Complication risk warning method, device, electronic device and storage medium

Abstract

The present invention provides a complication risk warning method, including: extracting all symptom features of the target organism to determine a symptom feature vector combination; inputting the symptom feature vector combination into a complication risk warning model to obtain the warning level of all complications corresponding to the symptom feature vector combination; the symptom feature vector combination is a combination of quantized representations of all symptom feature vectors; the complication risk warning model is obtained based on multiple feature vector combination samples and complication warning level label training corresponding to each feature vector combination sample. The present invention enables practitioners to be free from the constraints of professional knowledge level, and is widely used in intelligent risk warning assistance for various organisms, saving the time cost of combing decision rules one by one, so it can shorten the calculation feedback time of complication risk warning, making the organism complication risk warning model easier to promote among grassroots breeders.

Description

Technical Field

[0001] The present invention relates to the technical field of medical auxiliary decision-making, and in particular to a complication risk warning method, device, electronic equipment and storage medium. Background Art

[0002] In recent years, with the vigorous development of artificial intelligence, new technical means have been brought to the health management of traditional laying hens. More and more machine learning and deep learning technologies have been applied to health risk prediction, realizing the update and iteration of health risk models from traditional statistical models to data-driven intelligent optimization algorithms. Among them, deep learning is a typical data-driven technology, which is often used to deal with some problems with complex environmental information, fuzzy knowledge background and unclear reasoning rules. It has the advantages of data-oriented, nonlinear fitting, and autonomous learning of potential rules. Compared with the traditional disease expert system with high knowledge threshold and cumbersome inquiry process, the deep learning inquiry model can quickly establish an accurate, efficient and flexible disease inquiry model by integrating and refining disease case information, saving knowledge extraction costs, breaking through the "bottleneck" of expert knowledge acquisition, and realizing intelligent risk prediction of diseases.

[0003] The construction of existing disease risk prediction models has its own shortcomings. The disadvantage of rule-based models is that they require expert veterinarians to comprehensively and systematically extract and represent disease knowledge, and during use, the original system cannot be updated as new cases are added, so it has low flexibility and unsatisfactory accuracy. The disadvantage of machine learning algorithm-based models is that model establishment requires reasonable mathematical assumptions about the background distribution of a small amount of data for modeling. When the data is far from the mathematical assumptions, complex variable functions cannot be effectively estimated.

[0004] In addition, previous studies have not mentioned solutions for risk warnings of one disease causing another disease or complication during its development, and the inquiry results are limited to outputting classification results, which are technically unable to provide further support for risk warnings.

[0005] However, there is currently no technical solution that can solve the above technical problems. Specifically, there is no complication risk warning method, device, electronic device and storage medium. Summary of the invention

[0006] The purpose of the present invention is to provide a complication risk early warning method, comprising:

[0007] extracting all symptom features of the target organism to determine a symptom feature vector combination;

[0008] Inputting the symptom feature vector combination into a complication risk warning model to obtain the warning levels of all complications corresponding to the symptom feature vector combination;

[0009] The symptom feature vector combination is a combination of all symptom feature vector quantization representations;

[0010] The complication risk warning model is obtained by training based on multiple feature vector combination samples and complication warning level labels corresponding to each feature vector combination sample.

[0011] According to a complication risk early warning method provided by the present invention, before extracting all symptom features of the target organism to determine the symptom feature vector combination, the method comprises:

[0012] Construct the feature vector combination samples corresponding to all sample cases;

[0013] Determine the complication warning level label corresponding to each sample case;

[0014] Constructing a complication risk warning model based on the feature vector combination samples and the complication warning level labels corresponding to the feature vector combination samples;

[0015] The feature vector combination sample is composed of a preset number of sample feature vectors.

[0016] According to a complication risk warning method provided by the present invention, before building a complication risk warning model based on the feature vector combination sample and the complication warning level label corresponding to the feature vector combination sample, the method further includes:

[0017] Get all sample features;

[0018] Obtaining the warning level of each first main symptom based on a single first-level symptom feature;

[0019] Obtaining the warning level of each second main symptom based on multiple secondary symptom characteristics;

[0020] Based on a single first-level symptom feature and multiple second-level symptom features, obtaining the warning levels of all first complications related to each first main symptom;

[0021] Obtaining the warning levels of all second complications associated with each first main symptom based on a single first-level symptom feature, multiple second-level symptom features, and multiple third-level symptom features;

[0022] Obtaining the warning levels of all third complications related to each second main symptom based on a plurality of second-level symptom characteristics and a plurality of third-level symptom characteristics;

[0023] The complication warning level label is a collection of the warning levels of all first main symptoms, the warning levels of all second main symptoms, the warning levels of all first complications, the warning levels of all second complications, and the warning levels of all third complications;

[0024] The sample characteristics include at least one of a primary symptom characteristic, a secondary symptom characteristic, and a tertiary symptom characteristic.

[0025] According to a complication risk warning method provided by the present invention, the complication risk warning model is constructed based on the feature vector combination sample and the complication warning level label corresponding to the feature vector combination sample, including:

[0026] Construct a feature dataset based on the sample features and complication warning level labels corresponding to each sample case;

[0027] Each of the feature data sets is vectorized to obtain feature vector combination samples to train the feature vector combination samples to build a complication risk warning model.

[0028] According to a complication risk warning method provided by the present invention, all symptom characteristics of the target organism include at least one of appearance characteristics, anatomical characteristics, respiratory characteristics, abnormal behavior characteristics, body temperature characteristics, and disease condition characteristics.

[0029] According to a complication risk warning method provided by the present invention, the target organism is a colony of target organisms;

[0030] Accordingly, all symptom characteristics of the target organism also include at least one of inoculation characteristics, breeding characteristics, feces characteristics, season characteristics, age characteristics, environmental temperature characteristics, environmental humidity characteristics, food intake characteristics, and feed characteristics.

[0031] According to a complication risk warning method provided by the present invention, the target organism is an egg-laying poultry;

[0032] Accordingly, all symptom characteristics of the target organism also include at least one of egg shape characteristics, feather shape characteristics, and egg-laying ability characteristics.

[0033] The present invention also provides a complication risk warning device, comprising:

[0034] Determination device: used for extracting all symptom features of the target organism to determine a symptom feature vector combination;

[0035] Acquisition device: used for inputting the symptom feature vector combination into the complication risk warning model to obtain the warning levels of all complications corresponding to the symptom feature vector combination.

[0036] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the complication risk warning method when executing the program.

[0037] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the complication risk warning method are implemented.

[0038] Compared with the prior art, the complication risk warning model provided by the present invention is not restricted by the level of veterinary knowledge and is widely used in intelligent inquiry assistance for various organisms. It fills the gap in the industry that a complication risk warning model based on an intelligent algorithm cannot be established due to the lack of real cases. The neural network model also realizes a more flexible and effective internal function form than the traditional statistical method. A complication risk warning model based on a regression algorithm is established to improve the shortcomings of previous models or systems that can only output symptom categories but cannot output symptom risk warnings. At the same time, the multi-label model solves the problem of complication decision-making assistance, making the decision-making assistance results more in line with real expectations and improving user satisfaction. The complication risk warning model is based on a standard symptom library and manual labeling by professional veterinarians. According to the standardized process, the construction of the entire complication risk warning model can be completed, saving the time cost of sorting out decision rules one by one, and shortening the calculation feedback time, making the organism complication risk warning model easier to promote among grassroots breeders. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 This is one of the flow charts of a complication risk early warning method provided by the present invention;

[0041] Figure 2 This is the second flow chart of a complication risk warning method provided by the present invention;

[0042] Figure 3 This is the third flow chart of a complication risk warning method provided by the present invention;

[0043] Figure 4 It is a schematic diagram of a process for determining a complication risk warning model provided by the present invention;

[0044] Figure 5 This is the fourth flow chart of a complication risk warning method provided by the present invention;

[0045] Figure 6It is a structural schematic diagram of a complication risk warning device provided by the present invention;

[0046] Figure 7 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Due to the remote breeding areas, backward breeding disease prevention and control technology, lack of information and shortage of breeding professionals, the morbidity and mortality of livestock in small-scale farms are generally high, the economic losses are very serious, and the health problems of livestock have always troubled grassroots farmers. In order to solve the dilemma of free-range groups having nowhere to seek medical treatment, an expert system based on statistical models and rules with a large amount of knowledge and experience of veterinary experts was initially developed, proving the feasibility of online health management. Subsequently, a remote health management service center based on a remote video conferencing system was established, creating a precedent for the application of remote health management technology in the field of animal husbandry and veterinary medicine. However, the service center has high requirements for the professional level of veterinarians, and has restrictions on working hours, locations and network transmission. It cannot guarantee 24-hour standby and solve the inquiries and questions of a large number of small-scale but large-scale farmers anytime and anywhere, so it has not been applied on a large scale.

[0049] The existing health management systems are divided into two categories: rule-based and machine learning algorithm-based. Rule-based intelligent management represents an expert system that includes three modules: human-computer interface, knowledge base, and inference engine. To obtain knowledge from the knowledge base, it is necessary to first consult relevant books, sort out the original knowledge, and then process and fuzzify the original knowledge through experts in various disciplines. Finally, after being judged by an expert evaluation group, the knowledge used to establish the knowledge base is formed. The developer of the inference engine module proposed a knowledge representation method, and thus realized the construction of a dairy cow health management expert system for more than 300 diseases based on more than 2,500 rules. Later, some researchers realized the intelligent health management of chicken diseases based on the basic principles of information theory, control theory, and system theory, fuzzy mathematics membership, and grayscale two degrees in gray systems. The above method is extremely dependent on the selection of judgment thresholds and symptom combinations, which requires professional veterinarians to extract a large amount of rule knowledge through cases. However, if uncollected knowledge is encountered during the work process, this type of health management system will not be able to issue accurate predictions, and even cannot output prediction results for uncollected symptom combinations. Compared with machine learning intelligent health management methods, it is less flexible.

[0050] Intelligent health management based on machine learning is represented by an intelligent health management system based on the Bayesian algorithm. The system first calculates the prior probability as the initial judgment data of the model, and then solves the probability of each disease under the condition of the occurrence of this symptom for the given symptom to be classified, so as to predict the corresponding result. It should be noted that the Bayesian model requires a strict distribution independence assumption premise, but real cases are often difficult to meet. At the same time, the calculation of the prior probability is too dependent on subjective data, resulting in poor generalization and fault tolerance, and poor discrimination ability for input samples with large noise or defects. Later, some researchers tried to use the machine learning algorithm (support vector machine) for binary classification to solve the problem of animal intelligent inquiry. However, the machine learning algorithm uses quadratic programming to solve the support vector, involving high-order matrix calculation. As the sample size increases, the computational complexity increases exponentially. When the number of samples is huge, the matrix storage and calculation will consume a lot of machine memory and computing time, and the solution complexity is high. Therefore, this type of model is not suitable for big data modeling, but smaller data sets cannot fully express the query knowledge to be extracted. At present, there is a lack of a balanced solution between the two. Other researchers have proposed an expert system that combines query rules and neural network algorithms. The system extracts fuzzy rules based on the weights of the neural network, generates reasoning modules, and provides intelligent monitoring inquiries for 30 common chicken infectious diseases, nutritional metabolism, and parasitic diseases. However, due to the lack of only 228 cases, the necessary conditions for fully expressing the query rules are insufficient. At the same time, the reasoning module based on fuzzy rules frequently leads to unreasonable query results, which reduces the universality of the system to a certain extent.

[0051] Figure 1 This is one of the flow diagrams of a complication risk warning method provided by the present invention. The present invention constructs a basic symptom library based on the pathological characteristics of typical biological diseases, and designs simulation data set extraction rules after vectorizing veterinary expert knowledge, and deeply mines the input simulation case data through a multi-label neural network regression model to extract the symptom type association hidden in the case data set, predicts the possible risk warning of complications and symptom risk warning, including:

[0052] extracting all symptom features of the target organism to determine a symptom feature vector combination;

[0053] Inputting the symptom feature vector combination into a complication risk warning model to obtain the warning levels of all complications corresponding to the symptom feature vector combination;

[0054] The symptom feature vector combination is a combination of all symptom feature vector quantization representations;

[0055] The complication risk warning model is obtained by training based on multiple feature vector combination samples and complication warning level labels corresponding to each feature vector combination sample.

[0056] In step S101, the target organism is preferably an animal, more preferably a cultivated object, and the cultivated object may be an egg-laying poultry. In other embodiments, the target organism may also be a human.

[0057] All symptom characteristics of the target organism include at least one of appearance characteristics, anatomical characteristics, respiratory characteristics, abnormal behavior characteristics, body temperature characteristics, and disease condition characteristics.

[0058] The appearance characteristics include the external features of the target organism, whether deformation occurs, whether changes in the color or shape of the external appearance occur, etc. The anatomical characteristics include the external appearance, structure and body fluid detection of various organs and tissues in the dead organism. The respiratory characteristics include parameters such as gas flow and respiratory frequency during the inhalation and exhalation phases of the target organism. The abnormal behavioral characteristics include whether the organism is excited, depressed, drowsy, and whether it performs actions that are not performed normally. The body temperature characteristics include the measured surface temperature and the historical surface temperature. The morbidity characteristics include obvious characteristics that can be distinguished from daily life, such as tears, runny nose, diarrhea, abdominal pain, weakness, etc.

[0059] Optionally, the target organism is a colony of target organisms;

[0060] Accordingly, all symptom characteristics of the target organism also include at least one of inoculation characteristics, breeding characteristics, feces characteristics, season characteristics, age characteristics, environmental temperature characteristics, environmental humidity characteristics, food intake characteristics, and feed characteristics.

[0061] When the target organism is a cluster of target organisms, that is, the object of research and monitoring is a breeding group, such as a herd of cattle, a herd of sheep, a herd of ostriches, a herd of pigs, etc., accordingly, it is necessary to monitor relevant parameters in the cluster that can predict complications, such as vaccination characteristics, that is, whether the cluster is vaccinated according to the standard at each growth stage; the breeding characteristics are information on whether there is breeding with other foreign organisms in the cluster, which is used for the prediction and prevention of infectious diseases; the fecal characteristics are the excrement excreted by the organisms in the cluster. Since the cluster is managed in a unified manner and cannot be managed separately all the time, the feces of each organism cannot be well distinguished. The approach adopted by the present invention is to provide two judgment criteria for normal or abnormal feces. When the feces are normal, no further judgment is required. If it is abnormal, the cluster will be divided into multiple small clusters for secondary feces collection and detection, and then the locking range will be gradually narrowed down to finally determine the organism with abnormal feces. The seasonal characteristics are to provide a better reference for the prediction of epidemics that are prevalent in each season; the age characteristics can be understood as the growth stage of the organism and the age of the organism; the ambient temperature characteristics are the daily temperature and historical temperature near the farm, which can be obtained through a thermometer or an intelligent terminal application; the ambient humidity characteristics are the daily humidity and historical humidity near the farm, which can be obtained through a hygrometer or an intelligent terminal application; the food intake characteristics are the amount of food that the organisms in the cluster need to be fed every day; the feed characteristics are information such as the composition, production date, and shelf life of the feed.

[0062] Optionally, the target organism is an egg-laying bird;

[0063] Accordingly, all symptom characteristics of the target organism also include at least one of egg shape characteristics, feather shape characteristics, and egg-laying ability characteristics.

[0064] The laying poultry is chicken, duck, goose, pigeon, quail, etc. In such an embodiment, all the symptom characteristics of the target organism also include egg shape characteristics, that is, the external shape of the egg, whether it is deformed, whether it is damaged, the shell quality of the egg, etc.; feather shape characteristics are the color of the feathers, the texture of the feathers, the composition analysis of the feathers, etc.; egg-laying ability characteristics are the egg production of laying poultry per unit time.

[0065] Those skilled in the art understand that extracting all symptom features of the target organism to determine the symptom feature vector combination includes first determining all symptom features of the target organism, then encoding all symptom features of the target organism, obtaining vectorized representations of all its symptom features, and then using them as a combination as an input to the model, that is, the symptom feature vector combination is a combination of the vectorized representations of all symptom features.

[0066] In step S102, the symptom feature vector combination is input into the complication risk warning model to obtain the warning levels of all complications corresponding to the symptom feature vector combination. In such an embodiment, the minimum requirement for the complication risk warning model is that for any symptom feature vector combination as input, there is a corresponding output result in the complication risk warning model. Those skilled in the art understand that the symptom characteristics contained in each symptom feature vector combination as input are different, so when they are input into the complication risk warning model, the warning levels predicted for all complications are also different. The present invention is not limited to studying the symptoms inquired by a single symptom, but can predict complications by combining symptom characteristics at different levels, thereby making up for the technical defect of difficulty in predicting complications in the prior art.

[0067] Furthermore, the complication risk warning model is obtained through training based on multiple feature vector combination samples and complication warning level labels corresponding to each feature vector combination sample. In order to better predict the warning level of complications that can be determined by each symptom feature combination, it is necessary to take the combination formed by all possible foreseeable symptom features as part of the sample input, and take the warning level of complications corresponding to the combination formed by all possible foreseeable symptom features as another part of the sample input. The data set formed by the combination of the two is used as a sample input to train the model, thereby determining the complication risk warning model. These will be further described in the specific implementation methods described later and will not be repeated here.

[0068] The present invention provides a complication risk warning method, which integrates and refines laying hen disease information and knowledge by utilizing the potential feature extraction capability of a neural network, forms an automated inquiry decision model through training and fitting based on case data, replaces the traditional logical rule-based disease self-service health management expert system with an inquiry model modeled based on real case data, and completes the universal iteration of the laying hen disease self-service health management system from model-driven to data-driven, thereby reducing the cost of knowledge extraction and accurately inquiring about the possibility and probability of illness, thereby improving the application extensibility of the self-service inquiry model paradigm.

[0069] The present invention uses neural network data-driven adaptive technology, omits the step of making a priori assumptions about the problem model, and uses internal neurons to learn and train simulated cases when the internal knowledge of the laying hen disease inquiry problem is unknown or not described, obtains the hidden functional relationship between symptoms and diseases, establishes a functional relationship, and realizes automatic inquiry prediction of laying hen diseases. The present invention first proposes a programmatic logic construction process of a simulation data set to solve the current dilemma of lack of real cases based on deep learning modeling, and then uses a multi-label neural network regression modeling method to provide a solution to the inquiry of disease complications, and then uses the fitting characteristics of the regression neural network to output accurate disease probability for the inquiry of the disease.

[0070] The complication risk warning model provided by the present invention is not restricted by the knowledge level of veterinarians and is widely used in intelligent assisted inquiries of various organisms. It fills the gap in the industry that the complication risk warning model based on intelligent algorithms cannot be established due to the lack of real cases. The neural network model also realizes a more flexible and effective internal function form than the traditional statistical method; the complication risk warning model based on the regression algorithm is established, which improves the shortcomings of previous models or systems that can only output the disease category but cannot output the risk warning level or disease risk probability. At the same time, the multi-label model solves the problem of complication decision-making assistance, making the decision-making assistance results more in line with real expectations and improving user satisfaction; the complication risk warning model is based on the standard disease symptom library and the manual labeling of professional veterinarians. According to the standardized process, the construction of the entire complication risk warning model can be completed, saving the time cost of sorting out the decision rules one by one, and shortening the calculation feedback time, making the organism complication risk warning model easier to promote among grassroots breeders.

[0071] It should be noted that the present invention can not only provide risk warnings for diseases, but more importantly, it can also provide health risk warnings for organisms or groups of organisms based on information such as the appearance characteristics of the organisms and the shape of the organisms, combined with the group characteristics of the groups of organisms, so that farmers can more comprehensively understand the health status, satiety, temperament, mental state, etc. of the organisms in their current state, so as to better manage the farms.

[0072] Figure 2 This is a second flow chart of a complication risk warning method provided by the present invention, which includes: extracting all symptom features of the target organism to determine the symptom feature vector combination before the symptom feature vector combination is determined;

[0073] Construct the feature vector combination samples corresponding to all sample cases;

[0074] Determine the complication warning level label corresponding to each sample case;

[0075] Constructing a complication risk warning model based on the feature vector combination samples and the complication warning level labels corresponding to the feature vector combination samples;

[0076] The feature vector combination sample is composed of a preset number of sample feature vectors.

[0077] In step S201, all sample symptom features of the sample organism are extracted, all sample symptom feature vectors are quantized, and then all feature vector combination samples in all sample symptom feature vectors are determined by selecting any combination form from all sample symptom features. The all feature vector combination samples include all combination forms. The all feature vector combination samples are sample feature vector clusters formed by combining all sample feature vectors based on the number of all sample feature vectors, as shown in the following table:

[0078]

[0079] The present invention preferably selects laying hens as the research object to illustrate the technical solution of the present invention. Specifically, in view of the problem of insufficient real cases in the current laying hen health risk early warning modeling, the present invention provides a simulation data set extraction logic method to meet the data requirements required for laying hen health risk early warning modeling based on neural networks. Assuming that the symptoms are divided into m levels of strength and weakness, and the minimum description of the symptoms that express the disease in real situations is n, for each disease to be warned by the laying hen health risk early warning model, the symptoms contained therein are randomly extracted and combined without distinguishing levels, and the extracted symptoms are allowed to be all of the same level. Then, for each disease, there are a total of n symptoms. As shown in the table above, for example, there are 5 symptom levels in a disease-symptom database, and each sample case needs more than 5 symptoms to be identified. Then for each disease, there are 126 extraction logics for 5 symptom sample cases, namely There are 210 extraction logics for the 6 symptom sample cases, namely

[0080] In step S202, all complication warning level labels corresponding to each sample case are determined based on the historical database or the prediction results given by the veterinarian. This step is a pre-set determination step, and in other embodiments, all complication warning level labels corresponding to each sample case can be re-determined based on the subsequently updated historical database or the updated prediction results given by the veterinarian.

[0081] In step S203, based on the quantized laying hen disease simulation data set, an animal disease inquiry model based on multi-label neural network regression is established. Multi-label regression is an extension of the traditional regression model. It predicts multiple continuous targets in the same input space and predicts that the disease level of a certain sample varies between [0,1]. Assume that the quantized data set is:

[0082] D labeled =(X,Y)(1)

[0083] A training set of N samples, where is composed of n eigenvectors (X 1 ,,…,X j , …X n ), is composed of k target vectors (Y 1 ,…,Y i ,…Y k ) constitutes the output space. (x (l) ,y (l) ) represents a sample, containing n-dimensional input variables and k-dimensional output variables Where i∈{1,…,k}, j∈{1,…,d}, l∈{1,…,N}. The construction of the animal disease inquiry model is to learn the mapping function h:X→Y so that the symptom combination vector of any unknown input It can simultaneously predict the target of all complication warning levels

[0084] The neural network multi-marker regression algorithm proposed in the present invention can effectively model the complex relationship between input symptoms and output symptoms while revealing the correlation between target symptoms, achieving the purpose of predicting multiple continuous variables with a set of common input variables, thereby solving the problem that the intelligent health management model of laying hen diseases is difficult to predict complications and accurately outputs the warning level of a certain disease.

[0085] Figure 3 This is a flow chart of a complication risk warning method provided by the present invention. Before building a complication risk warning model based on the feature vector combination sample and the complication warning level label corresponding to the feature vector combination sample, it also includes:

[0086] Get all sample features;

[0087] Obtaining the warning level of each first main symptom based on a single first-level symptom feature;

[0088] Obtaining the warning level of each second main symptom based on multiple secondary symptom characteristics;

[0089] Based on a single first-level symptom feature and multiple second-level symptom features, obtaining the warning levels of all first complications related to each first main symptom;

[0090] Obtaining the warning levels of all second complications associated with each first main symptom based on a single first-level symptom feature, multiple second-level symptom features, and multiple third-level symptom features;

[0091] Obtaining the warning levels of all third complications related to each second main symptom based on a plurality of second-level symptom characteristics and a plurality of third-level symptom characteristics;

[0092] The complication warning level label is a collection of the warning levels of all first main symptoms, the warning levels of all second main symptoms, the warning levels of all first complications, the warning levels of all second complications, and the warning levels of all third complications;

[0093] The sample characteristics include at least one of a primary symptom characteristic, a secondary symptom characteristic, and a tertiary symptom characteristic.

[0094] In step S301, all sample features are acquired based on historical query information, historical database information, or feature information recorded by a doctor.

[0095] In step S302, the warning levels of all first main symptoms are obtained based on a single first-level symptom feature. In such an embodiment, the first-level symptom feature can be identified as the corresponding first main symptom when it occurs, or a warning level of the first main symptom can be given.

[0096] In step S303, the warning level of each second main symptom is obtained based on multiple secondary symptom characteristics, that is, when multiple secondary symptoms appear in combination, they can be basically identified as corresponding symptoms, or the warning level of the second main symptom is given.

[0097] In step S304, based on a single first-level symptom feature and multiple second-level symptom features, the warning levels of all first complications related to each first main symptom are obtained. In such an embodiment, a single first-level symptom feature and multiple second-level symptom features will be used in combination to determine the warning level of the first complication corresponding to each first main symptom. The warning level of the first complication here can be one or more complications of a specific symptom, or one or more complications of a non-specific symptom. When a single first-level symptom feature appears alone in the sample, it is difficult to identify it as the corresponding symptom, and it needs to be assisted by multiple second-level symptom features to determine whether it is the first complication corresponding to the first main symptom.

[0098] In step S305, the warning levels of all second complications related to each first main symptom are obtained based on a single first-level symptom feature, multiple second-level symptom features, and multiple third-level symptom features. Those skilled in the art understand that the symptom features are divided according to the strength of the symptoms, from strong to weak as first-level symptom features, second-level symptom features, and third-level symptom features, that is, they are used in combination with first-level symptom features, second-level symptom features, and third-level symptom features. The third-level symptom features are used as weak symptoms to assist in judgment. When the second-level symptom features and the third-level symptom features appear alone in the sample, they cannot be identified as the corresponding symptoms. Only when the first-level symptom features are used as the main judgment basis to determine that it is the first main symptom, and then with the assistance of the second-level symptom features and the third-level symptom features, can the warning level of the second complication corresponding to each first main symptom be determined.

[0099] In step S306, the warning levels of all third complications related to each second main symptom are obtained based on multiple secondary symptom characteristics and multiple tertiary symptom characteristics. In such an embodiment, the second main symptom is determined based on multiple secondary symptom characteristics, and the warning level of the third complications corresponding to each second main symptom is determined with the auxiliary judgment of multiple tertiary symptom characteristics.

[0100] The complication warning level label is a collection of the warning levels of all first main symptoms, the warning levels of all second main symptoms, the warning levels of all first complications, the warning levels of all second complications and the warning levels of all third complications. The sample characteristics include at least one of the first-level symptom characteristics, the second-level symptom characteristics and the third-level symptom characteristics. In other embodiments, the symptom level can also be four levels, five levels or even more, and the judgment process can also be more complicated. However, the core of the present invention is to protect such a symptom level division method, and then determine the disease symptoms corresponding to different complications through such division method, so as to better realize the establishment of a complication risk warning model.

[0101] For example, through the first-level symptom characteristics, it is determined that the target organism may have disease A, and then through multiple second-level symptom characteristics, it is determined that the target organism not only has disease A, but also has a complication associated with disease A. For another example, through multiple second-level symptom characteristics, it is determined that the target organism may have disease B, and then through multiple third-level symptom characteristics, it is determined that the target organism not only has disease B, but also has multiple complications associated with disease B. These are all determined based on the set of warning levels of all first main symptoms, all second main symptoms, all first complications, all second complications, and all third complications.

[0102] Figure 4is a flow chart of determining a complication risk warning model provided by the present invention, wherein the complication risk warning model is constructed based on the feature vector combination sample and the complication warning level label corresponding to the feature vector combination sample, including:

[0103] Construct a feature dataset based on the sample features and complication warning level labels corresponding to each sample case;

[0104] Each of the feature data sets is vectorized to obtain feature vector combination samples to train the feature vector combination samples to build a complication risk warning model.

[0105] In step S2031, the extracted data set is characterized and divided according to the complication warning level possibility standard of each case from 0% to 100%. Assuming there are k types of diseases, the data set can be expressed by the following formula:

[0106]

[0107] In formula (1), N is the number of samples extracted;

[0108]

[0109] Formula (2) is the i-th sample case containing n symptoms,

[0110]

[0111] In formula (3), the i-th sample has a warning level label for k diseases. For N samples, the warning level of each sample in k diseases is calibrated to obtain a labeled dataset D labeled , which is the feature data set.

[0112] In step S2032, the simulation data set D is encoded using the one-hot encoding technique. labeled Quantification. Assuming that there are C symptoms in the disease-symptom library, the one-hot technique uses a C-bit status register to encode the C symptoms. Each symptom has an independent register bit, and at any time, only one of them is valid. The simulation case is expressed as:

[0113]

[0114] In formula (4), if j∈X i , e j =1 if 0<j≤n, thereby converting the text symptoms of the simulation data set into digital symptoms that can be read by the neural network, and dividing the quantified data set into a training set and a validation set in a ratio of 8:2 for the establishment of the complication risk warning model in the aforementioned step S203.

[0115] Figure 5 This is the fourth flow chart of a complication risk warning method provided by the present invention.

[0116] The present invention first obtains sample symptoms based on the standard disease-symptom library, extracts the sample set according to the logic, and then manually calibrates the data set. The sample set reads the sample symptoms, performs one-hot encoding after data shuffling, and then uses the one-hot encoded sample vector data for model training. After multiple evaluations and iterations, the model forms the final complication risk warning model, thereby realizing the prediction of complications.

[0117] Figure 6 : is a structural schematic diagram of a complication risk warning device provided by the present invention. The complication risk warning device provided by the present invention comprises:

[0118] Determination device 1: is used to extract all symptom features of the target organism to determine the symptom feature vector combination. The working principle of the determination device can refer to the aforementioned step S101 and will not be described in detail here.

[0119] The complication risk warning device also includes an acquisition device 2: used to input the symptom feature vector combination into the complication risk warning model to obtain the warning levels of all complications corresponding to the symptom feature vector combination. The working principle of the acquisition device can refer to the aforementioned step S102 and will not be repeated here.

[0120] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the complication risk warning method, which includes: extracting all symptom features of the target organism to determine a symptom feature vector combination; inputting the symptom feature vector combination into a complication risk warning model to obtain the warning level of all complications corresponding to the symptom feature vector combination; the symptom feature vector combination is a combination of all symptom feature vector quantization representations; the complication risk warning model is obtained based on a plurality of feature vector combination samples and the complication warning level label training corresponding to each feature vector combination sample.

[0121] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0122] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute a complication risk warning method provided by the above methods, which method includes: extracting all symptom characteristics of the target organism to determine a symptom feature vector combination; inputting the symptom feature vector combination into a complication risk warning model to obtain the warning level of all complications corresponding to the symptom feature vector combination; the symptom feature vector combination is a combination of quantitative representations of all symptom features; the complication risk warning model is obtained based on multiple feature vector combination samples and complication warning level label training corresponding to each feature vector combination sample.

[0123] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the complication risk warning method provided by the above-mentioned methods, the method comprising: extracting all symptom characteristics of the target organism to determine a symptom feature vector combination; inputting the symptom feature vector combination into a complication risk warning model to obtain the warning level of all complications corresponding to the symptom feature vector combination; the symptom feature vector combination is a combination of quantitative representations of all symptom features; the complication risk warning model is obtained based on multiple feature vector combination samples, and complication warning level label training corresponding to each feature vector combination sample.

[0124] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0125] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A complication risk early warning method, characterized in that: include: extracting all symptom features of the target organism to determine a symptom feature vector combination; The target organism is an egg-laying bird; Inputting the symptom feature vector combination into a complication risk warning model to obtain the warning levels of all complications corresponding to the symptom feature vector combination; The symptom feature vector combination is a combination of all symptom feature vector quantization representations; all symptom features of the target organism include at least one of appearance features, anatomical features, respiratory features, abnormal behavior features, body temperature features, and disease characteristics; the appearance features include the appearance features of the target organism, whether it is deformed, and whether the color and shape of the appearance have changed; the anatomical features include the appearance, structure, and body fluid detection of each organ and tissue in the dead target organism; the respiratory features include the gas flow and respiratory frequency of the target organism during the inhalation and exhalation phases; the abnormal behavior features include whether it is excited, depressed, sleepy, and whether it has taken actions that are not normally taken; the body temperature features include the measured body surface temperature and the historical body surface temperature; the disease characteristics include tears, runny nose, diarrhea, abdominal pain, and weakness; accordingly, all symptom features of the target organism also include at least one of egg shape features, feather shape features, and egg-laying ability features; The complication risk warning model is obtained by training based on multiple feature vector combination samples and complication warning level labels corresponding to each feature vector combination sample; Before extracting all symptom features of the target organism to determine the symptom feature vector combination, including: Construct the feature vector combination samples corresponding to all sample cases; Determine the complication warning level label corresponding to each sample case; Constructing a complication risk warning model based on the feature vector combination samples and the complication warning level labels corresponding to the feature vector combination samples; The feature vector combination sample is composed of a preset number of sample feature vectors; Before constructing a complication risk warning model based on the feature vector combination samples and the complication warning level labels corresponding to the feature vector combination samples, the method further includes: Get all sample features; Obtaining the warning level of each first main symptom based on a single first-level symptom feature; Obtaining the warning level of each second main symptom based on multiple secondary symptom characteristics; Based on a single first-level symptom feature and multiple second-level symptom features, obtaining the warning levels of all first complications related to each first main symptom; Obtaining the warning levels of all second complications associated with each first main symptom based on a single first-level symptom feature, multiple second-level symptom features, and multiple third-level symptom features; Obtaining the warning levels of all third complications related to each second main symptom based on a plurality of second-level symptom characteristics and a plurality of third-level symptom characteristics; The complication warning level label is a collection of the warning levels of all first main symptoms, the warning levels of all second main symptoms, the warning levels of all first complications, the warning levels of all second complications, and the warning levels of all third complications; The sample characteristics include at least one of a primary symptom characteristic, a secondary symptom characteristic, and a tertiary symptom characteristic.

2. The complication risk warning method according to claim 1, characterized in that: The constructing of a complication risk warning model based on the feature vector combination samples and the complication warning level labels corresponding to the feature vector combination samples includes: Construct a feature dataset based on the sample features and complication warning level labels corresponding to each sample case; Each of the feature data sets is vectorized to obtain feature vector combination samples to train the feature vector combination samples to build a complication risk warning model.

3. A complication risk warning device, characterized in that: include: Determination device: used for extracting all symptom features of the target organism to determine a symptom feature vector combination; The target organism is an egg-laying bird; Acquisition device: used for inputting the symptom feature vector combination into the complication risk warning model to obtain the warning levels of all complications corresponding to the symptom feature vector combination; The symptom feature vector combination is a combination of all symptom feature vector quantization representations; all symptom features of the target organism include at least one of appearance features, anatomical features, respiratory features, abnormal behavior features, body temperature features, and disease characteristics; the appearance features include the appearance features of the target organism, whether it is deformed, and whether the color and shape of the appearance have changed; the anatomical features include the appearance, structure, and body fluid detection of each organ and tissue in the dead target organism; the respiratory features include the gas flow and respiratory frequency of the target organism during the inhalation and exhalation phases; the abnormal behavior features include whether it is excited, depressed, sleepy, and whether it has taken actions that are not normally taken; the body temperature features include the measured body surface temperature and the historical body surface temperature; the disease characteristics include tears, runny nose, diarrhea, abdominal pain, and weakness; accordingly, all symptom features of the target organism also include at least one of egg shape features, feather shape features, and egg-laying ability features; The complication risk warning model is obtained by training based on multiple feature vector combination samples and complication warning level labels corresponding to each feature vector combination sample; Before extracting all symptom features of the target organism to determine the symptom feature vector combination, including: Construct the feature vector combination samples corresponding to all sample cases; Determine the complication warning level label corresponding to each sample case; Constructing a complication risk warning model based on the feature vector combination samples and the complication warning level labels corresponding to the feature vector combination samples; The feature vector combination sample is composed of a preset number of sample feature vectors; Before constructing a complication risk warning model based on the feature vector combination samples and the complication warning level labels corresponding to the feature vector combination samples, the method further includes: Get all sample features; Obtaining the warning level of each first main symptom based on a single first-level symptom feature; Obtaining the warning level of each second main symptom based on multiple secondary symptom characteristics; Based on a single first-level symptom feature and multiple second-level symptom features, obtaining the warning levels of all first complications related to each first main symptom; Obtaining the warning levels of all second complications associated with each first main symptom based on a single first-level symptom feature, multiple second-level symptom features, and multiple third-level symptom features; Obtaining the warning levels of all third complications related to each second main symptom based on a plurality of second-level symptom characteristics and a plurality of third-level symptom characteristics; The complication warning level label is a collection of the warning levels of all first main symptoms, the warning levels of all second main symptoms, the warning levels of all first complications, the warning levels of all second complications, and the warning levels of all third complications; The sample characteristics include at least one of a primary symptom characteristic, a secondary symptom characteristic, and a tertiary symptom characteristic.

4. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the complication risk warning method as described in any one of claims 1 to 2 is implemented.

5. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the complication risk warning method according to any one of claims 1 to 2 is implemented.

Images