Patient relationship association and information matching method and device based on knowledge graph

Abstract

The present application belongs to the field of intelligent medical treatment, and particularly relates to a patient relationship association and information matching method and device based on a knowledge graph. The method comprises S1: obtaining patient treatment information records; S2: obtaining a patient accompanying person triple based on the treatment information records, the three elements of the patient accompanying person triple being a patient, an accompanying relationship and an accompanying person; and S3: matching the patient accompanying person triple with a treatment information knowledge graph to obtain a matching result. In the patient relationship association and information matching task, an auto-encoder algorithm based on adaptive spatial transformation projection is used to reduce the patient and accompanying person triple to a low-dimensional space, capture the nonlinear relationship of the data and eliminate redundant information, and through dynamic projection matrix optimization of the spatial transformation module, the problem of difficult semantic unification of the triple data is solved, and the accuracy of the matching is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent medical treatment, and more particularly, to a patient relationship association and information matching method, device, medium and program product based on a knowledge graph. BACKGROUND

[0002] The present application relates to the field of data processing, and more particularly, to a patient relationship association and information matching method based on a knowledge graph.

[0003] The background of the present application is based on the specific problems of the Tibetan population in medical services, namely the high degree of name overlap phenomenon caused by naming conventions, religious culture and language characteristics. This phenomenon causes difficulties in patient identity recognition in medical scenarios, especially when patients do not carry an ID card or other proof documents, it is difficult to quickly and accurately retrieve their historical medical records by name alone, directly affecting the continuity and accuracy of medical services.

[0004] Existing patient information retrieval methods based on names often fail when dealing with high name overlap situations. Patients may need to return home to obtain an ID card or other proof materials, resulting in increased time costs, and even possible delay of necessary medical services. In addition, in remote areas or emergency medical scenarios, the difficulty of obtaining these materials further exacerbates the problem. Simply relying on patient self-reported information can also cause unnecessary manual searching and operational burden due to communication barriers or input errors, increasing the pressure on medical institutions.

[0005] Chinese invention patent CN113688255B proposes a knowledge graph construction method based on Chinese electronic medical records. Most of the currently constructed knowledge graphs contain a small number of medical record corpus, and the size of the knowledge graph is not large, and is often only applicable to a single department or disease, with poor universality, while some well-developed medical record knowledge graphs require a lot of manual participation, which is time-consuming and laborious, and has poor scalability. Due to different descriptions of disease categories, the corresponding series of checks and treatments, and other language environments are different between different departments and diseases of electronic medical records, and the medical jargon corresponding to different disease categories is different. These characteristics make some deep learning methods less effective, and the knowledge graph construction framework is not easy to expand. In view of the above problems, a knowledge graph data analysis processing method based on Chinese electronic medical records, a corpus annotation process specification, and an entity relationship extraction scheme are developed.

[0006] The Chinese invention patent with publication number CN118629571B proposes a clinical test result auditing method and system based on artificial intelligence and big data, relating to the technical field of medical information processing, which includes: constructing a first knowledge graph, obtaining physiological condition characteristics and quality control information, historical test data and medication conditions, generating a heterogeneous data set, performing intelligent fusion, extracting semantic features and expanding to obtain high-quality sample data; obtaining real-time message data of the test equipment, performing security authentication, comparing adjacent sample data, generating sample correlation verification results, constructing an initial automatic auditing model and outputting automatic auditing results, comparing the automatic auditing results with manual approval results to obtain a high-precision automatic auditing model; inputting the sample data into the high-precision automatic auditing model, performing semantic reasoning, judging abnormal risks and outputting an intelligent auditing report, calculating the confidence and comparing it with the confidence threshold, and outputting the intelligent auditing report with a confidence higher than the confidence threshold as a trusted auditing result.

[0007] The Chinese invention patent with publication number CN119046421A proposes an intelligent question and answer system and method based on deep semantic understanding and knowledge graph fusion, which includes a receiving end, an analysis end and a reply end. The method includes the receiving end collecting questions about a specific disease raised by a user, the analysis end performing deep semantic understanding and knowledge graph mining operations on the questions based on an intelligent question and answer model to obtain relevant medical knowledge and data structures of the questions, and the reply end performing knowledge fusion based on the relevant medical knowledge and data structures to obtain a reply result and output it to the user. The present invention can solve the problem of lack of deep semantic understanding and knowledge graph fusion function in existing intelligent question and answer systems, improve the accuracy and response speed of the system, provide an efficient and reliable medical consultation channel for users, realize full-process automation from user query processing to reply output, reduce the risk of manual intervention, and promote the popularization and development of intelligent technology.

[0008] The existing technology has the following shortcomings:

[0009] 1. In the task of patient relationship association and information matching, the traditional patient information retrieval method based on name cannot cope with the high rate of name duplication, which easily leads to identity recognition failure and increases the additional cost of patients and medical institutions.

[0010] 2. In the task of patient relationship association and information matching, the existing method does not make full use of patient relationship information, relying only on simple text or basic attribute comparison, which makes it difficult to fully capture the association between patients, accompanying persons and family members, resulting in low matching accuracy.

[0011] 3. In the task of patient relationship association and information matching, there is a lack of effective verification mechanism, and the matching result cannot be further judged, which easily affects medical decision-making due to noise data or false information, reducing the credibility of the information. SUMMARY

[0012] In view of the above problems, the present application provides a patient relationship association and information matching method based on a knowledge graph, which utilizes hospital historical information to construct a knowledge graph, thereby realizing rapid matching of patient information.

[0013] The present application (first aspect) discloses a patient relationship association and information matching method based on a knowledge graph, comprising:

[0014] S1: obtaining patient visit information records;

[0015] S2: obtaining patient accompanying person triples based on the visit information records, wherein the three elements of the patient accompanying person triples are a patient, an accompanying relationship, and an accompanying person;

[0016] S3: matching the patient accompanying person triples with visit information knowledge graphs to obtain matching results; wherein the visit information knowledge graphs are obtained in the following manner:

[0017] Step 1: obtaining patient visit information record data sets;

[0018] Step 2: constructing knowledge graphs based on the visit information record data sets to obtain the visit information knowledge graphs.

[0019] Further, the method further comprises: S4: verifying the matching results to obtain verified matching results.

[0020] Further, the specific method of matching in S3 is: obtaining a triple vector after vectorizing the patient accompanying person triples, converting the triple vector to a low-dimensional space expression of the triple vector through an encoder, vectorizing all triple sets in the visit information knowledge graphs and inputting them into the encoder to convert to a low-dimensional space expression of knowledge graph triple vectors, and one-by-one matching the low-dimensional space expression of the triple vector with the low-dimensional space expression of the knowledge graph triple vector set to obtain a matching degree, taking the knowledge graph triple vector with the highest matching degree as the matching result;

[0021] Optionally, the encoder is a self-encoder of adaptive spatial transformation projection, and the construction method comprises:

[0022] Step 1: obtaining patient relationship triple training set data, labeling the true association relationship of the patient relationship triples, constructing an initial self-encoder with a multi-layer structure, and initializing the parameters of each layer;

[0023] Step 2: obtaining patient relationship triple vectors after vectorizing the patient relationship triples;

[0024] Step 3: The patient relationship triplet vector is transformed by a projection matrix to obtain a reduced low-dimensional space;

[0025] Step 4: The patient relationship triplet vector is input into a projection matrix adjusted autoencoder to obtain a compressed low-dimensional space;

[0026] Step 5: The parameters are updated by backpropagation through the loss function until the stopping condition is met, and the self-adaptive space transformation projection autoencoder is obtained, wherein the loss function is calculated based on the patient relationship triplet vector, the reduced low-dimensional space, the parameters and the output of the autoencoder;

[0027] Optionally, the patient relationship triplet vector is standardized and then input into the space transformation module and the autoencoder respectively.

[0028] Further, the standardization processing method is represented as:

[0029]

[0030] In the formula, μ a and Σ a are the mean vector and standard deviation vector of X a respectively; X′ a is the standardized patient triplet data.

[0031] Optionally, the space transformation module dimension reduction process is represented as:

[0032]

[0033] In the formula, is the reduced patient triplet data, X′ a is the standardized patient triplet data, and P r is a projection matrix dynamically optimized according to the characteristics of the input standardized patient triplet data.

[0034] Optionally, the projection matrix is calculated by principal component analysis;

[0035] Optionally, the standardized patient triplet data X′ a is subjected to covariance matrix decomposition, and the first n ty characteristic vectors of the obtained covariance matrix are selected to form the projection matrix;

[0036] Optionally, if the dimension of the standardized patient triplet data X′ a is n a1 × n a2 , then the dimension of the projection matrix is n a2 × n ty ;

[0037] Optionally, the projection matrix is iteratively optimized based on error back propagation learning to obtain an optimized projection matrix, and the spatial module dimension reduction is performed using the optimized projection matrix.

[0038] Optionally, the projection matrix is optimized using a gradient-based adaptive adjustment term, and the optimization mode is represented as:

[0039]

[0040] In the formula, is the projection matrix of the tth iteration; is the projection matrix of the (t+1)th iteration; η r is the learning rate of the projection matrix, is the gradient of the loss function of the autoencoder with respect to P r is the adjustment term; r is the adjustment term; is the local sensitivity gradient adjustment factor of the tth iteration; α r is the adjustment factor;

[0041] Optionally, α r is set to 0.2, and δ r is the information entropy of the projection matrix;

[0042] Optionally, the update mode of the local sensitivity gradient adjustment factor is represented as:

[0043]

[0044] In the formula, is the local sensitivity gradient adjustment factor of the (t+1)th iteration; η tre is the learning rate of the adjustment factor, represents the gradient of the loss function with respect to the projection matrix of the tth iteration, is the L2 norm of the gradient of the loss function with respect to the projection matrix, which is used to measure the change amplitude of the gradient.

[0045] Further, the network structure of the autoencoder includes multiple layers, the parameters of each layer include a weight matrix and a bias term, and the initialization of the parameters of each layer is represented as: let the number of layers of the autoencoder be L red is the initialization mode of the weight matrix and the bias term of the layer,

[0046]

[0047] In the formula, represents the weight matrix of the lth layer of the autoencoder, is the bias term of the lth layer of the autoencoder; ~ represents being subject to a specific distribution; represents a normal distribution with mean 0 and variance 0.01;

[0048] Optionally, the output of the autoencoder is represented as:

[0049]

[0050] wherein, is the activation output of the l-th layer of the encoder; is the activation output of the l-1-th layer of the encoder; and Sig() is a Sigmoid activation function; is the low estimation correction factor of the t-th iteration; represents the weight matrix of the l-th layer of the autoencoder, is the bias term of the l-th layer of the autoencoder;

[0051] Optionally, a dynamic adjustment method based on error feedback is used to continuously update the low estimation correction factor, and the updating method is represented as:

[0052]

[0053] wherein, is the low estimation correction factor of the t+1-th iteration; and η λ is the learning rate of the low estimation correction factor, and are the first and second order derivatives of the loss function of the autoencoder with respect to the low estimation correction factor of the t-th iteration, respectively;

[0054] Further, the loss function is represented as:

[0055]

[0056] The parameter updating rule of the autoencoder is:

[0057]

[0058] wherein, the vectorized patient triad data input to the autoencoder is X a , is the reduced dimension patient triad data, || ||1represents the L1 norm, and || ||2represents the L2 norm; η der is the learning rate of the autoencoder; and L r is the loss function of the autoencoder; is the activation output of the last layer of the encoder; is the weight matrix of the last layer of the autoencoder; and are the partial derivatives of the loss function of the autoencoder with respect to the weight and the bias, respectively; represents the gradient of the loss function with respect to the projection matrix of the t-th iteration; is the L2 norm of the gradient of the loss function with respect to the projection matrix; pes is a power adjustment parameter for controlling the amplification effect of the gradient; a red is a first weighting factor; β red is a second weighting factor.

[0059] Further, the relationship verification adopts a correlation vector machine algorithm based on adaptive feature learning for verification; the acquisition method of the correlation vector machine algorithm based on adaptive feature learning comprises:

[0060] Step 1: Obtain a training patient-accompanying person relationship data set, and label high confidence and low confidence;

[0061] Step 2: Initialize the kernel function parameters, learning rate and regularization parameter of the correlation vector machine;

[0062] Step 3: Perform vectorization on the patient data feature in the patient-accompanying person relationship data set to obtain a data feature vector, and perform feature mapping on the data feature vector to obtain a feature vector;

[0063] Step 4: Perform correlation vector machine training iteration based on the Bayesian framework in the feature vector space until the stopping condition is met to obtain the correlation vector machine algorithm based on adaptive feature learning;

[0064] Optionally, the feature mapping is represented as:

[0065] In the formula, q i is the i th patient data feature vector input into the correlation vector machine; q′ i is the mapped i th patient data feature vector; M() is a mapping function, and Θ qM is a parameter set required for mapping;

[0066] Optionally, the mapping function is adjusted by optimizing the following learning objective function, which is represented as:

[0067]

[0068] In the formula, w ij is a similarity weight based on q i and q j in the original feature space; q j is the j th patient data feature vector input into the correlation vector machine;

[0069] q′ i =M(q i ,Θ qM )

[0070] machine.

[0071] Optionally, the calculation of the mapping function is represented as:

[0072]

[0073] wherein, represents the sample set nearest to the i-th patient data feature vector input into the relevant vector machine, w ij is the contribution weight of the i-th patient data feature vector input into the relevant vector machine to the j-th patient data feature vector input into the relevant vector machine.

[0074] Optionally, the calculation of the contribution weight is represented as:

[0075]

[0076] wherein, || || is the L2 norm; q k is the k-th patient data feature vector input into the relevant vector machine;

[0077] Optionally, the relevant vector machine of the Bayesian framework selects the patient data point with the highest posterior probability as the support vector by calculating the posterior probability distribution, and the calculation is represented as:

[0078]

[0079] wherein, y is the label vector of the input patient data, for example, the label vector includes 0 and 1, corresponding to “high confidence” and “low confidence” respectively; p(y|q i ,θ qk ) represents the posterior probability distribution, representing the predicted probability based on the parameter θ qk ; θ qk is the kernel function parameter of the relevant vector machine.

[0080] Optionally, the relevant vector machine of the Bayesian framework calculates the energy entropy of the model and optimizes the parameters of the model in each iteration, and the calculation of the energy entropy is represented as:

[0081]

[0082] wherein, H q represents the energy entropy; Ncs is the number of samples input into the relevant vector machine in the current batch.

[0083] Optionally, the relevant vector machine of the Bayesian framework adaptively adjusts the model parameters, and the dynamic updating of the kernel function parameter is represented as:

[0084]

[0085] wherein, is the updated kernel function parameter of the relevant vector machine, γq is a learning rate of the relevance vector machine, represents a gradient of a kernel function parameter of the relevance vector machine.

[0086] Optionally, the gradient is calculated as:

[0087]

[0088] where y i is a label of i-th patient data.

[0089] The second aspect of the present application discloses a patient relationship association and information matching system based on a knowledge graph, comprising:

[0090] An acquisition module 201 is configured to acquire patient medical information records.

[0091] An extraction module 202 is configured to obtain a patient accompanying person triple based on the medical information records, wherein the three elements of the patient accompanying person triple are a patient, an accompanying relationship, and an accompanying person.

[0092] A matching module 203 is configured to match the patient accompanying person triple with a medical information knowledge graph to obtain a matching result, wherein the medical information knowledge graph is obtained in the following manner:

[0093] Step 1: Acquire a patient medical information record dataset.

[0094] Step 2: Construct a knowledge graph based on the medical information record dataset to obtain the medical information knowledge graph.

[0095] The third aspect of the present application discloses a computer device, comprising a memory and a processor, wherein the memory is configured to store program instructions, and the processor is configured to invoke the program instructions, and when the program instructions are executed, the steps of the above method are executed.

[0096] The fourth aspect of the present application discloses a computer readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0097] The fifth aspect of the present application discloses a computer program product, comprising a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0098] The present application has the following beneficial effects:

[0099] 1. In the patient relationship association and information matching task, the knowledge graph construction technology is adopted, the relationship data between the patient, the accompanying person and the family members are extracted to generate triples, the problem of difficult identification of the patient's identity caused by high overlap of the names of the Tibetan population is solved, the attribute storage and standardized processing of the relationship nodes are enhanced, the structured and organized nature of the personnel identity data is enhanced, and the basis is provided for subsequent rapid matching.

[0100] 2. In the patient relationship association and information matching task, the self-encoder algorithm based on adaptive spatial transformation projection is adopted, the patient and accompanying person triples are reduced to a low-dimensional space, the nonlinear relationship of the data is captured and the redundant information is eliminated, the dynamic projection matrix optimization of the spatial transformation module solves the problem of difficult semantic unification of the triples data, and the accuracy of the matching is improved.

[0101] 3. In the patient relationship association and information matching task, the related vector machine algorithm based on adaptive feature learning is adopted to verify the relationship of the matching result, the multi-dimensional features such as co-occurrence frequency, regional similarity and language rules are used to filter high-confidence relationships, the problem of high misjudgment rate of simple name similarity judgment in actual scene is overcome, and the reliability of the matching is improved. BRIEF DESCRIPTION OF DRAWINGS

[0102] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0103] Figure 1 is a method flowchart provided by the first aspect of the embodiment of the present application;

[0104] Figure 2 is a program product schematic diagram provided by the second aspect of the embodiment of the present application;

[0105] Figure 3 is a schematic diagram of a computer device provided by the embodiment of the present application;

[0106] Figure 4 is a schematic diagram of the architecture of an exemplary computing device provided by the embodiment of the present application;

[0107] Figure 5 is a schematic diagram of a storage medium provided by the embodiment of the present application;

[0108] Figure 6 is a schematic diagram of a patient triple and knowledge graph patient relationship association and information matching provided by the embodiment of the present application. DETAILED DESCRIPTION

[0109] In order to make the person skilled in the art better understand the technical scheme of the present application, the technical scheme in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application.

[0110] In some of the processes described in this specification and in the accompanying drawings, multiple operations are described in a specific order. However, it should be clear that unless otherwise specifically stated, these operations can be performed in any order, or in parallel, and the order of the operations should not be construed as a requirement. The serial numbers of the operations, such as S101, S102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes can include more or fewer operations, and the operations can be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this paper are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence, nor do "first" and "second" represent different types.

[0111] The technical scheme in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0112] Figure 1 It is a kind of method flow diagram based on the evaluation of patient relationship association and information matching based on knowledge graph provided by the embodiment of the present application, specifically, the method comprises the following steps:

[0113] S101: obtaining patient medical information record;

[0114] S102: obtaining patient accompanying medical triad based on the medical information record, the three elements of the patient accompanying medical triad are patient, accompanying relationship and accompanying person;

[0115] S103: matching the patient accompanying medical triad with the medical information knowledge graph to obtain a matching result; wherein the medical information knowledge graph is obtained in the following way:

[0116] Step 1: Obtain patient medical information record dataset;

[0117] Step 2: Construct the medical information knowledge graph based on the medical information record dataset.

[0118] The present application aims to solve the problem of patient identity recognition difficulty caused by high name coincidence in the medical treatment process of Tibetan population. In the Tibetan population, the phenomenon of high name coincidence is determined by specific naming tradition, religious culture influence and language characteristics. For example, the naming method of Buddhist terminology, auspicious words or family tradition is easy to cause name duplication. This phenomenon is particularly prominent in the medical scene. When the patient does not carry an identity card or other identity documents, the name alone cannot quickly and accurately retrieve the historical medical records, which brings significant challenges to the continuity and accuracy of medical services.

[0119] The traditional name-based retrieval method often fails due to the name duplication phenomenon, which may cause the patient to return home to take identity cards and other proof materials, not only increasing the time cost of the patient, but also delaying medical services, especially in remote areas or emergency medical scenarios. In addition, relying solely on patient self-reports may result in unnecessary manual searching and communication costs, further increasing the operational burden of medical institutions.

[0120] Therefore, the present application proposes a patient relationship association and information matching method based on knowledge graph Figure 6 The main steps are as follows:

[0121] 1. Construct a personnel relationship knowledge graph of all patients

[0122] In the process of constructing the personnel relationship knowledge graph, first, data collection and preprocessing are performed;

[0123] The data is collected from the hospital's electronic health record system to obtain patient name, identity card number, accompanying person name, family member name, contact information, etc.

[0124] Further, data cleaning is performed, including: removing redundant records (such as repeated information of the same patient entered multiple times), handling multi-syllable words, pinyin or misspelled words, etc. All names are unified into a standard format (such as full spelling or regional language form) to ensure data consistency.

[0125] Further, extract the personnel relationship related to each patient and generate the triples of the knowledge graph. The generation rule of the triples is:

[0126] Patient and accompanying person: (patient name, relationship, accompanying person name), relationship is "accompanying person"

[0127] Patient and family member: (patient name, relationship, family member name), relationship is "family member";

[0128] Accompanying person and family member: (family member name, relationship, accompanying person name), relationship is "contact person".

[0129] It should be noted that in the generation of triples, each triple should be able to uniquely describe a node and its relationship to ensure the accuracy of subsequent graph matching.

[0130] Further, the generated triples are stored using a graph database (such as Neo4j), and the database structure is as follows:

[0131] Node: represents patients, accompanying persons, family members, etc., attributes include name, identity (patient / accompanying person / family member), contact information, etc.

[0132] Edge: represents the relationship between two nodes, types include "accompanying person", "family member", etc.

[0133] For clarity, in one embodiment, the Tibetan name is simplified, and the constructed triples are as follows:

[0134] (Zhashi, accompanying person, Ciren Danzeng)

[0135] (Zhashi, family member, Zhuoma)

[0136] (Zhuoma, contact person, Ciren)

[0137] (Lam, family member, Ciren)

[0138] (Dorje, accompanying person, Rinchen)

[0139] It should be noted that the complexity, length, and frequency of rare words of Tibetan names are higher than the example data in this embodiment.

[0140] In an ideal case, the knowledge graph in the graph database is continuously accumulated and updated, and eventually forms a complete knowledge graph of the hospital's patient personnel relationships.

[0141] 2. Construct triples for the current patient and accompanying person

[0142] In the case where the patient does not carry an ID card, the hospital registration system needs to collect the patient's basic information, provided by the patient or accompanying person, including:

[0143] Patient name: the name of the current patient;

[0144] Accompanying person's name: if there is an accompanying person, collect their name;

[0145] Relationship description: mark the relationship between the accompanying person and the patient (such as accompanying person, family member, etc.).

[0146] Further, according to the collected information, generate triples related to the patient, in one embodiment, the generated triples are as follows:

[0147] (Zhashi, accompanying person, Ciren)

[0148] 3, Triple matching

[0149] The triple formed by the patient and the accompanying person is mapped into a low-dimensional space by the autoencoder, so as to be efficiently matched with the triples in the knowledge graph;

[0150] Each part of the triple, including two nodes (patient name, accompanying person name) and relationship type, is subjected to vectorization processing to adapt to the input requirements of the autoencoder, and the application adopts a Word2Vec algorithm for vectorization processing of the text, the Word2Vec algorithm being a commonly used vectorization algorithm in the field, which scans the text to be vectorized according to a preset large-scale corpus, and represents each word as a one-hot encoding vector, the dimension of the one-hot encoding vector being equal to the size of the vocabulary in the corpus.

[0151] Further, the encoder of the autoencoder compresses the high-dimensional vector into a low-dimensional space to extract core semantic features;

[0152] And the decoder of the autoencoder restores the original triple from the low-dimensional space, ensuring that the model understands the relationship semantics.

[0153] The autoencoder adopts the existing patient relationship triples in the hospital as training data for model training, and labels the real association relationship in the historical triples.

[0154] The application adopts an autoencoder algorithm based on adaptive space transformation projection, which finds the optimal expression of the vectorized patient triple data in the low-dimensional space, and uses the projection matrix generated by the space transformation module to adjust each layer in the encoding process, so that the processing process of the autoencoder on the vectorized patient triple data is more adaptive to the internal structure of the data, effectively removes redundant information and better captures the nonlinear relationship of the vectorized patient triple data.

[0155] Specifically, the training process of the autoencoder algorithm based on adaptive space transformation projection is as follows:

[0156] 1) Construct a multi-layer structure of the autoencoder and initialize the parameters of each layer, including the weight matrix and the bias term, and let the number of layers of the autoencoder be L red The initialization mode of the weight matrix and the bias term of the l-th layer of the autoencoder is represented as:

[0157]

[0158] In the formula, Wl represents the weight matrix of the l-th layer of the autoencoder, bl represents the bias term of the l-th layer of the autoencoder; ~ represents being subject to a specific distribution; N (0, 0.01) represents a normal distribution with a mean of 0 and a variance of 0.01.

[0159] 2) The vectorized patient triad data input into the autoencoder is first standardized to eliminate the influence of different dimensions on training, and let the vectorized patient triad data input into the autoencoder be X a The standardization method is represented as:

[0160]

[0161] In the formula, μ a and Σ a are the mean vector and standard deviation vector of X a , respectively; and X′ a is the standardized patient triad data.

[0162] 3) The standardized patient triad data is input into the spatial transformation module, which learns an adaptive projection matrix P r for mapping the high-dimensional standardized patient triad data to a low-dimensional space, and the dimension reduction process is represented as:

[0163]

[0164] In the formula, is the dimension-reduced patient triad data, and P r is a projection matrix dynamically optimized according to the characteristics of the input standardized patient triad data.

[0165] The projection matrix is learned based on the error back propagation method, and the initialization generation method of the projection matrix is to calculate the projection matrix by using the principal component analysis method. Specifically, first, the standardized patient triad data X′ a is decomposed into a covariance matrix, and the first n ty characteristic vectors of the obtained covariance matrix are selected to form the projection matrix. If the dimension of the standardized patient triad data X′ a is n a1 × n a2 , then the dimension of the projection matrix is n a2 × n ty .

[0166] To enhance the projection ability, an adaptive adjustment term based on gradient is used to optimize the projection matrix, and the optimization method is represented as:

[0167]

[0168] In the formula, is the projection matrix of the tth iteration; is the projection matrix of the t+1th iteration; η r is the learning rate of the projection matrix, is the gradient of the loss function of the autoencoder with respect to P r is the gradient of the loss function of the autoencoder with respect to P r is the adjustment term; is the local sensitivity gradient adjustment factor of the tth iteration; α r is the adjustment factor. Preferably, α r is set to 0.2, and δ r is the information entropy of the projection matrix.

[0169] Further, the role of the local sensitivity gradient adjustment factor is to dynamically adjust the learning rate of the projection matrix in each training, so that the model can make appropriate local adjustment and amplification when facing high noise or features with high sensitivity, to avoid information loss, and the update mode is represented as:

[0170]

[0171] In the formula, is the local sensitivity gradient adjustment factor of the t+1th iteration; η tre is the learning rate of the adjustment factor, represents the gradient of the loss function with respect to the projection matrix of the tth iteration, is the L2 norm of the gradient of the loss function with respect to the projection matrix, used to measure the change amplitude of the gradient. Preferably, η tre is set to 0.05.

[0172] 4) The autoencoder will encode the normalized patient triad data through a plurality of neurons, and gradually compress it to a low-dimensional space. In this process, the latent representation of the patient triad data is learned by minimizing the reconstruction error, that is, the optimal expression of the patient triad data in the low-dimensional space is found. The projection matrix generated by the space transformation module will adjust each layer in the encoding process, so that the dimension reduction process is more suitable for the internal structure of the patient triad data. In the dimension reduction process, a low estimation adjustment optimization strategy is adopted, and the output features are weighted and corrected in the dimension reduction process. The output of the encoder is represented as:

[0173]

[0174] In the formula, is the activation output of the lth layer of the encoder; is the activation output of the l-1th layer of the encoder; Sig() is the Sigmoid activation function; is the low estimation correction factor of the tth iteration.

[0175] 5) The role of the low estimation correction factor is to prevent information loss in the dimension reduction process. The low estimation correction factor is continuously updated by the dynamic adjustment method based on error feedback, and the update mode is represented as:

[0176]

[0177] wherein, is the low estimate correction factor of the tth iteration; η λ is the learning rate of the low estimate correction factor, and are the first and second order derivatives of the autoencoder loss function with respect to the low estimate correction factor of the tth iteration, respectively. Preferably, η λ is set to 0.3.

[0178] 6) During the training process, the autoencoder adjusts the parameters to optimize the dimensionality reduction effect by using the back propagation algorithm, and adopts the norm constraint to sparsity of the features, which not only optimizes the reconstruction error, but also promotes the sparsity of the features, thereby improving the effectiveness of the patient triplet data representation after dimensionality reduction. The loss function is calculated as follows:

[0179]

[0180] and the parameter update rule of the autoencoder is:

[0181]

[0182] wherein, || ||1represents the L1 norm; || ||2represents the L2 norm; η der is the learning rate of the autoencoder; L r is the loss function of the autoencoder; is the activation output of the last layer of the encoder; is the weight matrix of the last layer of the autoencoder; and are the partial derivatives of the loss function of the autoencoder with respect to the weight and bias, respectively; represents the gradient of the loss function with respect to the projection matrix of the tth iteration; is the L2 norm of the gradient of the loss function with respect to the projection matrix; pes is a power adjustment parameter for controlling the amplification effect of the gradient; α red is the first weighting factor; β red is the second weighting factor. Preferably, η der is set to 0.01, α red is set to 0.3, and β red is set to 0.7.

[0183] 7) Repeat the above steps until the preset stopping iteration condition is met, i.e., the model training is completed. In an embodiment, the preset stopping iteration condition is to reach a preset maximum number of iterations, and preferably, the preset maximum number of iterations is set to 1000 times.

[0184] After the autoencoder model is trained, the triplets input by the current patient are reduced to the low-dimensional vector space by the trained autoencoder, and the obtained low-dimensional features are used as input triplet features.

[0185] Further, the similarity between the input triplets and the graph triplets is calculated using cosine similarity or Euclidean distance, and high similarity indicates that the input triplets have a high matching possibility in the knowledge graph.

[0186] Further, the triplets are sorted according to the similarity, and the most similar triplets are selected as candidates, and the patient node information corresponding to the candidate triplets is returned.

[0187] Input vs (Zhashi, accompanying person, Tsering Danzeng): similarity = 0.98 (highest match);

[0188] Input vs (Zhashi, family member, Zhuoma): similarity = 0.92 (higher match);

[0189] Input vs (Dorje, accompanying person, Rinchen): similarity = 0.45 (low match);

[0190] Input vs (Lham, family member, Tsering): similarity = 0.78 (general match).

[0191] 4. Relationship verification

[0192] The matching result with the highest similarity is verified, and the relationship between the patient and the accompanying person or family member is verified to be credible, thereby improving the accuracy of the knowledge graph matching.

[0193] The present application adopts a related vector machine algorithm based on adaptive feature learning for verification, and the related vector machine algorithm is used as a classifier model, and the training data sources mainly include related information of patients and accompanying persons, family members, etc. in the hospital electronic health record system, relationship data in the historical medical record, and patient-family-accompanying person triplet data already existing in the knowledge graph;

[0194] The data features cover multi-dimensional information, including historical medical features (such as the frequency and the last co-occurrence time of the patient and the accompanying person), geographical features (such as the similarity of the addresses of the two), relationship type features (whether the provided relationship conforms to the logic, such as family member, accompanying person, etc.), demographic features (such as whether the age difference between the patient and the accompanying person is reasonable), language features (the language consistency of the name, such as the naming rules of the Tibetan phonetic alphabet), and data source features (such as whether the record is manually input or system generated).

[0195] Further, by manually annotating data, a dataset containing high and low confidence relationships is constructed, where positive samples represent known high confidence relationships, such as records of patients with real family members or accompanying persons, and negative samples represent low confidence relationships, such as incorrect matches or fake data. After each sample is vectorized, it is represented in the form of a feature vector, for example, containing co-occurrence times, address similarity, relationship type, and manually annotated confidence labels such as "high confidence" or "low confidence".

[0196] The adaptive feature learning-based relevance vector machine algorithm used as the classification algorithm is a sparse learning method based on Bayesian inference. In the traditional relevance vector machine algorithm, adaptive feature learning is used, and the classification decision boundary is optimized by combining an energy entropy model, so that the classifier not only reflects the local structural characteristics of the patient data, but also effectively captures the global distribution of the patient data.

[0197] Specifically, the training process of the adaptive feature learning-based relevance vector machine algorithm is as follows:

[0198] 1) Initialize the kernel function parameters of the relevance vector machine and the basic configuration of the model, including the learning rate and the regularization parameter, and the initialization method is represented as:

[0199]

[0200] γ q =0.01

[0201] λ q =0.1

[0202] In the formula, θ qk represents the parameter of the kernel function, is the initialization variance, γ q is the learning rate, and λ q is the regularization parameter; ~ indicates that it is subject to a certain distribution; represents a normal distribution.

[0203] 2) Process the input patient data feature vector, so that each data feature vector is mapped to a new feature space to obtain a new feature vector, which can better reflect the internal structure and correlation of the features. The mapping method is represented as:

[0204] q′ i =M(q i ,Θ qM )

[0205] In the formula, q i is the i-th patient data feature vector input to the relevance vector machine; q′ iis the i-th mapped patient data feature vector; M() is the mapping function, Θ qM is the set of parameters required for mapping.

[0206] Further, the mapping function is adjusted by optimizing the following learning objective function, denoted as:

[0207]

[0208] where w ij is the similarity weight based on q i and q j in the original feature space; q j is the j-th patient data feature vector input to the relevance vector machine.

[0209] Further, the mapping function is adjusted by optimizing the following learning objective function, denoted as:

[0210]

[0211] where denotes the set of nearest neighbors of the i-th patient data feature vector input to the relevance vector machine, w ij is the contribution weight of the i-th patient data feature vector input to the relevance vector machine to the j-th patient data feature vector input to the relevance vector machine.

[0212] Further, the mapping function is adjusted by optimizing the following learning objective function, denoted as:

[0213]

[0214] where || || is the L2 norm; q k is the k-th patient data feature vector input to the relevance vector machine.

[0215] 3) In the new feature space, the relevance vector machine based on the Bayesian framework is used for training, and the patient data point with the highest posterior probability is selected as the support vector by calculating the posterior probability distribution, and the calculation method is represented as:

[0216]

[0217] where y is the label vector of the input patient data, for example, the label vector includes 0 and 1, corresponding to "high confidence" and "low confidence" respectively; p(y|q i , θ qk ) represents the posterior probability distribution, representing the predicted probability based on the parameter θ qk ; θ qk is the kernel function parameter of the relevance vector machine.

[0218] 4) In each iteration, the energy entropy of the model is calculated and the parameters of the model are optimized, and the calculation method of the energy entropy is represented as:

[0219]

[0220] wherein H q represents the energy entropy; Ncs is the number of samples of the current batch input to the relevant vector machine.

[0221] 5) The model parameters are adaptively adjusted, and in an embodiment, the dynamic updating method of the kernel function parameters is represented as:

[0222]

[0223] wherein is the updated kernel function parameter of the relevant vector machine, γ q is the learning rate of the relevant vector machine, represents the gradient of the kernel function parameter of the relevant vector machine.

[0224] Further, the calculation method of the gradient is represented as:

[0225]

[0226] wherein y i is the label of the i-th patient data.

[0227] 6) Repeat the above steps until the preset stopping iteration condition is met, that is, the model training is completed. In an embodiment, the preset stopping iteration condition is that the preset maximum number of iterations is reached, and preferably, the preset maximum number of iterations is set to 500.

[0228] After the relevant vector machine training based on adaptive feature learning is completed, for each new patient-accompanying record, the feature vector is extracted and input into the trained relevant vector machine for classification, and "high confidence relationship" or "low confidence relationship" is output.

[0229] Further, the patient relationship data of the high confidence relationship is sent to the doctor to assist in judging the true information of the patient.

[0230] Figure 3 is a schematic diagram of a computer device provided by an embodiment of the present application, as Figure 3 shown, the device 2000 can include one or more processors 2010 and one or more memories 2020; wherein the memory has computer readable code stored therein, and the computer readable code can execute the method as described above when run by the one or more processors.

[0231] The processor in this embodiment can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, operations, and logic block diagrams disclosed in this embodiment. The general-purpose processor can be a microprocessor or any conventional processor, and can be based on an x86 or ARM architecture.

[0232] In general, the various exemplary embodiments of this disclosure can be implemented in hardware or dedicated circuitry, software, firmware, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device. When aspects of embodiments of this disclosure are illustrated or described as block diagrams, flowcharts, or using some other graphical representation, it will be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as non-limiting examples in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0233] For example, the method or apparatus according to embodiments of this disclosure can also be used by means of Figure 4 The architecture of the computing device 3000 shown is used for implementation. For example... Figure 4 As shown, the computing device 3000 may include a bus 3010, one or more CPUs 3020, a read-only memory (ROM) 3030, a random access memory (RAM) 3040, a communication port 3050 connected to a network, an input / output component 3060, a hard disk 3070, etc. The storage devices in the computing device 3000, such as the ROM 3030 or the hard disk 3070, may store various data or files used for processing and / or communication of the methods provided in this disclosure, as well as program instructions executed by the CPU. The computing device 3000 may also include a user interface 3080. Of course, Figure 4 The architecture shown is merely exemplary and can be omitted as needed when implementing different devices. Figure 4 One or more components in the computing device shown.

[0234] This invention also includes a computer-readable storage medium, such as... Figure 5As shown, the storage medium 4000 provided by the embodiment of the present application is a schematic diagram of the storage medium 4000, and the computer readable instructions 4010 are stored on the computer storage medium 4020. When the computer readable instructions 4010 are run by the processor, the method according to the embodiment of the present application can be executed as described above with reference to the accompanying drawings. The computer readable storage medium in the embodiment of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. The non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM) used as an external cache. By way of example and not limitation, many forms of RAM can be used, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DRAM). It should be noted that the memory of the method described herein is intended to include but not limited to these and any other suitable types of memory. It should be noted that the memory of the method described herein is intended to include but not limited to these and any other suitable types of memory.

[0235] The embodiment of the present application also provides a computer program product or computer program which, when executed by a processor, implements the steps of the above method. Figure 2 As shown, the computer program product or computer program includes:

[0236] The acquisition module 201 is configured to acquire patient visit information records.

[0237] The extraction module 202 is configured to obtain patient accompanying person triples based on the visit information records, and the three elements of the patient accompanying person triples are patient, accompanying relationship and accompanying person.

[0238] The matching module 203 is configured to match the patient accompanying person triples with a visit information knowledge graph to obtain a matching result. The visit information knowledge graph is obtained in the following manner:

[0239] Step 1: Acquire patient visit information record data set.

[0240] Step 2: Construct a knowledge graph based on the visit information record data set to obtain the visit information knowledge graph.

[0241] It should be noted that the flowchart and block diagrams in the drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustrations, and combinations thereof, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and

[0242] In general, various example embodiments of the present disclosure can be implemented in hardware or special-purpose circuits, software, firmware, logic, or any combination thereof. Some aspects can be implemented in hardware, while other aspects can be implemented in firmware or software which can be executed by a controller, microprocessor or other computing device, Although various example embodiments of the disclosure can be illustrated and described as block diagrams, flow charts, or using some other pictorial representation, it is well understood that these blocks, apparatus, systems, techniques or methods described herein can be implemented in, as non-limiting examples, hardware, software, firmware, special purpose circuits or logic, general purpose hardware or controler or other computing devices, or some combination thereof.

[0243] The specific operations disclosed herein can be implemented as one or more computer programs or software modules being executed by one or more processors. Each computer program or software module can include an operating system that is executed by the one or more processors. The operating system can be stored in any apparatus's memory that is accessible by the processor(s), including, for example, the above-described memory. The operating system can include procedures or routines for handling various basic operations. The operating system can also include procedures for handling communication between the apparatus and one or more other apparatuses. The operating system can also include procedures for handling communication between the apparatus and one or more user input devices. The operating system can also include procedures for handling communication between the apparatus and one or more user output devices.

[0244] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the division of the above-described device embodiment is only a logical function division, and there can be another division manner for actual implementation, for example, multiple devices or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different parts can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrically, mechanically or in other forms.

[0245] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0246] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0247] The example embodiments of the present disclosure described in detail above are only illustrative, not limiting. Those skilled in the art should understand that various modifications and combinations of these embodiments or their features can be made without departing from the principles and spirits of the present disclosure, and such modifications should fall within the scope of the present disclosure.

Claims

1. A knowledge graph-based patient relationship association and information matching method, characterized in that, The method comprises: S1: obtaining patient visit information records; S2: obtaining patient accompanying person triplets based on the visit information records, the three elements of the patient accompanying person triplets being patient, accompanying relationship, and accompanying person; S3: obtaining a triplet vector after vectorizing the patient accompanying person triplets, and converting the triplet vector into a low-dimensional space expression of the triplet vector through an encoder; vectorizing all triplet sets in a visit information knowledge graph and inputting them into the encoder to convert the low-dimensional space expression of the knowledge graph triplet vector set; matching the low-dimensional space expression of the triplet vector with the low-dimensional space expression of the knowledge graph triplet vector set to obtain a matching degree, and taking the knowledge graph triplet vector with the highest matching degree as a matching result; wherein the visit information knowledge graph is obtained in the following manner: Step 1: obtaining a patient visit information record data set; Step 2: constructing a knowledge graph based on the visit information record data set to obtain the visit information knowledge graph; The encoder is a self-encoder of adaptive spatial transformation projection, and the construction method comprises: Step 1: obtaining a patient relationship triplet training set data, and labeling the true correlation of the patient relationship triplets; constructing an initial self-encoder with a multi-layer structure, and initializing the parameters of each layer; Step 2: obtaining a patient relationship triplet vector after vectorizing the patient relationship triplets; Step 3: obtaining a low-dimensional space after the patient relationship triplet vector is transformed by a projection matrix; Step 4: inputting the patient relationship triplet vector into a self-encoder adjusted by a projection matrix to obtain a compressed low-dimensional space; Step 5: updating the parameters through a loss function until a stop condition is met to obtain the self-encoder of adaptive spatial transformation projection, the loss function being calculated based on the patient relationship triplet vector, the low-dimensional space, the parameters and output of the self-encoder; The standardized patient triad data is input into a spatial transformation module and the autoencoder respectively after the patient relationship triad vector is standardized; the network structure of the autoencoder comprises multiple layers, the parameters of each layer comprise a weight matrix and a bias term, and the initialization of the parameters of each layer is represented as: let the number of layers of the autoencoder be The initialization mode of the weight matrix and the bias term of the layer is represented as: wherein denotes the weight matrix of the first layer of the autoencoder, denotes the bias term of the first layer of the autoencoder; denotes being subject to a certain distribution; denotes a normal distribution with mean 0 and variance 0.

01. The output of the self-encoder is represented as: wherein is the activation output of the first layer of the encoder; is the activation output of the first layer of the encoder; is a Sigmoid activation function; is the low estimate correction factor of the th iteration; is the weight matrix of the th layer of the autoencoder, is the bias term of the th layer of the autoencoder; the low estimate correction factor is constantly updated by using a dynamic adjustment method based on error feedback, and the updating method is represented as: In the formula, The first The underestimation correction factor for the t-th iteration; To underestimate the learning rate of the correction factor, and These are the autoencoder loss functions with respect to the th The first and second derivatives of the underestimation correction factor in the next iteration. 2.The knowledge graph based patient relationship association and information matching method according to claim 1, characterized in that, The method further comprises: S4: verifying the matching result to obtain a verified matching result. 3.The knowledge graph based patient relationship association and information matching method of claim 1, wherein, The standardization processing mode is represented as: wherein, represents a patient relationship triplet vector, and are the mean vector and standard deviation vector of respectively; is the normalized patient triplet data. 4.The knowledge graph based patient relationship association and information matching method of claim 1, wherein, The spatial transform module dimensionality reduction process is represented as: or ; wherein, is the reduced dimensionality patient triad data, represents the patient relationship triad vector, is the standardized patient triad data, is a projection matrix dynamically optimized according to the input standardized patient triad data characteristics. 5.The knowledge graph based patient relationship association and information matching method of claim 1, wherein, The projection matrix is calculated by principal component analysis. 6.The knowledge graph based patient relationship association and information matching method of claim 4, wherein, The standardized patient triad data The covariance matrix is decomposed, and the obtained covariance matrix selects the first characteristic vector to form a projection matrix; if the dimension of the standardized patient triad data is , the dimension of the projection matrix is . 7.The knowledge graph based patient relationship association and information matching method of claim 4, wherein, The projection matrix is optimized based on error back propagation learning iteration to obtain an optimized projection matrix, and the low-dimensional space is obtained using the optimized projection matrix. 8.The knowledge graph based patient relationship association and information matching method of claim 7, wherein, The projection matrix is optimized using a gradient-based adaptive adjustment term, and the optimization mode is represented as: In the formula, For the first The projection matrix of the next iteration; For the first The projection matrix of the next iteration; The learning rate is the projection matrix. The loss function of the autoencoder is a pair gradient, It is an adjustment item; For the first The local sensitivity gradient adjustment factor for the next iteration; It is a regulatory factor. 9.The knowledge graph based patient relationship association and information matching method of claim 8, wherein, is set to 0.2, is the information entropy of the projection matrix. 10.The knowledge graph based patient relationship association and information matching method of claim 8, wherein, The update mode of the local sensitivity gradient adjustment factor is represented as: wherein is the local sensitivity gradient adjustment factor for the th iteration; is the local sensitivity gradient adjustment factor for the th iteration; is the learning rate of the adjustment factor, denotes the gradient of the loss function with respect to the projection matrix for the th iteration, is the L2-norm of the gradient of the loss function with respect to the projection matrix. 11.The knowledge graph based patient relationship association and information matching method of claim 8, wherein, The loss function is represented as: wherein, represents a patient relation triple vector, is the dimension-reduced patient triple data, represents an L1 norm; represents an L2 norm; is a first weighting factor; is a second weighting factor; is a weight matrix of the last layer of the autoencoder; is an activation output of the last layer of the encoder.

12. The knowledge graph based patient relationship association and information matching method according to claim 11, characterized in that, The parameter update rule of the self-encoder is: wherein, is a weight matrix of the self-encoder l-th layer; is a learning rate of the self-encoder; is a loss function of the self-encoder; and are partial derivatives of the loss function of the self-encoder with respect to the weights and biases, respectively; denotes the gradient of the loss function with respect to the projection matrix at the -th iteration; is the L2-norm of the gradient of the loss function with respect to the projection matrix; is a power adjustment parameter. 13.The knowledge graph based patient relationship association and information matching method of claim 2, wherein, The relationship verification is verified using a related vector machine algorithm based on adaptive feature learning; and the related vector machine algorithm based on adaptive feature learning is obtained in the following manner: Step 1: obtaining a patient- accompanying person relationship data set for training, and labeling high and low confidence levels; Step 2: initializing the kernel function parameters, learning rate, and regularization parameters of the related vector machine; Step 3: obtaining a data feature vector by vectorizing the patient data feature in the patient- accompanying person relationship data set, and obtaining a feature vector by feature mapping the data feature vector; Step 4: Based on the eigenvector space, the relevant vector machine of the Bayesian framework is trained iteratively until the stopping condition is met, and the adaptive feature learning relevant vector machine algorithm is obtained. 14.The knowledge graph based patient relationship association and information matching method of claim 13, wherein, The feature mapping is represented as: wherein is the i-th patient data feature vector input to the relevance vector machine; is the i-th patient data feature vector input to the relevance vector machine; is the mapped i-th patient data feature vector; is the mapped i-th patient data feature vector; is the mapping function, is the set of parameters required for the mapping.

15. The knowledge graph based patient relationship association and information matching method of claim 14, wherein, The mapping function is adjusted by optimizing the following learning objective function, represented as: wherein is based on the similarity weight of and in the original feature space; is the i-th patient data feature vector input to the relevance vector machine. is the i-th patient data feature vector input to the relevance vector machine. 16.The knowledge graph based patient relationship association and information matching method of claim 14, wherein, The calculation method of the mapping function is represented as: In the formula, This represents the first input to the correlation vector machine. The set of nearest neighbors of the feature vectors of each patient's data. It is the first input to the relevance vector machine. The feature vector of the patient data is input into the correlation vector machine. The contribution weights of the feature vectors of each patient's data.

17. The knowledge graph based patient relationship association and information matching method of claim 16, wherein, Optionally, the calculation method of the contribution weight is represented as: wherein is the L2 norm; respectively represent the i-th patient data feature vector input to the relevant vector machine. respectively represent the i-th patient data feature vector input to the relevant vector machine.

18. The knowledge graph based patient relationship association and information matching method of claim 13, wherein, The relevant vector machine of the Bayesian framework selects the patient data point with the highest posterior probability as the support vector by calculating the posterior probability distribution, and the calculation method is represented as: wherein is a label vector of input patient data; is a mapped feature vector of the patient; denotes a posterior probability distribution characterizing the probability of a prediction based on parameters ; and is a kernel function parameter of the relevance vector machine.

19. The knowledge graph based patient relationship association and information matching method of claim 18, wherein, In each iteration, the relevant vector machine of the Bayesian framework calculates the energy entropy of the model and optimizes the parameters of the model, and the calculation method of the energy entropy is represented as: wherein represents the energy entropy; is the number of samples of the current batch input to the relevance vector machine; represents the posterior probability distribution.

20. The knowledge graph based patient relationship association and information matching method of claim 18, wherein, The relevant vector machine of the Bayesian framework adaptively adjusts the model parameters, and the dynamic updating method of the kernel function parameters is represented as: wherein is the kernel function parameter of the updated relevance vector machine, is the learning rate of the relevance vector machine, denotes the gradient of the kernel function parameter of the relevance vector machine; denotes the energy entropy; denotes the regularization parameter.

21. The knowledge graph based patient relationship association and information matching method of claim 20, wherein, gradient The calculation of the gradient is represented as: wherein is a label of the first patient data; is the number of samples input to the relevance vector machine in the current batch; denotes the posterior probability distribution; denotes the energy entropy; denotes the regularization parameter; is a kernel function parameter of the relevance vector machine.

22. A computer device, comprising: The device comprises a memory and a processor; the memory is used to store a computer program; the processor executes the computer program to realize the steps of the method of any one of claims 1-21.

23. A computer-readable storage medium, characterized in that, A computer program is stored thereon, and the computer program is executed by a processor to realize the steps of the method of any one of claims 1-21.

24. A computer program product comprising a computer program, characterized in that, The computer program is executed by a processor to realize the steps of the method of any one of claims 1-21.

Images