Ophthalmology department hospital guide knowledge graph construction system based on improved gradient lifting

Abstract

The invention belongs to the technical field of intelligent medical treatment, and discloses an ophthalmology department hospital guide knowledge graph construction system based on improved gradient lifting. Comprising the steps of collecting multi-source ophthalmology original data; performing word segmentation processing on the multi-source ophthalmology original data to obtain ophthalmology medical vocabularies, and dividing the ophthalmology medical vocabularies into disease vocabularies and disease vocabularies; identifying professional vocabularies in the disease vocabularies, and constructing a rough reasoning algorithm and a fine reasoning algorithm; according to the disease vocabularies, the specialized vocabularies and a predefined disease weight set, an ophthalmology entity relation network is constructed, a rough reasoning algorithm and a fine reasoning algorithm are fused, and an ophthalmology hospital guide knowledge graph is constructed; according to the invention, complex symptoms can be dynamically analyzed, the diagnosis accuracy is improved, the medical resource configuration is optimized, and the self-evaluation requirement of the patient is met, so that the rapid response and accurate recommendation of the user symptoms are realized, and the medical treatment efficiency and the medical service quality of the patient are improved.

Description

Technical Field

[0001] The present invention relates to the field of smart medical technology, and more specifically, to an ophthalmology guidance knowledge graph construction system based on improved gradient boosting. Background Art

[0002] With the aging of the population and the popularization of electronic devices, the incidence of eye diseases continues to rise, and the problem of uneven distribution of ophthalmic medical resources has become increasingly prominent; patients often lack professional self-assessment capabilities before seeking medical treatment, resulting in inefficient allocation of medical resources; traditional ophthalmic guidance systems mainly rely on simple rule matching or keyword retrieval, which cannot handle complex symptom descriptions and vague expressions, and have limited accuracy and practicality; in order to assist doctors and patients in making scientific and efficient diagnoses, building an intelligent ophthalmic guidance system has become a research hotspot.

[0003] As an effective tool for expressing domain knowledge and semantic relationships, knowledge graphs can integrate and represent diseases, symptoms and their internal connections in the medical field, providing a solid data foundation and reasoning support for intelligent medical guidance. However, existing medical knowledge graphs focus more on static knowledge representation and lack effective reasoning mechanisms, making them difficult to adapt to scenarios such as ophthalmic medical guidance that require dynamic analysis and precise reasoning. At the same time, the field of ophthalmology is full of professional terminology, and the same symptom may be expressed in multiple ways, which increases the complexity and challenges of knowledge graph construction.

[0004] In view of this, the present invention proposes an ophthalmology guidance knowledge graph construction system based on improved gradient boosting to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned objectives, the present invention provides the following technical solution: an ophthalmology guidance knowledge graph construction system based on improved gradient boosting, comprising: Data acquisition module, used to collect multi-source ophthalmic raw data; The word segmentation processing module is used to perform word segmentation processing on multi-source ophthalmology raw data, obtain ophthalmology medical vocabulary, and divide ophthalmology medical vocabulary into symptom vocabulary and disease vocabulary; The vocabulary building module is used to identify professional terms in disease vocabulary and build a synonym database; The rough reasoning module is used to build a rough reasoning algorithm. The rough reasoning algorithm performs preliminary reasoning on the collected user input symptoms based on a predefined symptom weight set and a synonym library to obtain rough reasoning results. The fine reasoning module is used to build a fine reasoning algorithm. The fine reasoning algorithm uses a pre-built improved gradient model to perform deep reasoning on the rough reasoning results to obtain fine reasoning results. The graph construction module is used to construct an ophthalmic entity relationship network based on disease vocabulary, professional vocabulary and predefined disease weight sets, and integrate rough reasoning algorithms with fine reasoning algorithms to construct an ophthalmic guidance knowledge graph.

[0006] Furthermore, the method of dividing ophthalmology medical vocabulary into symptom vocabulary and disease vocabulary includes: Obtain disease seeds and symptom seeds in the ophthalmic medical vocabulary, and mark all words in the ophthalmic medical vocabulary that are not disease seeds or symptom seeds as words to be divided; convert each word in the ophthalmic medical vocabulary into a corresponding word vector, and calculate the average disease similarity between each word to be divided and all disease seeds, and the average symptom similarity between each word to be divided and all symptom seeds based on the word vector, where the average symptom similarity is the average of the symptom similarities between the word to be divided and each symptom seed; According to the comparison result of the average disease similarity of each word to be divided and the corresponding average symptom similarity, each word to be divided is divided into a symptom word or a disease word.

[0007] Furthermore, the method of constructing a synonym database includes: All disease seeds are regarded as professional vocabulary, and disease vocabulary that is not a professional vocabulary is regarded as the vocabulary to be analyzed; based on the vocabulary vector, the comprehensive similarity between each word to be analyzed and each professional vocabulary is calculated in turn; the comprehensive similarity with the largest value corresponding to each word to be analyzed is marked as the maximum similarity; the words to be analyzed with the maximum similarity greater than the preset similarity threshold are regarded as a group of synonyms with the corresponding professional vocabulary; based on all synonym groups, a synonym library is constructed.

[0008] Furthermore, the calculation method of the comprehensive similarity includes: Assign a character weight to each character position in the professional vocabulary and determine the editing operation corresponding to each character position; compare the professional vocabulary with the vocabulary to be analyzed to determine the editing operation corresponding to each character position in the vocabulary to be analyzed; set a corresponding operation weight for each editing operation, and calculate the weighted edit distance and the maximum possible distance based on the character weight of each character position and the operation weight of the corresponding editing operation; calculate the literal similarity based on the weighted edit distance and the maximum possible distance; A preset ratio set includes ratio coefficients corresponding to literal similarity and disease similarity; based on the ratio set, the disease similarity and literal similarity between the analyzed vocabulary and the professional vocabulary are weightedly summed to obtain a comprehensive similarity.

[0009] Furthermore, methods for obtaining rough inference results include: Collect ophthalmology professional vocabulary and build an extended dictionary; select a word segmentation tool, and segment the user-entered symptoms based on the tool's default dictionary and extended dictionary to extract the user's symptom vocabulary; map each user's symptom vocabulary to the corresponding professional vocabulary based on the synonym library and mark it as the current professional vocabulary; the symptom weight set includes the vocabulary weight between each professional vocabulary and each disease vocabulary; The lexical weight of each current professional vocabulary is obtained from the disease weight set, the lexical weights of the same disease vocabulary are added in sequence, the credibility corresponding to each disease vocabulary is obtained, and all credibility and corresponding disease vocabulary are used as rough inference results.

[0010] Furthermore, methods for obtaining refined reasoning results include: The disease words in the rough inference results are marked as candidate words, and the professional words corresponding to each candidate word are obtained according to the disease weight set; the professional words corresponding to the same candidate words are regarded as a set of standard symptom sets, and all current professional words are regarded as user symptom sets; the standard symptom sets and user symptom sets are collectively referred to as symptom sets, and the word frequency-inverse document frequency of each word in each symptom set is calculated in turn, and the vector space of each symptom set is modeled based on the word frequency-inverse document frequency to obtain the symptom vector of each symptom set; Based on the symptom vector, the symptom similarity between the user's symptom set and each set of standard symptom sets is calculated and used as the confidence of the candidate words corresponding to the standard symptom set; the rough reasoning results and all confidences are input into the improved gradient model to predict the reliability set; the candidate words in the reliability set whose reliability is greater than the preset reliability threshold are used as diagnostic words, and all diagnostic words are used as the results of fine reasoning.

[0011] Furthermore, the method for obtaining the symptom vector is as follows: construct a vocabulary based on the professional vocabulary in the synonym library; arrange the word frequency-inverse document frequency of each word in each group of symptom sets in sequence according to the order of the professional vocabulary in the vocabulary to form the symptom vector of each group of symptom sets; wherein, if there is a word in the vocabulary that does not appear in the symptom set, the word frequency-inverse document frequency of the corresponding word in the corresponding symptom set is set to 0.

[0012] Furthermore, the steps for building the improved gradient model include: Step S1: construct the initial feature space; Step S2: constructing a disease relationship network based on the initial feature space; Step S3: extracting the kink feature vector from the disease relationship network and merging it with the initial feature space to form an enhanced feature space; Step S4: Iteratively train the gradient boosting model based on the enhanced feature space and introduce the kink entropy regularization term. After the iterative training is completed, an improved gradient model is obtained.

[0013] Furthermore, the kink eigenvector includes node centrality and kink invariant; The calculation method of the kink entropy regularization term includes: According to the kink invariant, all nodes in the disease relationship network are divided into kink types and count the number of nodes in each kink type; obtain the relative frequency of each kink type based on the ratio of the number of nodes in each kink type to the total number of nodes; calculate the kink entropy based on the relative frequency of each kink type; and obtain the kink entropy regularization term based on the product of the kink entropy and the preset regularization strength parameter.

[0014] Furthermore, the method of introducing the kink entropy regularization term is: adding the kink entropy regularization term to the loss function during the iterative training process of the gradient boosting model.

[0015] The technical effects and advantages of the ophthalmology guidance knowledge graph construction system based on improved gradient boosting of the present invention are as follows: Through the collection and word segmentation of multi-source ophthalmic data, ophthalmic medical vocabulary is accurately divided into symptom vocabulary and disease vocabulary, laying a solid vocabulary foundation for the subsequent knowledge graph establishment; a comprehensive similarity calculation method is adopted, combining literal similarity and semantic similarity, to effectively identify professional vocabulary in symptom vocabulary, build a professional vocabulary library covering synonyms, and improve the ability to understand user input descriptions; based on the predefined symptom weight set and synonym library, a rough inference algorithm is designed, which can quickly and preliminarily identify possible disease candidate sets from user symptoms, providing a basis for subsequent fine reasoning; an improved gradient boosting model is adopted, integrating similarity analysis based on symptom vectors, and the algorithm can effectively identify professional vocabulary in symptom vocabulary, build a professional vocabulary library covering synonyms, and improve the ability to understand user input descriptions; based on the predefined symptom weight set and synonym library, a rough inference algorithm is designed, which can quickly and preliminarily identify possible disease candidate sets from user symptoms, providing a basis for subsequent fine reasoning ... vector similarity analysis, a rough inference algorithm is designed, which can quickly and preliminarily identify possible disease candidate sets from user symptoms, providing a basis for subsequent fine reasoning; based on the predefined symptom vector similarity analysis, a rough inference algorithm is designed, combining symptom vector similarity analysis, and the algorithm can effectively identify professional vocabulary in symptom vocabulary, build a professional vocabulary library covering synonyms, and improve the understanding of user input descriptions; based on the predefined symptom weight set and synonym library, a rough inference algorithm is designed, which can quickly and preliminarily identify possible disease candidate sets from user symptoms, providing a basis for subsequent fine Analysis and relationship modeling based on kink features can significantly enhance the diagnostic accuracy and reliability of the reasoning algorithm, and improve the accuracy and reliability of fine reasoning results; the rough reasoning results and fine reasoning results are integrated to construct an ophthalmic knowledge graph containing disease entities, symptom entities and their relationships, which can effectively support the knowledge representation and reasoning functions of the intelligent guidance system; realize the automation of the entire process from raw data to knowledge representation, can dynamically analyze complex symptoms, improve diagnostic accuracy, optimize medical resource allocation, meet the needs of patients for self-assessment, thereby achieving rapid response and accurate recommendation to user symptoms, improving patients' medical efficiency and medical service quality, and promoting the development of smart medical care. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a schematic diagram of a system for constructing an ophthalmology guidance knowledge graph based on improved gradient boosting according to Example 1 of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 making creative efforts are within the scope of protection of the present invention.

[0018] Example 1 See also Figure 1 As shown, the ophthalmic guidance knowledge graph construction system based on improved gradient boosting described in this embodiment includes a data acquisition module, a word segmentation processing module, a vocabulary establishment module, a rough reasoning module, a fine reasoning module and a graph construction module; each module is connected by wired and / or wireless means to realize data transmission between modules.

[0019] Data acquisition module, used to collect multi-source ophthalmic raw data.

[0020] Multi-source ophthalmology raw data includes but is not limited to medical institution data, open medical data, and online platform data; among them, medical institution data refers to real clinical data from medical institutions such as hospitals and ophthalmology centers, including but not limited to electronic medical record data and ophthalmology examination reports; open medical data refers to data from authoritative public resources, standard terminology libraries, and academic databases with open access or semi-open authorization features, including but not limited to medical terminology libraries and medical literature abstracts; online platform data refers to user interaction content from Internet medical platforms, Q&A communities, and guidance apps, including but not limited to online consultation dialogues and health Q&A community content; Electronic medical record data includes but is not limited to chief complaints, current medical history, diagnostic records, etc.; ophthalmic examinations include but are not limited to optometry results, slit lamp examination results, etc.; medical terminology libraries such as ICD-11 (disease classification), SNOMED CT (clinical terminology library), MeSH (Medical Subject Headings), etc.; medical literature abstracts such as literature abstracts from platforms such as PubMed, CNKI, Wanfang, Google Scholar, etc., containing the latest clinical research conclusions; online consultation dialogues such as structured dialogue texts such as patients describing their symptoms and doctors giving preliminary suggestions; health question and answer community content such as community users posting questions independently and doctors or other users replying, etc. Health question and answer communities include Baidu Health, Dingxiangyuan Forum, Zhihu Health Topics, etc.

[0021] The word segmentation processing module is used to perform word segmentation processing on multi-source ophthalmology raw data, obtain ophthalmology medical vocabulary, and divide ophthalmology medical vocabulary into symptom vocabulary and disease vocabulary.

[0022] Ways to acquire ophthalmology vocabulary include: Common Chinese word segmentation tools (such as Jieba, IK Analyzer, NLPIR, etc.) are used to segment multi-source ophthalmology raw data and extract candidate medical terms; the candidate medical terms are sequentially subjected to part-of-speech filtering and rule filtering to eliminate non-entity terms and invalid terms in the candidate medical terms; technical personnel in this field review the filtered candidate medical terms to obtain ophthalmic medical terms; part-of-speech filtering is to remove words with specific parts of speech, such as verbs, adverbs, conjunctions, prepositions, etc.; rule filtering is to remove invalid or irrelevant words according to preset rules, such as stop words (such as de, le, shi, etc.), specific characters (such as numbers, special symbols, etc.), and field-irrelevant words (i.e., common words in non-medical fields, such as feeling, yesterday, me, etc.); the specific contents of part-of-speech filtering and rule filtering are pre-determined by technical personnel in this field according to actual needs.

[0023] Symptom vocabulary refers to words that describe the patient's subjective symptoms or abnormal manifestations found in objective examinations, such as blurred vision, eye pain, corneal edema, etc.; disease vocabulary refers to words that have clear diagnostic significance and represent disease entities, such as glaucoma, cataract, macular degeneration, etc.

[0024] Methods for dividing ophthalmology vocabulary into symptom vocabulary and disease vocabulary include: In ophthalmology medical vocabulary, all words collected from ICD-11 are marked as disease seeds, and all words collected from SNOMED CT are marked as symptom seeds; all words in ophthalmology medical vocabulary that are not marked as disease seeds or symptom seeds are marked as words to be divided; a pre-trained word vector model (such as Word2Vec, FastText, etc.) is used to convert each word in the ophthalmology medical vocabulary into a corresponding word vector; based on the word vectors, the average disease similarity between each word to be divided and all disease seeds, as well as the average symptom similarity between each word to be divided and all symptom seeds are calculated in turn; the average disease similarity of each word to be divided is compared with the average symptom similarity; if the average disease similarity is greater than the average symptom similarity, the corresponding word to be divided is divided into a disease word; if the average disease similarity is less than the average symptom similarity, the corresponding word to be divided is divided into a symptom word; if the average disease similarity is equal to the average symptom similarity, the corresponding word to be divided is manually divided by a person skilled in the art; the disease seeds are divided into disease words, and the symptom seeds are divided into symptom words; Among them, the average disease similarity is the mean of the disease similarities between the words to be divided and each disease seed, and the average symptom similarity is the mean of the symptom similarities between the words to be divided and each symptom seed. Both disease similarity and symptom similarity are cosine similarities.

[0025] The vocabulary building module is used to identify professional terms in disease vocabulary and build a synonym database.

[0026] Methods for building a synonym library include: All disease seeds are regarded as professional words, and disease words that are not professional words are regarded as words to be analyzed; based on the word vector, the comprehensive similarity between each word to be analyzed and each professional word is calculated in turn; the comprehensive similarity with the largest value corresponding to each word to be analyzed is marked as the maximum similarity; each maximum similarity is compared with a preset similarity threshold; the words to be analyzed whose maximum similarity is greater than the similarity threshold are regarded as a synonym group with the corresponding professional words; the words to be analyzed whose maximum similarity is less than or equal to the similarity threshold are not regarded as a synonym group with the corresponding professional words; a synonym library is constructed based on all synonym groups; the similarity threshold is pre-set by a person skilled in the art according to actual conditions.

[0027] The calculation method of comprehensive similarity includes: Assign a character weight to each character position in the professional vocabulary, and determine the editing operation corresponding to each character position; wherein the character weight increases with the character position, that is, the character position closer to the end of the vocabulary has a higher character weight, so as to reflect the important contribution of the character at the end of the vocabulary to the word meaning; compare the professional vocabulary with the vocabulary to be analyzed, and determine the editing operation corresponding to each character position in the vocabulary to be analyzed; set a corresponding operation weight for each editing operation, and calculate the weighted editing distance and the maximum possible distance based on the character weight of each character position and the operation weight of the corresponding editing operation; wherein editing operations include insertion, replacement and deletion, and the specific values ​​of the character weight and operation weight are determined by those skilled in the art based on actual application scenarios and actual experience; The literal similarity is calculated based on the weighted edit distance and the maximum possible distance; a preset ratio set is set, and the ratio set includes proportional coefficients corresponding to the literal similarity and the disease similarity, which are pre-set by technical personnel in this field according to actual conditions; based on the ratio set, the disease similarity between the word to be analyzed and the professional vocabulary, as well as the literal similarity, are weighted and summed to obtain a comprehensive similarity.

[0028] The weighted edit distance is calculated as follows: the product of the character weight at the corresponding character position and the operation weight is used as the weight product of the corresponding character position; the weight products of all character positions are sequentially added together to obtain the weighted edit distance; it should be noted that when the editing operation is insertion, since there is no original character at the insertion position, that is, there is no corresponding character position, the character weight of the insertion position is defined as the average of the character weights of the two character positions before and after the insertion position; if the insertion position is the beginning or end of the word, that is, there is only one adjacent character position, the character weight of the adjacent character position is used as the character weight of the insertion position; The maximum possible distance is calculated by sequentially deleting and inserting all character positions in the word to be analyzed, and calculating the corresponding weighted edit distance as the maximum possible distance; the maximum possible distance is the maximum value of the weighted edit distance in all character edit paths; for example, the word to be analyzed is gastritis, and the professional word is chronic gastritis, so gastritis in the word to be segmented is first deleted sequentially, and then chronic gastritis in the professional word is inserted; The method for calculating the literal similarity is to take the ratio of the weighted edit distance to the maximum possible distance as the normalized edit distance, and take the difference between one and the normalized edit distance as the literal similarity.

[0029] The rough reasoning module is used to build a rough reasoning algorithm. The rough reasoning algorithm performs preliminary reasoning on the collected user input symptoms based on a predefined symptom weight set and a synonym library to obtain rough reasoning results.

[0030] Methods for obtaining rough inference results include: Collect ophthalmic professional vocabulary and build an extended dictionary; select a word segmentation tool, and segment the user-input symptoms based on the default dictionary and extended dictionary of the word segmentation tool to extract user symptom vocabulary; map each user symptom vocabulary to the corresponding professional vocabulary based on the synonym library and mark it as the current professional vocabulary; the symptom weight set includes the lexical weight between each professional vocabulary and each disease vocabulary, which is used to measure the degree of diagnostic support of the symptoms for the disease; obtain the lexical weight of each current professional vocabulary from the symptom weight set, add the lexical weights of the same disease vocabulary in sequence, obtain the corresponding credibility of each disease vocabulary, and use all credibility and corresponding disease vocabulary as rough inference results; among them, the credibility in the rough inference results is all greater than 0.

[0031] Among them, ophthalmology professional vocabulary refers to professional symptom terms used in ophthalmology diagnosis, description, and communication, which are obtained through expert interviews, diagnosis and treatment guide textbooks, etc. In this embodiment, the word segmentation tool selected is IKAnalyzer. IKAnalyzer is a lightweight open source Chinese word segmentation tool based on the Java language. It mainly includes components such as IKAnalyzer.cfg.xml (word segmentation extension configuration file), IKAnalyzer2012_u6.jar (main jar package), stopword.dic (stopword dictionary), LICENSE.txt, and NOTICE.txt (Apache copyright notice). The default dictionary is the basic dictionary that comes with the word segmentation tool when it is installed or deployed. It contains common words, general terms, common phrases, etc., which help the word segmentation tool perform basic segmentation on ordinary text. The extended dictionary uses a configuration-based dictionary expansion method. Each ophthalmology professional word in the extended dictionary is placed line by line in the diseasename.txt text file. Notepad++ converts it into the diseasename.dic file in UTF-8 encoding format without BOM. The user-defined extended dictionary path in the IKAnalyzer.cfg.xml file is modified to implement the configuration of the extended dictionary. The disease weight set is pre-defined by technicians in this field based on comprehensive factors such as clinical knowledge, standard diagnosis and treatment pathways, professional terminology system and expert consensus.

[0032] The fine reasoning module is used to build a fine reasoning algorithm. The fine reasoning algorithm uses a pre-built improved gradient model to perform deep reasoning on the rough reasoning results to obtain fine reasoning results.

[0033] Methods for obtaining refined inference results include: The disease words in the rough inference results are marked as candidate words, and the professional words corresponding to each candidate word are obtained according to the symptom weight set, that is, the professional words whose word weights between the candidate words and the symptom weight set are not 0 are obtained; the professional words corresponding to the same candidate words are regarded as a set of standard symptom sets, and all current professional words are regarded as user symptom sets; the standard symptom sets and the user symptom sets are collectively referred to as symptom sets, and the word frequency-inverse document frequency of each word in each symptom set is calculated in turn, and the vector space of each symptom set is modeled based on the word frequency-inverse document frequency to obtain the symptom vector of each symptom set; Based on the symptom vector, the symptom similarity between the user's symptom set and each set of standard symptom sets is calculated and used as the confidence of the candidate words corresponding to the standard symptom set; the symptom similarity is the cosine similarity; the rough inference results and all confidences are input into the improved gradient model to predict the reliability set; the reliability set includes the reliability of each candidate word; the candidate words in the reliability set whose reliability is greater than the preset reliability threshold are used as diagnostic words, and all diagnostic words are used as the refined inference results.

[0034] Among them, word frequency-inverse document frequency is used to measure the importance of a word in a set of symptom sets, specifically the product of word frequency and inverse document frequency; word frequency is used to reflect the frequency of a word in a set of symptom sets, specifically the ratio of the number of word occurrences to the maximum number of word occurrences, the number of word occurrences is the number of times a word appears in a set of symptom sets, and the maximum number of word occurrences is the number of times the word appears the most in the corresponding symptom set; inverse document frequency is used to reduce the weight of common words, and the expression of inverse document frequency is: Where, is the inverse document frequency, is the total number of symptom sets, is the number of symptom sets that contain the corresponding vocabulary when calculating the word frequency.

[0035] The method for obtaining the symptom vector is as follows: based on the professional vocabulary in the synonym library, a vocabulary is constructed as the dimensional basis of the symptom vector; the word frequency-inverse document frequency of each word in each group of symptom sets is arranged in sequence according to the order of the professional vocabulary in the vocabulary to form the symptom vector of each group of symptom sets; among them, if there is a word in the vocabulary that does not appear in the symptom set, the word frequency-inverse document frequency of the corresponding word in the corresponding symptom set is set to 0; that is, the number of word frequency-inverse document frequencies contained in each symptom vector is consistent with the number of professional vocabulary contained in the vocabulary, which helps to quantify, compare and analyze the relationship between the user symptom set and the standard symptom set.

[0036] The steps to build the improved gradient model include: Step S1: construct the initial feature space; Step S2: constructing a disease relationship network based on the initial feature space; Step S3: extracting the kink feature vector from the disease relationship network and merging it with the initial feature space to form an enhanced feature space; Step S4: Iteratively train the gradient boosting model based on the enhanced feature space and introduce the kink entropy regularization term. After the iterative training is completed, an improved gradient model is obtained.

[0037] In the above step S1, the initial feature space includes the credibility and confidence corresponding to multiple different disease terms collected by technical personnel in this field, as well as the reliability set according to actual experience; among them, the credibility and confidence corresponding to the same disease term are used as a set of initial feature vectors, and the corresponding reliability is used as the target value.

[0038] In the above step S2, the method for constructing the disease relationship network includes: Using a pre-trained word vector model, each disease word in the initial feature space is converted into a corresponding disease vector. Based on the disease vectors, the disease similarity between each two disease words is calculated and marked as the relationship similarity. Based on the relationship similarity, an adjacency matrix is ​​constructed. Based on all disease words, disease similarities, and the adjacency matrix in the initial feature space, a disease relationship network is constructed. The size of the adjacency matrix is , is the number of disease words in the initial feature space. The method of constructing the adjacency matrix is ​​as follows: when the relationship similarity between two disease words is greater than the preset relationship threshold, the corresponding position in the adjacency matrix is ​​assigned the corresponding relationship similarity; according to the relationship similarity, the similarity set corresponding to each disease word is obtained, and the similarity set includes the disease word with the highest relationship similarity. Disease vocabulary, When the similarity set of one of the two disease terms contains the other disease term, the corresponding position in the adjacency matrix is ​​assigned the corresponding relationship similarity; when the relationship similarity between the two disease terms is less than or equal to the preset relationship threshold, and neither of them is in the other's similarity set, the corresponding position in the adjacency matrix is ​​assigned zero; the relationship threshold is preset by those skilled in the art according to actual conditions; All disease words in the initial feature space are regarded as nodes in the disease relationship network, and edges are constructed between two nodes corresponding to non-zero elements in the adjacency matrix. The weight of the edge is the value of the corresponding non-zero element.

[0039] The method of extracting the kink feature vector from the disease relationship network in step S3 above includes: The kink eigenvector includes node centrality and kink invariant; Methods for extracting node centrality from disease relationship networks include: Initialize the center vector, which includes numerical values, each of which has an initial value of 1; defining an iterative process, the iterative process is: normalizing the product of the adjacency matrix and the center vector to obtain a modulation vector; updating the center vector according to the modulation vector; looping the iterative process, and calculating the modulation difference after each iterative process is completed until the modulation difference is less than a preset difference threshold, stopping the iterative process, and obtaining the center vector, which includes the node centrality of each node; the modulation difference is the Euclidean distance between the center vector after the update and the center vector before the update; the difference threshold is pre-set by a person skilled in the art according to actual conditions.

[0040] Methods for extracting kink invariants from disease relationship networks include: Calculate the clustering coefficient of each node in turn, and subtract each clustering coefficient from one to get the kink crossing number of each node, and use the kink crossing number of all nodes as the kink invariant; The expression of clustering coefficient is: Where, For the The clustering coefficient of nodes, For the The number of edges between neighbor nodes of a node, For the The number of neighbor nodes of the node, the neighbor nodes are There are edges connecting the nodes.

[0041] The method for forming the enhanced feature space is as follows: merging the initial feature vector of each node with the kink feature vector to form an enhanced feature vector; replacing the initial feature vector of each node in the initial feature space with the corresponding enhanced feature vector to form an enhanced feature space.

[0042] In the above step S4, the calculation method of the kink entropy regularization term includes: According to the kink invariant, all nodes are divided into kink types and counting the number of nodes in each kink type; obtaining the relative frequency of each kink type based on the ratio between the number of nodes of each kink type and the total number of nodes; calculating the kink entropy based on the relative frequency of each kink type; obtaining the kink entropy regularization term based on the product of the kink entropy and a preset regularization strength parameter; wherein the kink entropy is the information entropy, and the regularization strength parameter is preset by those skilled in the art according to actual conditions; The method of introducing the knot entropy regularization term is: adding the knot entropy regularization term to the loss function during the iterative training process of the gradient boosting model; the gradient boosting model is an existing technology, and the specific training process will not be described in detail here.

[0043] The graph construction module is used to construct an ophthalmic entity relationship network based on disease vocabulary, professional vocabulary and predefined disease weight sets, and integrate rough reasoning algorithms with fine reasoning algorithms to construct an ophthalmic guidance knowledge graph.

[0044] The method for constructing an ophthalmology entity relationship network is as follows: disease vocabulary is used as disease entity and professional vocabulary is used as symptom entity; a connection edge is constructed between each disease entity and each symptom entity; according to the symptom weight set, the vocabulary weight between each disease entity and symptom entity is obtained and used as the connection weight of the corresponding connection edge; according to the connection weight of the disease entity, symptom entity and the connection weight of the connection edge between disease entity and symptom entity, an ophthalmology entity relationship network is constructed.

[0045] This embodiment accurately divides ophthalmic medical vocabulary into symptom vocabulary and disease vocabulary through the collection and word segmentation of multi-source ophthalmic data, laying a solid vocabulary foundation for the subsequent knowledge graph establishment; adopts a comprehensive similarity calculation method, combining literal similarity and semantic similarity, effectively identifies professional vocabulary in symptom vocabulary, builds a professional vocabulary library covering synonyms, and improves the ability to understand user input descriptions; based on the predefined symptom weight set and synonym library, designs a rough inference algorithm, which can quickly and preliminarily identify possible disease candidate sets from user symptoms, providing a basis for subsequent fine reasoning; adopts an improved gradient boosting model, integrates similarity based on symptom vectors, and uses the symptom vector-based similarity algorithm to generate a comprehensive similarity calculation method. Degree analysis and relationship modeling based on kink features significantly enhance the diagnostic accuracy and reliability of the reasoning algorithm, and improve the accuracy and reliability of fine reasoning results; the coarse reasoning results and fine reasoning results are integrated to construct an ophthalmic knowledge graph containing disease entities, symptom entities and their relationships, which can effectively support the knowledge representation and reasoning functions of the intelligent guidance system; realize the automation of the entire process from raw data to knowledge representation, can dynamically analyze complex symptoms, improve diagnostic accuracy, optimize medical resource allocation, meet the needs of patients for self-assessment, thereby achieving rapid response and accurate recommendation to user symptoms, improving patients' medical efficiency and medical service quality, and promoting the development of smart medical care.

[0046] Example 2 The present application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories may store computer-readable code that, when executed by the one or more processors, may execute the above-described improved gradient boosting-based ophthalmology guidance knowledge graph construction system.

[0047] The method or system according to the embodiment of the present application can also be implemented with the help of the architecture of the electronic device shown in this application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as a ROM or a hard disk, can store the ophthalmic guidance knowledge graph construction system based on improved gradient boosting provided by this application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in this application is only exemplary. When implementing different devices, one or more components in the electronic device shown in this application can be omitted according to actual needs.

[0048] Example 3 One embodiment of the present application discloses a computer-readable storage medium. Computer-readable instructions are stored on the computer-readable storage medium. When the computer-readable instructions are executed by a processor, the ophthalmology guidance knowledge graph construction system based on improved gradient boosting according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache). Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0049] In addition, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, for example: an ophthalmology guidance knowledge graph construction system based on improved gradient boosting. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are performed.

[0050] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will be able to modify the technical solutions described in the foregoing embodiments or to substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

[0051] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0052] In the description of the present invention, it should be understood that the terms "first", "second", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0053] In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0054] In the description of the present invention, “several” means one or more, and “a large number” means two or more.

[0055] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0056] The formulas in this manual are all dimensionless and calculated using numerical values. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field based on actual conditions.

[0057] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. An ophthalmology guidance knowledge graph construction system based on improved gradient boosting, characterized by: include: Data acquisition module, used to collect multi-source ophthalmic raw data; The word segmentation processing module is used to perform word segmentation processing on multi-source ophthalmology raw data, obtain ophthalmology medical vocabulary, and divide ophthalmology medical vocabulary into symptom vocabulary and disease vocabulary; The vocabulary building module is used to identify professional terms in disease vocabulary and build a synonym database; The rough reasoning module is used to build a rough reasoning algorithm. The rough reasoning algorithm performs preliminary reasoning on the collected user input symptoms based on a predefined symptom weight set and a synonym library to obtain rough reasoning results. The fine reasoning module is used to build a fine reasoning algorithm. The fine reasoning algorithm uses a pre-built improved gradient model to perform deep reasoning on the rough reasoning results to obtain fine reasoning results. The graph construction module is used to construct an ophthalmic entity relationship network based on disease vocabulary, professional vocabulary and predefined disease weight sets, and integrate rough reasoning algorithms with fine reasoning algorithms to construct an ophthalmic guidance knowledge graph.

2. The ophthalmology guidance knowledge graph construction system based on improved gradient boosting according to claim 1 is characterized in that: Methods for dividing ophthalmology vocabulary into symptom vocabulary and disease vocabulary include: Obtain disease seeds and symptom seeds in the ophthalmic medical vocabulary, and mark all words in the ophthalmic medical vocabulary that are not disease seeds or symptom seeds as words to be divided; convert each word in the ophthalmic medical vocabulary into a corresponding word vector, and calculate the average disease similarity between each word to be divided and all disease seeds, and the average symptom similarity between each word to be divided and all symptom seeds based on the word vector, where the average symptom similarity is the average of the symptom similarities between the word to be divided and each symptom seed; According to the comparison result of the average disease similarity of each word to be divided and the corresponding average symptom similarity, each word to be divided is divided into a symptom word or a disease word.

3. The ophthalmology guidance knowledge graph construction system based on improved gradient boosting according to claim 2 is characterized in that: Methods for building a synonym library include: All disease seeds are regarded as professional vocabulary, and disease vocabulary that is not a professional vocabulary is regarded as the vocabulary to be analyzed; based on the vocabulary vector, the comprehensive similarity between each word to be analyzed and each professional vocabulary is calculated in turn; the comprehensive similarity with the largest value corresponding to each word to be analyzed is marked as the maximum similarity; the words to be analyzed with the maximum similarity greater than the preset similarity threshold are regarded as a group of synonyms with the corresponding professional vocabulary; based on all synonym groups, a synonym library is constructed.

4. The ophthalmology guidance knowledge graph construction system based on improved gradient boosting according to claim 3 is characterized in that: The calculation method of comprehensive similarity includes: Assign a character weight to each character position in the professional vocabulary and determine the editing operation corresponding to each character position; compare the professional vocabulary with the vocabulary to be analyzed to determine the editing operation corresponding to each character position in the vocabulary to be analyzed; set a corresponding operation weight for each editing operation, and calculate the weighted edit distance and the maximum possible distance based on the character weight of each character position and the operation weight of the corresponding editing operation; calculate the literal similarity based on the weighted edit distance and the maximum possible distance; A preset ratio set includes ratio coefficients corresponding to literal similarity and disease similarity; based on the ratio set, the disease similarity and literal similarity between the analyzed vocabulary and the professional vocabulary are weightedly summed to obtain a comprehensive similarity.

5. The ophthalmology guidance knowledge graph construction system based on improved gradient boosting according to claim 4 is characterized in that: Methods for obtaining rough inference results include: Collect ophthalmology professional vocabulary and build an extended dictionary; select a word segmentation tool, and segment the user-entered symptoms based on the tool's default dictionary and extended dictionary to extract the user's symptom vocabulary; map each user's symptom vocabulary to the corresponding professional vocabulary based on the synonym library and mark it as the current professional vocabulary; the symptom weight set includes the vocabulary weight between each professional vocabulary and each disease vocabulary; The lexical weight of each current professional vocabulary is obtained from the disease weight set, the lexical weights of the same disease vocabulary are added in sequence, the credibility corresponding to each disease vocabulary is obtained, and all credibility and corresponding disease vocabulary are used as rough inference results.

6. The ophthalmology guidance knowledge graph construction system based on improved gradient boosting according to claim 5 is characterized in that: Methods for obtaining refined inference results include: The disease words in the rough inference results are marked as candidate words, and the professional words corresponding to each candidate word are obtained according to the disease weight set; the professional words corresponding to the same candidate words are regarded as a set of standard symptom sets, and all current professional words are regarded as user symptom sets; the standard symptom sets and user symptom sets are collectively referred to as symptom sets, and the word frequency-inverse document frequency of each word in each symptom set is calculated in turn, and the vector space of each symptom set is modeled based on the word frequency-inverse document frequency to obtain the symptom vector of each symptom set; Based on the symptom vector, the symptom similarity between the user's symptom set and each set of standard symptom sets is calculated and used as the confidence of the candidate words corresponding to the standard symptom set; the rough reasoning results and all confidences are input into the improved gradient model to predict the reliability set; the candidate words in the reliability set whose reliability is greater than the preset reliability threshold are used as diagnostic words, and all diagnostic words are used as the results of fine reasoning.

7. The ophthalmology guidance knowledge graph construction system based on improved gradient boosting according to claim 6 is characterized in that: The method for obtaining the symptom vector is as follows: construct a vocabulary based on the professional vocabulary in the synonym library; arrange the word frequency-inverse document frequency of each word in each group of symptom sets in the order of the professional vocabulary in the vocabulary to form the symptom vector of each group of symptom sets; among them, if there is a word in the vocabulary that does not appear in the symptom set, the word frequency-inverse document frequency of the corresponding word in the corresponding symptom set is set to 0.

8. The ophthalmology guidance knowledge graph construction system based on improved gradient boosting according to claim 7 is characterized in that: The steps to build the improved gradient model include: Step S1: construct the initial feature space; Step S2: constructing a disease relationship network based on the initial feature space; Step S3: extracting the kink feature vector from the disease relationship network and merging it with the initial feature space to form an enhanced feature space; Step S4: Iteratively train the gradient boosting model based on the enhanced feature space and introduce the kink entropy regularization term. After the iterative training is completed, an improved gradient model is obtained.

9. The ophthalmology guidance knowledge graph construction system based on improved gradient boosting according to claim 8, characterized in that: The kink eigenvector includes node centrality and kink invariant; The calculation method of the kink entropy regularization term includes: According to the kink invariant, all nodes in the disease relationship network are divided into kink types and count the number of nodes in each kink type; obtain the relative frequency of each kink type based on the ratio between the number of nodes in each kink type and the total number of nodes; The kink entropy is calculated based on the relative frequency of each kink type; the kink entropy regularization term is obtained by multiplying the kink entropy with the preset regularization strength parameter.

10. The ophthalmology guidance knowledge graph construction system based on improved gradient boosting according to claim 9, characterized in that: The method of introducing the kink entropy regularization term is to add the kink entropy regularization term to the loss function during the iterative training process of the gradient boosting model.

Images