New infectious disease intelligent identification method based on generative open set identification technology
Through the intelligent identification method of newly-infectious diseases based on generative open-set recognition technology, the encoder network and decoder network are used to identify newly-infectious diseases, which solves the problem that the existing technology cannot identify newly-infectious diseases and achieves disease outbreaks Initial timely identification and control.
Patent Information
- Application Number
- CN202410927130.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-11
- Publication Date
- 2024-11-08
AI Technical Summary
The existing technology cannot effectively identify new infectious diseases and it is difficult to provide timely medical intervention measures in the early stages of a disease outbreak.
Using the intelligent identification method of newly-occurred infectious diseases based on generative open-set recognition technology, the newly-occurred identification model of infectious diseases is constructed by acquiring and preprocessing electronic health record data, and the encoder network and decoder network are used to construct the newly-occurred identification model of infectious diseases, combining the optimal word embedding method and Classifiers, close-set and open-set identification of structured and unstructured data, identify unknown classes as newly-occurred infectious diseases.
It has achieved timely identification in the early stages of new infectious disease outbreaks, and helped relevant institutions to provide specific medical intervention measures to maximize the control of disease transmission.
Smart Images

Figure CN118919092A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data mining, and in particular to an intelligent identification method for emerging infectious diseases based on generative open set identification technology. Background Art
[0002] Emerging infectious diseases are newly identified and previously unknown infectious diseases that cause regional or international public health problems, and they pose a serious threat to human life and health. With the continuous growth of economic globalization and the world's population, economic exchanges and population mobility between countries in the world have become more frequent, which has also accelerated the spread of infectious diseases and expanded their scope.
[0003] How to promptly identify emerging infectious diseases in the early stages of an outbreak remains a challenge. Currently, there are studies on the detection, risk prediction and classification of infectious diseases using electronic health record data. In addition, some scholars have used network data to identify the source of infectious diseases, risk factors, and key nodes in the transmission process. However, these studies focus on monitoring and early warning of known infectious diseases, and it is difficult to identify emerging infectious diseases. Summary of the invention
[0004] 1. Technical issues to be resolved
[0005] In view of the shortcomings of the prior art, the present invention provides a method and system for intelligent identification of emerging infectious diseases based on generative open set recognition technology, which solves the technical problem that the prior art cannot identify emerging infectious diseases.
[0006] (II) Technical solution
[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions:
[0008] In a first aspect, the present invention provides a method for intelligently identifying emerging infectious diseases based on generative open set recognition technology, comprising:
[0009] Acquire and preprocess electronic health record data to obtain normalized data, wherein the normalized data includes structured data and unstructured data;
[0010] Performing closed set recognition on the structured data and the unstructured data respectively to determine the optimal word embedding method, encoder network and optimal classifier;
[0011] Determine a decoder network based on the encoder network, and construct an infectious disease emerging recognition model based on the encoder network, the decoder network and the optimal classifier;
[0012] The unstructured data is processed by the optimal word embedding method to obtain sentence vectors; the sentence vectors and structured data are open-set identified by the new infectious disease identification model to identify unknown classes as new infectious diseases.
[0013] Preferably, closed set recognition is performed on the structured data and the unstructured data respectively to determine the optimal word embedding method, encoder network and optimal classifier, including:
[0014] Different word embedding methods are used to obtain sentence vectors of unstructured data, the effects of different word embedding methods on classification effects are verified, and the optimal word embedding method is selected as the optimal word embedding method; the word embedding methods include Tfidf, Word2vec, Glove, FastText and CNN;
[0015] Adopt the encoder architecture in DeepMicro, that is, use an intermediate hidden layer, set different numbers of hidden units, set the hidden layer dimensions to different units, and choose the architecture with the best classification effect as the encoder network;
[0016] Use different classifiers for classification, and use the one with the smallest loss function as the optimal classifier.
[0017] Preferably, determining a decoder network based on the encoder network comprises:
[0018] Based on the encoder network, a variational autoencoder is used to generate the decoder network.
[0019] Preferably, the unstructured data is processed by the optimal word embedding method to obtain sentence vectors; the sentence vectors and structured data are open-set identified by the new infectious disease identification model to identify unknown classes as new infectious diseases, including
[0020] The optimal word embedding method is used to process unstructured data, and the sentence vector is represented as {V1, V2, ..., V N}, and the related labels are {y1, y2, ..., y N},
[0021] The sentence vectors and structured data are processed through the infectious disease emerging recognition model to identify known classes, and the centroids of each known class are calculated respectively. l} is the centroid set of known classes; calculate the cosine similarity distance between the vectors in each class and the centroid:
[0022]
[0023]
[0024] According to the classification of each type of closed set, the threshold τ is selected d , identify the unknown class, and the calculation formula is as follows:
[0025]
[0026] Among them, 0≤i≤N; 0≤j≤l; C k Represents a known class, C k ∈{C1,C2,...,C l}; C l+1 It is an unknown type, that is, an emerging infectious disease.
[0027] Preferably, the emerging infectious disease identification method further comprises:
[0028] The new infectious disease identification model outputs a reconstruction error, and the tail of the reconstruction error distribution is modeled through an extreme value model, and the unknown classes identified by the new infectious disease identification model are estimated, and the known classes and unknown classes are further divided.
[0029] Preferably, the preprocessing includes: data deduplication, missing value processing, and outlier processing.
[0030] In a second aspect, the present invention provides an emerging infectious disease intelligent identification system based on generative open set identification technology, comprising:
[0031] A data acquisition and preprocessing module, used to acquire and preprocess electronic health record data to obtain standardized data, wherein the standardized data includes structured data and unstructured data;
[0032] The closed set recognition module is used to perform closed set recognition on structured data and unstructured data respectively, and determine the optimal word embedding method, encoder network and optimal classifier;
[0033] A model building module, used to determine a decoder network based on an encoder network, and to build an emerging infectious disease recognition model based on the encoder network, the decoder network, and the optimal classifier;
[0034] The open set recognition module is used to process unstructured data through the optimal word embedding method to obtain sentence vectors; the sentence vectors and structured data are open-set recognized through the emerging infectious disease recognition model to identify unknown classes as emerging infectious diseases.
[0035] Preferably, the emerging infectious disease identification system further includes:
[0036] The secondary division module is used to model the tail of the reconstruction error distribution through an extreme value model according to the reconstruction error output by the new infectious disease identification model, estimate the unknown class identified by the new infectious disease identification model, and further divide the known class into the unknown class.
[0037] In a third aspect, the present invention provides a storage medium storing a computer program for constructing an emerging infectious disease intelligent identification method based on generative open set recognition technology, wherein the computer program enables a computer to execute the emerging infectious disease identification method as described above.
[0038] In a fourth aspect, the present invention provides an electronic device, comprising:
[0039] one or more processors;
[0040] Memory; and
[0041] One or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the programs include methods for executing the emerging infectious disease identification method as described above.
[0042] (III) Beneficial effects
[0043] The present invention provides a method and system for intelligent identification of emerging infectious diseases based on generative open set identification technology. Compared with the prior art, it has the following beneficial effects:
[0044] The present invention develops an emerging infectious disease identification model for emerging infectious disease monitoring. The construction of the model includes a closed set classification stage and a new class identification stage. It can effectively identify emerging infectious diseases as unknown infectious diseases in the early stages of an outbreak, which is beneficial for relevant institutions to provide specific medical intervention measures in a timely manner in the early stages of an outbreak of emerging infectious diseases. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only 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.
[0046] Figure 1 A block diagram of an emerging infectious disease intelligent identification method based on generative open set identification technology according to an embodiment of the present invention;
[0047] Figure 2 This is the closed set classification stage of the embodiment of the present invention;
[0048] Figure 3This is the open set recognition stage of the embodiment of the present invention. DETAILED DESCRIPTION
[0049] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described. 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.
[0050] The embodiments of the present application provide a method and system for intelligently identifying emerging infectious diseases based on generative open set recognition technology, thereby solving the technical problem that the prior art cannot identify emerging infectious diseases and effectively identifying emerging infectious diseases.
[0051] The technical solution in the embodiment of the present application is to solve the above technical problems, and the overall idea is as follows:
[0052] The outbreak of infectious diseases poses a serious threat to human life and health and the global economy. Timely intervention with effective preventive measures can prevent its wider spread. Compared with known infectious diseases, emerging infectious diseases lack effective drug treatment in the early stages of an outbreak. Therefore, rapid and accurate identification of emerging infectious diseases is an important part of controlling the spread of diseases, improving clinical care and public health measures. Therefore, the embodiments of the present invention focus on the identification of emerging infectious diseases, helping medical staff and public health decision makers to determine whether it is an infectious disease that has not appeared before, determine whether relevant preventive measures need to be taken, and control the spread of emerging infectious diseases to the greatest extent possible.
[0053] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.
[0054] The embodiment of the present invention provides a method for intelligently identifying emerging infectious diseases based on generative open set recognition technology. Figure 1 As shown, the method includes:
[0055] S1. Acquire and preprocess electronic health record data to obtain standardized data, wherein the standardized data includes structured data and unstructured data;
[0056] S2, perform closed set recognition on structured data and unstructured data respectively, and determine the optimal word embedding method, encoder network and optimal classifier;
[0057] S3. Determine the decoder network based on the encoder network, and build an emerging infectious disease recognition model based on the encoder network, the decoder network and the optimal classifier;
[0058] S4. Process the unstructured data using the optimal word embedding method determined in S2 to obtain sentence vectors; perform open set recognition on the sentence vectors and structured data using the emerging infectious disease recognition model to identify unknown classes as emerging infectious diseases.
[0059] The embodiment of the present invention develops an emerging infectious disease identification model for emerging infectious disease monitoring. The construction of the model includes a closed set classification stage and a new class identification stage. It can effectively identify emerging infectious diseases as unknown infectious diseases in the early stages of an outbreak, which is beneficial for relevant institutions to provide specific medical intervention measures in a timely manner in the early stages of an outbreak of emerging infectious diseases.
[0060] The following is a detailed description of each step:
[0061] It should be noted that, in the embodiment of the present invention, it is assumed that D = {X1, X2, ..., X n} is the set of electronic health records, H = {C1, C2, ..., C k} is the associated label set. Suppose there are M disease types {1, 2, ..., M}. In the label H, the subset H1 = {C1, C2, ..., C l}∈H as the known class, and the remaining labels H2={C l+1 , C l+2 , ..., C k}∈H as the unknown class. The goal is to build a model based on open set recognition to identify the label H2 that does not exist in the training set by exploring the relationship between the electronic health record D and the label H1.
[0062] In one embodiment, S1, obtaining and preprocessing electronic health record data to obtain standardized data, wherein the standardized data includes structured data and unstructured data. Specifically, the following steps are performed:
[0063] S101. Obtain electronic health record data, including:
[0064] The electronic health record data comes from tertiary hospitals or open data sources. The embodiment of the present invention uses four types of diseases, namely pneumonia, lung cancer, chronic obstructive pulmonary disease, and tuberculosis, as experimental subjects.
[0065] S102. Preprocessing electronic health record data, specifically including:
[0066] Data duplication, missing value processing, outlier processing, etc. are performed to obtain normalized data.
[0067] Normalized data includes structured data and unstructured data. Structured data mainly includes patient information, physical examination and test results, and unstructured data includes chief complaint / current medical history, past history and examination results.
[0068] In one embodiment, S2, respectively, performs closed set recognition on structured data and unstructured data to determine the optimal word embedding method, encoder network and optimal classifier. Specifically, it includes:
[0069] Assume {X1, X2, ..., X N} are the electronic health records of N patients, and the relevant labels are {y1, y2, ..., y N}, The optimal word embedding method is used to obtain the sentence vector representation of the electronic health record, which is used as the input of the encoder. The latent variables are obtained through the hidden layer and input into the optimal classifier to achieve closed-set classification.
[0070] S21. Word embedding method. Use Tfidf, Word2vec, Glove, FastText, CNN and other word embedding methods to obtain sentence vectors of unstructured data, verify the impact of different word embedding methods on classification results, and select the optimal word embedding method as the optimal word embedding method in this paper.
[0071] S22, encoder network. The encoder structure uses the architecture of the encoder in DeepMicro, that is, an intermediate hidden layer is used, where the number of hidden units ranges from 32, 64, 128, 256 to 512, and the hidden layer dimension is set to 4, 8, and 16 units. Sentence vectors, structured data, and the names of known infectious diseases are input into these encoder structures respectively. A total of 15 different model structures are tested, and the architecture with the best classification effect is selected.
[0072] S23, optimal classifier. The classifiers use support vector machine, random forest, multi-layer perceptron, and classification network for classification. The loss function of the classification network is the binary cross entropy loss function, and the activation function uses the sigmoid function.
[0073] The closed set classification stage is as follows Figure 2 shown.
[0074] In one embodiment, S3, determining a decoder network based on the encoder network, and constructing an emerging infectious disease recognition model based on the encoder network, the decoder network and the optimal classifier. Specifically, it includes:
[0075] According to the encoder determined in step S2, a variational autoencoder model is generated. Specifically:
[0076] The VAE (Variational Autoencoder) model is a type of generative model. It has a small reconstruction error for known classes, but a large reconstruction error for unknown classes. It identifies unknown classes based on the size of the reconstruction error. The VAE model consists of an encoder, a decoder, and a loss function. The encoder is a neural network with parameters φ, the input is sample x and the output is hidden representation z. The decoder is a neural network with parameters θ, the input is hidden representation z and the output is the probability distribution of the sample. The loss function is as follows:
[0077]
[0078] Among them, q φ (z|x) is the approximate posterior, p θ (z) is the prior distribution of the potential representation z, p θ (x|z) is the probability of input x given the latent representation. The first term on the right side of the equation is the KL divergence between the approximate posterior and prior, which can be seen as a regularizer and is very close to the prior probability p θ The second term can be viewed as the reconstruction error.
[0079] The VAE architecture of the embodiment of the present invention uses the architecture in DeepMicro, which is a deep representation learning framework suitable for predicting individual representations of high-dimensional vectors.
[0080] In the training phase, the model is trained to minimize the reconstruction loss L r , KL divergence L KL and classification loss L c The sum of the predicted and true labels is used to measure the classification loss, using the input data x and the reconstructed data The reconstruction loss is measured by the L1 distance of the decoder output data.
[0081]
[0082]
[0083]
[0084] The overall loss function of the model is as follows:
[0085] L=-(L r +L KL +L c ).
[0086] In one embodiment, S4, the unstructured data is processed by the optimal word embedding method determined in S2 to obtain a sentence vector; the sentence vector and the structured data are open-set identified by the new infectious disease identification model to identify the unknown class as an emerging infectious disease. The open-set identification stage is shown in 3, and specifically includes:
[0087] The optimal word embedding method selected in S2 is used to process the unstructured data, and the sentence vector is represented as {V1, V2, ..., VN}, and the related labels are {y1, y2, ..., y N},
[0088] The structured data, sentence vectors and the names of known infectious diseases are used as the input data of the new infectious disease recognition model. The known classes are identified and the centroid of each known class is calculated. Let {Ce1, Ce2, ..., Ce l} is the centroid set of known classes. Calculate the cosine similarity distance between the vectors in each class and the centroid:
[0089]
[0090]
[0091] According to the classification of each type of closed set, the threshold τ is selected d , identify the unknown class.
[0092]
[0093] Among them, 0≤i≤N; 0≤j≤l; C k Represents a known class, C k ∈{C1,C2,...,C l}; C l+1 It is an unknown type, that is, an emerging infectious disease.
[0094] In the specific implementation process, it also includes S5, the new infectious disease identification model outputs the reconstruction error, models the tail of the reconstruction error distribution through the extreme value model, estimates the unknown class identified by S4, and further divides the known class and the unknown class.
[0095] The embodiment of the present invention reconstructs the tail of the error distribution for modeling, and achieves a better estimation of the tail data. The Fisher Tippett-Gnedenko theorem (FTG) and the Picklands-Balkema deHaan formula are usually used to model statistical extreme values. This study uses the Picklands-Balkema deHaan formula. For a sufficiently large threshold w, for a class distribution V, there are {V1, V2, ..., V n n independent and identically distributed samples can be approximated by the generalized Pareto distribution, denoted as
[0096]
[0097] Where Pr(vw≤v|v>w) is a conditional probability, which is fitted using the generalized Pareto distribution.V (w+v), F V (v) and F V (w) are all cumulative distribution functions. For all real numbers x, the cumulative distribution function is defined as follows: X (x) = P (X≤x).
[0098]
[0099] v in is a random variable, where is the cumulative distribution function of the generalized Pareto distribution, It represents the probability that the extreme value of the distribution V is less than a certain value r (r>w), which is used to determine whether a given test sample belongs to a known class. ζ and μ are the tail index parameter and scale parameter, which can be estimated from the given tail data using maximum likelihood estimation. The parameters tail size and threshold w need to be defined by yourself to divide known classes and unknown classes.
[0100] In order to verify the effectiveness of the embodiment of the present invention, the embodiment of the present invention verifies the proposed model through real electronic health records.
[0101] The data comes from a tertiary hospital in eastern China (the level of medical institutions is divided according to the provisions of China's current "Hospital Grading Management Measures", etc., and it is the highest level in the "three levels and six grades" classification of hospitals in mainland China). The hospital is a sentinel hospital for infectious disease monitoring.
[0102] The present invention obtained 1983 electronic medical record data from September 7, 2020 to November 28, 2022, and selected four types of diseases, pneumonia, lung cancer, chronic obstructive pulmonary disease, and tuberculosis, as experimental subjects. In the entire data set, there are 1121 pneumonia records, 38 lung cancer records, 218 chronic obstructive pulmonary disease records, and 197 tuberculosis records, totaling 1574 valid records.
[0103] Experiment 1:
[0104] The open set classification is compared by different methods. The results are shown in Table 1.
[0105] Herein, OpenEIDI refers to the method of an embodiment of the present invention.
[0106] Softmax means: calculate the softmax classification probability of each sample. If the maximum classification probability appears in the unknown category, or this maximum probability is lower than a specific threshold, it is classified as an unknown category.
[0107] Openmax means: calculate the openmax score of each sample and use SoftMax to map the probability of each category. If the maximum classification probability appears in the unknown category, or this maximum probability is lower than a specific threshold, it is classified as an unknown category.
[0108] DOC means: a 1-vs-rest sigmoids layer is constructed. When all predicted probabilities are less than their corresponding thresholds, they are classified as unknown classes. On this basis, the Gaussian fitting method is used to narrow the judgment range of the sigmoid function.
[0109] MLOSR means: multi-task training using encoder, classifier and decoder networks, EVT models the tail of the reconstruction error of known classes. When the probability that the reconstruction error comes from a known class is less than 0.5, the test sample is classified as an unknown class.
[0110] CGDL means: after the model training is completed, the information hidden in the potential representation and the reconstruction error are used to construct the unknown detector D, which is used as a binary classifier to identify unknown classes.
[0111] Table 1. Comparative experimental results of open set classification
[0112]
[0113] Note: * p<0.05, ** p<0.01, *** p<0.0001; the values in the table are the means of the 10-fold cross-validation results, and the standard deviations are enclosed in parentheses.
[0114] It can be seen from Table 1 that the emerging infectious disease identification model based on open set recognition according to the embodiment of the present invention has leading performance compared with the existing emerging infectious disease identification baseline method.
[0115] The embodiment of the present invention provides an emerging infectious disease intelligent identification system based on generative open set identification technology, including:
[0116] A data acquisition and preprocessing module, used to acquire and preprocess electronic health record data to obtain standardized data, wherein the standardized data includes structured data and unstructured data;
[0117] The closed set recognition module is used to perform closed set recognition on structured data and unstructured data respectively, and determine the optimal word embedding method, encoder network and optimal classifier;
[0118] A model building module, used to determine a decoder network based on an encoder network, and to build an emerging infectious disease recognition model based on the encoder network, the decoder network, and the optimal classifier;
[0119] The open set recognition module is used to process unstructured data through the optimal word embedding method to obtain sentence vectors; the sentence vectors and structured data are open-set recognized through the emerging infectious disease recognition model to identify unknown classes as emerging infectious diseases.
[0120] It can be understood that the emerging infectious disease intelligent identification system based on generative open set recognition technology provided in the embodiment of the present invention corresponds to the above-mentioned emerging infectious disease intelligent identification method based on generative open set recognition technology. The explanations, examples, beneficial effects and other parts of the relevant contents can refer to the corresponding contents in the emerging infectious disease intelligent identification method based on generative open set recognition technology, and will not be repeated here.
[0121] An embodiment of the present invention also provides a computer-readable storage medium, which stores a computer program for intelligent identification of emerging infectious diseases based on generative open set recognition technology, wherein the computer program enables a computer to execute the method for intelligent identification of emerging infectious diseases based on generative open set recognition technology as described above.
[0122] An embodiment of the present invention further provides an electronic device, including:
[0123] one or more processors;
[0124] Memory; and
[0125] One or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the programs include a method for executing the emerging infectious disease intelligent identification method based on generative open set recognition technology as described above.
[0126] In summary, compared with the prior art, the present invention has the following beneficial effects:
[0127] 1. The embodiment of the present invention develops an emerging infectious disease identification model for emerging infectious disease monitoring. The construction of the model includes a closed set classification stage and a new class identification stage. It can effectively identify emerging infectious diseases as unknown infectious diseases in the early stages of an outbreak, which is conducive to relevant institutions to provide specific medical intervention measures in a timely manner in the early stages of an outbreak of emerging infectious diseases.
[0128] 2. The embodiment of the present invention introduces a generative open set recognition method into the identification of new infectious diseases, and proposes a method combining VAE and cosine similarity distance to more accurately identify new infectious diseases.
[0129] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0130] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. 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 method for intelligent identification of emerging infectious diseases based on generative open set recognition technology, characterized in that: include: Acquire and preprocess electronic health record data to obtain normalized data, wherein the normalized data includes structured data and unstructured data; Performing closed set recognition on the structured data and the unstructured data respectively to determine the optimal word embedding method, encoder network and optimal classifier; Determine a decoder network based on the encoder network, and construct an infectious disease emerging recognition model based on the encoder network, the decoder network and the optimal classifier; The unstructured data is processed by the optimal word embedding method to obtain sentence vectors; the sentence vectors and structured data are open-set identified by the new infectious disease identification model to identify unknown classes as new infectious diseases.
2. The method for intelligent identification of emerging infectious diseases based on generative open set identification technology according to claim 1, characterized in that: Performing closed set recognition on the structured data and the unstructured data respectively, determining an optimal word embedding method, an encoder network, and an optimal classifier, including: Different word embedding methods are used to obtain sentence vectors of unstructured data, the effects of different word embedding methods on classification effects are verified, and the optimal word embedding method is selected as the optimal word embedding method; the word embedding methods include Tfidf, Word2vec, Glove, FastText and CNN; Adopt the encoder architecture in DeepMicro, that is, use an intermediate hidden layer, set different numbers of hidden units, set the hidden layer dimensions to different units, and choose the architecture with the best classification effect as the encoder network; Use different classifiers for classification, and use the one with the smallest loss function as the optimal classifier.
3. The method for intelligent identification of emerging infectious diseases based on generative open set identification technology according to claim 1, characterized in that: The determining a decoder network based on the encoder network comprises: Based on the encoder network, a variational autoencoder is used to generate the decoder network.
4. The method for intelligent identification of emerging infectious diseases based on generative open set identification technology according to claim 1, characterized in that: The unstructured data is processed by the optimal word embedding method to obtain sentence vectors; the sentence vectors and structured data are open-set identified by the new infectious disease identification model to identify unknown classes as new infectious diseases, including The optimal word embedding method is used to process unstructured data, and the sentence vector is represented as {V1, V2, …, V N }, and the relevant labels are {y1,y2,…,y N }, The sentence vector and structured data are processed through the new infectious disease recognition model to identify the known classes, and the centroid of each known class is calculated respectively. l } is the centroid set of known classes; calculate the cosine similarity distance between the vectors in each class and the centroid: According to the classification of each type of closed set, the threshold τ is selected d , identify the unknown class, and the calculation formula is as follows: Among them, 0≤i≤N; 0≤j≤l; C k Represents a known class, C k ∈{C1,C2,…,C l }; C l+1 It is an unknown type, that is, an emerging infectious disease.
5. The method for intelligent identification of emerging infectious diseases based on generative open set identification technology according to any one of claims 1 to 4, characterized in that: The emerging infectious disease identification method further comprises: The new infectious disease identification model outputs a reconstruction error, and the tail of the reconstruction error distribution is modeled through an extreme value model, and the unknown classes identified by the new infectious disease identification model are estimated, and the known classes and unknown classes are further divided.
6. The method for intelligent identification of emerging infectious diseases based on generative open set identification technology according to any one of claims 1 to 4, characterized in that: The preprocessing includes: data deduplication, missing value processing, and outlier processing.
7. An emerging infectious disease intelligent identification system based on generative open set recognition technology, characterized in that: include: A data acquisition and preprocessing module, used to acquire and preprocess electronic health record data to obtain standardized data, wherein the standardized data includes structured data and unstructured data; The closed set recognition module is used to perform closed set recognition on structured data and unstructured data respectively, and determine the optimal word embedding method, encoder network and optimal classifier; A model building module, used to determine a decoder network based on an encoder network, and to build an emerging infectious disease recognition model based on the encoder network, the decoder network, and the optimal classifier; The open set recognition module is used to process unstructured data through the optimal word embedding method to obtain sentence vectors; the sentence vectors and structured data are open-set recognized through the emerging infectious disease recognition model to identify unknown classes as emerging infectious diseases.
8. The emerging infectious disease intelligent identification system based on generative open set identification technology according to claim 7, characterized in that: The emerging infectious disease identification system also includes: The secondary division module is used to model the tail of the reconstruction error distribution through an extreme value model according to the reconstruction error output by the new infectious disease identification model, estimate the unknown class identified by the new infectious disease identification model, and further divide the known class into the unknown class.
9. A storage medium, characterized in that: It stores a computer program for constructing an emerging infectious disease intelligent identification method based on generative open set recognition technology, wherein the computer program enables a computer to execute the emerging infectious disease identification method as described in any one of claims 1 to 6.
10. An electronic device, characterized in that: include: one or more processors; Memory; as well as One or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the programs include programs for executing the emerging infectious disease identification method according to any one of claims 1 to 6.
Citation Information
Patent Citations
Open set recognition method, information processing equipment and storage medium
CN110110734A
Infectious disease monitoring and early warning method and device
CN112185586A
Image open set identification method based on comparison thought and integrated discrimination
CN115331055A
Open set identification method for urine red blood cells, white blood cells and calcium oxalate crystals
CN115471697A
Comparative incremental learning-based model training method and malicious traffic classification method and system
CN116244645A
