection, purification and the like are timely carried out on a potentially pathogenic pig farm through risk early warning, so that the flow rate and the production, death and panning rate of the pig farm are reduced, and further economic losses are avoided.

Inventors

Zeng Zhongjie
CHEN JIAQUAN
CHEN JIAWEI
Huang Juejun
LIN XIAOWAN
ZHANG DEQUAN

Assignees

温氏食品集团股份有限公司

Dates

Publication Date: 20260508
Application Date: 20230324

Claims (10)

1. A pig farm disease risk prediction method based on a small sample scene, which is characterized by comprising the following steps: Acquiring original recorded data corresponding to pig farm diseases in real time, and judging and generating pig farm disease recorded data corresponding to small sample scenes; Constructing a corresponding knowledge fusion model according to pig farm disease original record data in a small sample scene, and generating early warning data corresponding to pig farm diseases in real time by combining the knowledge fusion model, wherein the process for constructing the knowledge fusion model comprises the following steps: Training an epidemic situation recognition model based on a small amount of labeling data; identifying the unlabeled historical data based on the epidemic situation identification model to generate pseudo-labeling data; Combining the original annotation data with the pseudo annotation data, combining a multi-strategy learning strategy, and training to obtain a prediction model for disease early warning through a training process of multi-task transfer learning and knowledge fusion; and constructing a corresponding risk point identification model, and generating pig farm disease risk prediction data corresponding to the early warning data in real time according to the risk point identification model.
2. The method for predicting risk of a pig farm disease based on a small sample scene as claimed in claim 1, wherein the acquiring in real time raw record data corresponding to the pig farm disease and determining and generating pig farm disease record data corresponding to the small sample scene further comprises: Deep learning corresponding small sample scene disease data, and labeling and processing unlabeled sample data corresponding to pig farm disease original record data in a small sample scene; and constructing an epidemic situation identification model in real time according to the small sample scene disease data.
3. The pig farm disease risk prediction method based on a small sample scene according to claim 2, wherein the deep learning of the small sample scene disease data corresponding to the small sample scene and the labeling of the unlabeled sample data corresponding to the pig farm disease raw record data in the small sample scene further comprises: preprocessing small sample scene disease characteristic data, and performing pseudo-label labeling processing on a pig farm unlabeled sample.
4. The method for predicting risk of pig farm diseases based on small sample scene as claimed in claim 1, wherein the constructing a corresponding knowledge fusion model according to pig farm disease original record data in small sample scene, and combining the knowledge fusion model, generating early warning data corresponding to pig farm diseases in real time, further comprises: And training and constructing a plurality of disease prediction models by combining a multi-strategy learning strategy, wherein the training process comprises multi-task knowledge learning and knowledge fusion.
5. The method for predicting risk of pig farm diseases based on small sample scene as claimed in claim 1, wherein said constructing a corresponding risk point recognition model, and generating pig farm disease risk prediction data corresponding to the pre-warning data in real time according to the risk point recognition model, further comprises: Generating and acquiring early warning risk data; And training and constructing a proxy interpretation model, and attributing and processing observation factors corresponding to the early warning risk data in real time according to the proxy interpretation model.
6. The method for predicting risk of pig farm diseases based on small sample scenes according to claim 1 or 5, wherein after constructing a corresponding risk point identification model and generating pig farm disease risk prediction data corresponding to early warning data in real time according to the risk point identification model, further comprises: Detecting and checking corresponding pig farm disease risk prediction data, and transmitting the corresponding pig farm disease risk prediction data in real time; the pig farm disease risk prediction data is visually displayed.
7. A pig farm disease risk prediction system based on a small sample scenario, the system comprising: the acquisition judging unit is used for acquiring original record data corresponding to the pig farm diseases in real time, and judging and generating pig farm disease record data corresponding to the small sample scene; the first construction generating unit is used for constructing a corresponding knowledge fusion model according to pig farm disease original record data in a small sample scene and generating early warning data corresponding to pig farm diseases in real time by combining the knowledge fusion model, wherein the process for constructing the knowledge fusion model comprises the following steps of: Training an epidemic situation recognition model based on a small amount of labeling data; identifying the unlabeled historical data based on the epidemic situation identification model to generate pseudo-labeling data; Combining the original annotation data with the pseudo annotation data, combining a multi-strategy learning strategy, and training to obtain a prediction model for disease early warning through a training process of multi-task transfer learning and knowledge fusion; the second construction generating unit is used for constructing a corresponding risk point identification model and generating pig farm disease risk prediction data corresponding to the early warning data in real time according to the risk point identification model.
8. The piggery disease risk prediction system based on a small sample scenario of claim 7, wherein the acquisition determination unit further comprises: The labeling processing module is used for deeply learning the corresponding small sample scene disease data and labeling and processing the unlabeled sample data corresponding to the pig farm disease original record data in the small sample scene; the first construction module is used for constructing an epidemic situation recognition model in real time according to the small sample scene disease data; And/or, the labeling processing module further comprises: The preprocessing module is used for preprocessing small sample scene disease characteristic data and performing pseudo-label labeling processing on the pig farm unlabeled samples; and/or, the first construction generating unit further comprises: the training construction module is used for combining a multi-strategy learning strategy to train and construct a plurality of disease prediction models, wherein the training process comprises multi-task knowledge learning and knowledge fusion; and/or, the second construction generating unit further includes: the first generation module is used for generating and acquiring early warning risk data; The second construction module is used for training and constructing a proxy interpretation model, and attributing and processing observation factors corresponding to the early warning risk data in real time according to the proxy interpretation model; and/or, the system further comprises: the detection and investigation module is used for detecting and investigating corresponding pig farm disease risk prediction data and transmitting the corresponding pig farm disease risk prediction data in real time; and the visual display module is used for visually displaying the pig farm disease risk prediction data.
9. The pig farm disease risk prediction platform based on the small sample scene is characterized by comprising a processor, a memory and a pig farm disease risk prediction platform control program based on the small sample scene; Wherein executing the small sample scene based pig farm disease risk prediction platform control program in the processor, the small sample scene based pig farm disease risk prediction platform control program being stored in the memory, the small sample scene based pig farm disease risk prediction platform control program implementing the small sample scene based pig farm disease risk prediction method of any of claims 1 to 6.
10. A computer readable storage medium, wherein the computer readable storage medium stores a pig farm disease risk prediction platform control program based on a small sample scene, and the pig farm disease risk prediction platform control program based on a small sample scene implements the pig farm disease risk prediction method based on a small sample scene as set forth in any one of claims 1 to 6.

Description

Pig farm disease risk prediction method, system and platform based on small sample scene Technical Field The invention belongs to the technical field of pig farm disease risk prediction, and particularly relates to a pig farm disease risk prediction method, system and platform based on a small sample scene. Background At present, part of pig farm disease informatization systems on the market mainly aim at informatization management of a certain single disease, and are generally used for disease information management and improvement of detection efficiency of related diseases. However, the existing pig farm disease information system in the market still has a plurality of defects, specifically as follows: The coverage is low, and the pig farm disease informatization system on the market is generally only limited for a certain single disease and common pig farm major diseases. The pig farm disease management system on the market only supports the disease detection function, or risks are quantified in the diseased pig farm, so that the risk prediction of the disease cannot be realized, and advanced prevention and control cannot be realized. Some pig farm informatization systems only output overall risks, and cannot provide the identification and quantification capabilities of related risk points, so that accurate management is difficult to achieve. The labeling data has high requirements, namely a plurality of intelligent systems on the market have higher requirements on standard data quantity, and at least one hundred thousand or even millions of labeling data are needed, so that the difficulty of system construction and popularization is increased. Therefore, in order to overcome the technical defects, there is an urgent need to design and develop a pig farm disease risk prediction method, system and platform based on a small sample scene. Disclosure of Invention In order to overcome the defects and difficulties in the prior art, the invention aims to provide a pig farm disease risk prediction method, a pig farm disease risk prediction system, a pig farm disease risk prediction platform and a pig farm disease risk prediction storage medium based on a small sample scene, which can assist a first-line pig farm manager to realize early discovery, accurate positioning, timely prevention, control and continuous tracking, and timely perform intervention measures such as immunization, disinfection, purification and the like on a potentially ill pig farm through risk early warning, so that the flow rate and the yield, the death and panning rate of the pig farm are reduced, and further economic losses are avoided. The first aim of the invention is to provide a pig farm disease risk prediction method based on a small sample scene; the second object of the invention is to provide a pig farm disease risk prediction system based on a small sample scene; the third object of the invention is to provide a pig farm disease risk prediction platform based on a small sample scene; a fourth object of the present invention is to provide a computer-readable storage medium; the first object of the invention is achieved in that the method comprises the steps of: Acquiring original recorded data corresponding to pig farm diseases in real time, and judging and generating pig farm disease recorded data corresponding to small sample scenes; constructing a corresponding knowledge fusion model according to pig farm disease original record data in a small sample scene, and generating early warning data corresponding to pig farm diseases in real time by combining the knowledge fusion model; and constructing a corresponding risk point identification model, and generating pig farm disease risk prediction data corresponding to the early warning data in real time according to the risk point identification model. Further, the acquiring, in real time, the original recorded data corresponding to the pig farm disease, and determining whether the original recorded data is pig farm disease original recorded data in a small sample scene, further includes: Deep learning corresponding small sample scene disease data, and labeling and processing unlabeled sample data corresponding to pig farm disease original record data in a small sample scene; and constructing an epidemic situation identification model in real time according to the small sample scene disease data. Further, the deep learning of the small sample scene disease data corresponding to the deep learning, the labeling processing of the unlabeled sample data corresponding to the pig farm disease original record data in the small sample scene, and the further comprises: preprocessing small sample scene disease characteristic data, and performing pseudo-label labeling processing on a pig farm unlabeled sample. Further, the constructing a corresponding knowledge fusion model according to the pig farm disease original record data in the small sample scene, and generating early warning data corresp