Event correlation analysis method and system based on large model graph rules

Abstract

The application discloses an event correlation analysis method and system based on a large model graph rule, relates to the technical field of sick event correlation analysis, and comprises the following steps: establishing data connection with medical big data, and acquiring publicly available case data; extracting symptom manifestations in the case data, and analyzing symptom correlation between different single symptoms in the symptom manifestations; based on an initial symptom input by a user, the estimated disease of the user is analyzed in combination with the symptom correlation; the application is used for solving the problem that the existing sick event correlation analysis technology is not deep enough in analyzing the correlation between symptoms and diseases, so that the analysis result cannot effectively assist medical treatment.

Description

Technical Field

[0001] This invention relates to the field of disease event association analysis technology, specifically to an event association analysis method and system based on large model graph rules. Background Technology

[0002] Disease event association analysis is a data mining technique used to discover correlations and patterns between different diseases or symptoms. By analyzing large amounts of medical data, including case records, symptoms, examination results, and treatment plans, disease event association analysis can help medical professionals discover potential associations between diseases or determine the degree of association between a certain symptom and a specific disease. This technique can help doctors better understand the mutual influence between diseases and provide more accurate references for diagnosis and treatment.

[0003] Existing disease event association analysis techniques typically list multiple possible symptoms based on the initial symptoms provided by the user, commonly seen in online consultations. However, users usually lack diagnostic capabilities. When listing multiple possible symptoms, if they cannot provide a precise or closest diagnosis, providing a diagnosis through multiple results is ultimately ineffective. Users cannot correctly identify the symptom when faced with multiple possibilities. Existing methods do not accurately categorize these symptoms, and the analysis results are largely useless when multiple possibilities are listed, offering limited medical assistance. Furthermore, existing disease event association analysis techniques lack in-depth analysis of the correlation between symptoms and diseases, resulting in analysis results that cannot effectively assist in medical care. Summary of the Invention

[0004] This invention aims to at least partially address one of the technical problems in the prior art. By establishing a data connection with medical big data, it obtains publicly available case data, extracts keywords from symptom manifestations to obtain individual symptoms, analyzes the symptom correlation between different individual symptoms, obtains the initial symptoms input by the user, analyzes the co-related symptoms based on the symptom correlation, and finally, based on the initial symptoms and co-related symptoms, accurately predicts the user's actual condition. This addresses the problem that existing disease event correlation analysis technologies do not provide in-depth analysis of the correlation between symptoms and conditions, resulting in analysis results that cannot effectively assist in medical treatment.

[0005] To achieve the above objectives, firstly, this application provides an event association analysis method based on large model graph rules, comprising the following steps:

[0006] Establish data connections with medical big data to obtain publicly available case data;

[0007] Extract symptom presentations from case data and analyze the symptom correlations between different individual symptoms within the symptom presentations;

[0008] Based on the user's initial symptoms, the predicted symptoms are analyzed by combining symptom correlation analysis.

[0009] Furthermore, establishing a data connection with medical big data to obtain publicly available case data includes the following sub-steps:

[0010] Establish data connections with medical big data;

[0011] Obtain publicly available case data;

[0012] The case data includes symptoms and confirmed conditions.

[0013] Furthermore, extracting symptom presentations from case data and analyzing the symptom correlations between different individual symptoms includes the following sub-steps:

[0014] Keyword extraction is performed on symptom manifestations to obtain single symptoms;

[0015] Analyze the symptom associations between different individual symptoms within the symptom presentation.

[0016] Furthermore, keyword extraction is performed on the symptom manifestations to obtain the following sub-steps for a single symptom:

[0017] Based on NLP keyword extraction technology, keywords are extracted from symptom manifestations to obtain manifestation word segmentation;

[0018] The performance segmentation words are compared with the symptom database, and the performance segmentation words that exist in the symptom database are marked as single symptoms.

[0019] Furthermore, analyzing the symptom associations between different individual symptoms within the symptom presentation includes the following sub-steps:

[0020] The symptoms and the confirmed diagnosis constitute a case data, and the number of individual symptoms included in the symptoms of different case data is different.

[0021] For any single symptom, mark it as the target symptom, find all case data containing the target symptom, and mark them as target data;

[0022] The number of target symptoms in the target data is marked as the target quantity. For any single symptom in the target data, its quantity is counted and marked as the quantity to be associated. The target quantity and the quantity to be associated are collectively referred to as the symptom quantity.

[0023] Sort and number the quantities to be associated in ascending order of quantity, using the symbol P. nThis represents a sequence where n is a non-zero natural number and n is the index of P. The maximum value of n is incremented by one and marked as N. This is then processed through P. N Indicates the target quantity;

[0024] Establish a Cartesian coordinate system with the numbers 1 and 2 as the X-axis and the number of symptoms as the Y-axis, and name it the Correlation Proportion Analysis Chart. Place the coordinate point (2, P) N ) and coordinates (1, P) n Enter the correlation ratio analysis chart;

[0025] Set the coordinates (2, P) N Name it the target point, and set the coordinates (1, P) as the target point. n The target point is named the point to be associated. Connect the target point to each of the different points to be associated using straight line segments to obtain the associated line segments. Then, assign the coordinates (1, P) to the target point. n ) and coordinates (2, P) N The associated line segment is marked as L. n ;

[0026] Get L n The slope, denoted as K n K in ascending order n Number them using the symbol T i Let i represent the sequence number of T, where i is a non-zero natural number and i is the index of T;

[0027] With i as the horizontal axis, T i Establish a Cartesian coordinate system for the vertical axis, name it the Effective Correlation Analysis Chart, and set T... i Enter the valid association analysis graph according to i;

[0028] Based on effective association analysis graphs, identify the symptoms to be associated with the target symptoms.

[0029] Furthermore, the analysis of symptoms to be associated with the target symptom based on the effective association analysis graph includes the following sub-steps:

[0030] Perform multinomial regression on the effective correlation analysis graph and name the curve corresponding to the regression function as the effective correlation curve.

[0031] Connect the two ends of an effective associated curve with a straight line segment, and name the connected straight line segment as the effective associated line.

[0032] Draw auxiliary lines, which are infinitely extended straight lines and are perpendicular to the effective associated lines;

[0033] Obtain the intersection points of the auxiliary line with the effective associated straight line and the effective associated curve, and name them as the first intersection point and the second intersection point, respectively;

[0034] Calculate the distance between the first intersection point and the second intersection point, and name it the regression drop distance. Move the auxiliary line along the horizontal axis in ascending order, and calculate the regression drop distance corresponding to each value of the horizontal axis in real time.

[0035] Find the maximum value in the regression drop distance and mark it as the maximum drop distance. Mark the value of the y-axis of the second intersection point corresponding to the maximum drop distance as the association segmentation threshold.

[0036] K n Compare with the associated segmentation threshold, if K n If the value is less than or equal to the correlation segmentation threshold, the output indicates the presence of a correlation signal; otherwise, the output indicates the absence of a correlation signal.

[0037] If the output contains a correlated signal, then K n The corresponding symptoms to be associated are marked as related symptoms of the target symptoms. The ratio of the number of related symptoms to be associated to the number of target symptoms is the degree of association.

[0038] Furthermore, based on the initial symptoms input by the user, and combined with symptom correlation analysis, the predicted symptoms of the user include the following sub-steps:

[0039] Obtain the initial symptoms input by the user, and analyze the co-related symptoms based on the symptom correlation.

[0040] Based on initial symptoms and related symptoms, the system can accurately predict the user's actual condition.

[0041] Further, obtaining the initial symptoms input by the user and analyzing the co-related symptoms based on the symptom correlation includes the following sub-steps:

[0042] Obtain existing symptoms input by the user, mark them as initial symptoms, and assign them numbers using the symbol D. j This indicates that j is a non-zero natural number and j is the index of D;

[0043] Find the associated symptoms of the initial symptoms, for any D j , for D j The corresponding related symptoms are sorted and numbered in descending order of their degree of correlation, denoted by the symbol G(j,h), where h is a non-zero natural number and (j,h) is the index of G, and G(j,h) represents D. j The related symptom ranked hth in the middle;

[0044] Get each D j If there are identical related symptoms, they are marked as co-related symptoms. The sum of h in G(j,h) corresponding to co-related symptoms is obtained and marked as the symptom matching value.

[0045] Symptoms are numbered in ascending order of their symptom match values, using the symbol Q. c Let represent , where c is a non-zero natural number and c is the index of Q.

[0046] Furthermore, based on initial symptoms and related symptoms, accurately predicting the user's actual condition includes the following sub-steps:

[0047] Starting with c=1, Q c The initial symptoms are included, and the initial symptoms are compared with the case data. If there are case data with symptoms that are exactly the same as the initial symptoms, a diagnosis signal is output; otherwise, a diagnosis verification signal is output.

[0048] If a symptom verification signal is output, the case data containing all initial symptoms with the fewest number of individual symptoms is obtained and marked as data to be verified.

[0049] Obtain a single symptom other than the initial symptom from the data to be tested, mark it as the symptom to be tested, check whether all the symptom to be tested are within the same related symptom, if so, output the diagnosis signal, otherwise output the exclusion signal.

[0050] If the output shows a symptom exclusion signal, then Q will be... c Remove from the initial symptoms and increment c by one, then add Q again. c Incorporate initial symptoms and repeatedly compare them with case data, repeating the process until a diagnosis signal is output;

[0051] If a diagnosis signal is output, the corresponding confirmed symptom in the case data will be marked as the user's predicted symptom.

[0052] Secondly, this application provides an event association analysis system based on large model graph rules, including a data acquisition module, a symptom association analysis module, and a symptom prediction module; the data acquisition module and the symptom prediction module are respectively connected to the symptom association analysis module.

[0053] The data acquisition module is used to establish a data connection with medical big data to obtain publicly available case data;

[0054] The symptom association analysis module is used to extract symptom manifestations from case data and analyze the symptom associations between different individual symptoms within the symptom manifestations;

[0055] The symptom prediction module is used to predict the user's symptom based on the initial symptoms input by the user and by analyzing the correlation between symptoms.

[0056] The beneficial effects of this invention are as follows: This invention establishes a data connection with medical big data to obtain publicly available case data, then extracts keywords from symptom manifestations to obtain single symptoms, and analyzes the symptom correlation between different single symptoms within the symptom manifestations. The advantage is that the same disease usually has many different symptoms. By analyzing the symptom correlation between different symptoms, a reliable data foundation and data support can be provided for subsequent prediction of the user's disease, thereby improving the reliability and accuracy of disease event correlation analysis.

[0057] This invention obtains the initial symptoms input by the user, analyzes the correlation between the initial symptoms and related symptoms, and finally makes an accurate prediction of the user's actual condition based on the initial symptoms and related symptoms. The advantage is that, unlike existing medical aids that list multiple possibilities, this invention only provides the most accurate or closest symptom for the user or doctor to refer to, which greatly reduces the time for disease diagnosis and improves the effectiveness and accuracy of the disease time correlation analysis. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the system of the present invention;

[0059] Figure 2 This is a correlation ratio analysis diagram of the present invention;

[0060] Figure 3 This is an effective correlation analysis diagram of the present invention;

[0061] Figure 4 This is a schematic diagram of the effective correlation curve, effective correlation line, and auxiliary lines of the present invention;

[0062] Figure 5 This is a flowchart of the steps of the method of the present invention. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] Example 1, please refer to Figure 1 As shown, this application provides an event association analysis system based on large model graph rules, including a data acquisition module, a symptom association analysis module, and a symptom prediction module; the data acquisition module and the symptom prediction module are respectively connected to the symptom association analysis module.

[0065] The data acquisition module is used to establish a data connection with medical big data and obtain publicly available case data;

[0066] The data acquisition module is configured with data acquisition strategies, which include:

[0067] Establish data connections with medical big data;

[0068] Obtain publicly available case data;

[0069] Case data includes symptoms and confirmed illnesses;

[0070] In practical applications, this embodiment uses existing large-scale modeling techniques to perform the analysis process for symptom prediction. Graph rules refer to a set of rules that describe the logical relationships between nodes such as entities, events, and edges in association analysis based on graph structures, such as knowledge graphs and event networks. They are typically used to represent causal relationships, temporal relationships, spatial relationships, similarities, etc., helping the reasoning system generate new information from existing data or discover implicit associations. In the analysis process of this embodiment, the causal rules in graph rules are followed. The diagnosis of the disease leads to the appearance of symptoms. The diagnosis of the disease and the appearance of symptoms are the events, and the relationship between them is the triggering or causing. Part of the case data is shown in Table 1 below:

[0071] Table 1. Partial data from case studies.

[0072]

[0073] The case data in Table 1 are only used to illustrate the symptoms and the presentation of the diagnosed disease.

[0074] The symptom association analysis module is used to extract symptom manifestations from case data and analyze the symptom associations between different individual symptoms within the symptom manifestations; the symptom association analysis module includes a symptom extraction unit, an association analysis unit, and...

[0075] The symptom extraction unit is used to extract keywords from symptom manifestations to obtain single symptoms;

[0076] The symptom extraction unit is configured with a symptom extraction strategy, which includes:

[0077] Based on NLP keyword extraction technology, keywords are extracted from symptom manifestations to obtain manifestation word segmentation;

[0078] The performance segmentation words are compared with the symptom database, and the performance segmentation words that exist in the symptom database are marked as single symptoms;

[0079] In practical applications, taking the second case data in Table 1 as an example, the symptoms are described as "headache, fever, cough, and palpitations, which may be accompanied by shock". Using existing NLP keyword extraction technology, the segmented words for the symptoms include "headache", "fever", "cough", "palpitations", "may", "accompanied by", and "shock". Punctuation marks are not included in the segmented words. The symptom database contains descriptions of various symptoms, which are used to filter the words describing symptoms in the segmented words. By comparison, the single symptoms are found to include "headache", "fever", "cough", "palpitations", and "shock".

[0080] The association analysis unit is used to analyze the symptom associations between different individual symptoms within the symptom presentation;

[0081] The correlation analysis unit is configured with correlation analysis strategies, which include:

[0082] Symptoms and confirmed diagnosis constitute a case data point, and the number of individual symptoms included in the symptom data points varies from case data point to case data point.

[0083] For any single symptom, mark it as the target symptom, find all case data containing the target symptom, and mark them as target data;

[0084] The number of target symptoms in the target data is marked as the target quantity. For any single symptom in the target data, its quantity is counted and marked as the quantity to be associated. The target quantity and the quantity to be associated are collectively referred to as the symptom quantity.

[0085] In practical applications, taking the single symptom "cough" as an example, "cough" is marked as the target symptom. All case data containing "cough" in the symptom presentation are searched to obtain the target data. A total of 16,872 target data entries are obtained. Each target data entry includes "cough", so the target quantity is 16,872. As shown in Table 1, the first case data entry belongs to the target data, which includes the single symptom "sputum". The number of "sputum" entries in the target data is counted, which is the number of single symptom "sputum" to be associated. The number of single symptom to be associated is different for each single symptom.

[0086] Sort and number the quantities to be associated in ascending order of quantity, using the symbol P. n This represents a sequence where n is a non-zero natural number and n is the index of P. The maximum value of n is incremented by one and marked as N. This is then processed through P. N Indicates the target quantity;

[0087] Please see Figure 2 As shown, a Cartesian coordinate system is established with the numbers 1 and 2 as the X-axis and the number of symptoms as the Y-axis, named the Correlation Proportion Analysis Chart. The coordinate point (2, P) is set as follows. N ) and coordinates (1, P)n Enter the correlation ratio analysis chart;

[0088] Set the coordinates (2, P) N Name it the target point, and set the coordinates (1, P) as the target point. n The target point is named the point to be associated. Connect the target point to each of the different points to be associated using straight line segments to obtain the associated line segments. Then, assign the coordinates (1, P) to the target point. n ) and coordinates (2, P) N The associated line segment is marked as L. n ;

[0089] Get L n The slope, denoted as K n K in ascending order n Number them using the symbol T i Let i represent the sequence number of T, where i is a non-zero natural number and i is the index of T;

[0090] Please see Figure 3 As shown, with i as the horizontal axis, T i Establish a Cartesian coordinate system for the vertical axis, name it the Effective Correlation Analysis Chart, and set T... i Enter the valid association analysis graph according to i;

[0091] In practical applications, numbering yields P n 1 ≤ n ≤ 122, resulting in N = 123. The numbers 1 and 2 on the X-axis represent the single symptom and target symptom corresponding to the number of items to be associated, respectively, and are named symptom codes here. The resulting association ratio analysis chart is shown below. Figure 2 As shown, due to the excessive density of data, Figure 2 The text only shows a small number of points to be associated and associated line segments to clarify the definition of associated line segments, connecting them to obtain L1 to L2. 122 Obtain the slope to get K1 to K 122 K in ascending order n Number them to get T1 to T 122 An effective correlation analysis graph was constructed as follows: Figure 3 As shown;

[0092] Based on the effective association analysis diagram, analyze the symptoms to be associated with the target symptoms;

[0093] Please see Figure 4 As shown, a multinomial regression was performed on the effective correlation analysis graph, and the curve corresponding to the regression function was named the effective correlation curve.

[0094] Connect the two ends of an effective associated curve with a straight line segment, and name the connected straight line segment as the effective associated line.

[0095] Draw auxiliary lines, which are infinitely extended straight lines and are perpendicular to the effective associated lines;

[0096] Obtain the intersection points of the auxiliary line with the effective associated straight line and the effective associated curve, and name them as the first intersection point and the second intersection point, respectively;

[0097] Calculate the distance between the first intersection point and the second intersection point, and name it the regression drop distance. Move the auxiliary line along the horizontal axis in ascending order, and calculate the regression drop distance corresponding to each value of the horizontal axis in real time.

[0098] Find the maximum value in the regression drop distance and mark it as the maximum drop distance. Mark the value of the y-axis of the second intersection point corresponding to the maximum drop distance as the association segmentation threshold.

[0099] K n Compare with the associated segmentation threshold, if K n If the value is less than or equal to the correlation segmentation threshold, the output indicates the presence of a correlation signal; otherwise, the output indicates the absence of a correlation signal.

[0100] If the output contains a correlated signal, then K n The corresponding symptoms to be associated are marked as related symptoms of the target symptoms. The ratio of the number of related symptoms to be associated to the number of target symptoms is the degree of association.

[0101] In practical applications, Figure 4 The solid line segments represent the effective correlation lines, the dashed lines represent the auxiliary lines, and the curves represent the effective correlation curves. Based on the distance calculation formula for coordinate points, the regression drop distance is calculated to be 603. The result is rounded to the nearest integer, and the maximum drop distance is found to be 603. The corresponding correlation segmentation threshold is 9763, K. n Less than or equal to 9763 means that in most cases, K n The corresponding single symptom is very likely to appear simultaneously with the target symptom, showing a strong correlation; conversely, the correlation is weak, rendering it ineffective in medical assistance. Therefore, K... n Single symptoms less than or equal to the association segmentation threshold are labeled as related symptoms of the target symptom, such as P. n The corresponding number of items to be associated is 12344. The ratio of 12344 to the target number is calculated, and the association degree is 0.73. The calculation result is rounded to two decimal places.

[0102] The symptom prediction module is used to predict the user's symptoms based on the initial symptoms input by the user and the correlation analysis of the symptoms; the symptom prediction module includes a correlation symptom extraction unit and a symptom prediction unit.

[0103] The associated symptom extraction unit is used to obtain the initial symptoms input by the user and analyze the related symptoms based on the symptom correlation.

[0104] The associated symptom extraction unit is configured with associated symptom extraction strategies, which include:

[0105] Obtain existing symptoms input by the user, mark them as initial symptoms, and assign them numbers using the symbol D. j This indicates that j is a non-zero natural number and j is the index of D;

[0106] Find the associated symptoms of the initial symptoms, for any D j , for D j The corresponding related symptoms are sorted and numbered in descending order of their degree of correlation, denoted by the symbol G(j,h), where h is a non-zero natural number and (j,h) is the index of G, and G(j,h) represents D. j The related symptom ranked hth in the middle;

[0107] Get each D j If there are identical related symptoms, they are marked as co-related symptoms. The sum of h in G(j,h) corresponding to co-related symptoms is obtained and marked as the symptom matching value.

[0108] Symptoms are numbered in ascending order of their symptom match values, using the symbol Q. c It is represented as follows, where c is a non-zero natural number and c is the index of Q;

[0109] In practical applications, the initial symptoms are those that the user identifies as already present. The initial symptoms obtained include "cough" and "wheezing," numbered D1 and D2 respectively. For D1, among the related symptoms of "cough," the one with the highest correlation is G(1,1), and G(1,1) is "sputum." Since "sputum" also exists in G(2,h), it is considered a related symptom. "Sputum" in G(2,h) is G(2,9). Adding h together gives a symptom matching value of 10. Similarly, the symptom matching values ​​of all related symptoms are analyzed, and then the related symptoms are numbered in ascending order of their matching values, resulting in Q1 to Q2. 23 ;

[0110] The symptom prediction unit is used to accurately predict a user's actual symptom based on initial symptoms and related symptoms.

[0111] The symptom prediction unit is configured with symptom prediction strategies, which include:

[0112] Starting with c=1, Q cThe initial symptoms are included, and the initial symptoms are compared with the case data. If there are case data with symptoms that are exactly the same as the initial symptoms, a diagnosis signal is output; otherwise, a diagnosis verification signal is output.

[0113] If a symptom verification signal is output, the case data containing all initial symptoms with the fewest number of individual symptoms is obtained and marked as data to be verified.

[0114] Obtain a single symptom other than the initial symptom from the data to be tested, mark it as the symptom to be tested, check whether all the symptom to be tested are within the same related symptom, if so, output the diagnosis signal, otherwise output the exclusion signal.

[0115] If the output shows a symptom exclusion signal, then Q will be... c Remove from the initial symptoms and increment c by one, then add Q again. c Incorporate initial symptoms and repeatedly compare them with case data, repeating the process until a diagnosis signal is output;

[0116] If a diagnosis signal is output, the diagnosis of the corresponding case data will be marked as the user's predicted diagnosis.

[0117] In practical applications, when c=1, Q cFor the initial symptom "coughing up phlegm," we include it as part of the initial symptoms. The initial symptoms now include "cough," "wheezing," and "coughing up phlegm." We search for case data where the symptom presentation includes "cough," "wheezing," and "coughing up phlegm." Simultaneously, the case data containing the fewest other single symptoms besides "cough," "wheezing," and "coughing up phlegm" is designated as the test data. For example, the first case data in Table 1 contains "cough," "wheezing," and "coughing up phlegm," but no other single symptoms; therefore, it is marked as test data. Since the test data does not contain any other single symptoms, a diagnosis signal is directly output. If the test symptoms also include the single symptom "shortness of breath," we check whether "shortness of breath" is a related symptom of "cough," "wheezing," and "coughing up phlegm." If so, a diagnosis signal is output; otherwise, an output is not provided. If a symptom exclusion signal is output, the initial symptoms are restored to "cough" and "wheezing," and 'c' is incremented and included back into the initial symptoms for re-analysis. Since this embodiment outputs a symptom diagnosis signal, the diagnosed symptom in the data to be tested is marked as the predicted symptom, resulting in the predicted symptom being chronic bronchitis, and the predicted symptom is output to the user. Because in most cases, when users query the network for their symptoms, the network usually returns multiple results, which cannot provide effective reference, and users' descriptions of symptoms are usually not comprehensive, only uploading part of the symptoms, in order to provide users with a more accurate symptom reference, this embodiment can output the most accurate or closest symptom as a reference based on the user's initial symptoms and at least two related symptoms.

[0118] Example 2, please refer to Figure 5 As shown, this application provides an event correlation analysis method based on large model graph rules, including the following steps:

[0119] Step S1: Establish a data connection with medical big data to obtain publicly available case data; Step S1 includes the following sub-steps:

[0120] Step S101: Establish a data connection with medical big data;

[0121] Step S102: Obtain publicly available case data;

[0122] Step S103, case data includes symptoms and confirmed illness;

[0123] Step S2 involves extracting symptom presentations from the case data and analyzing the symptom correlations between different individual symptoms. Step S2 includes the following sub-steps:

[0124] Step S201: Extract keywords from symptom manifestations to obtain single symptoms;

[0125] Step S201 includes the following sub-steps:

[0126] Step S201.1: Extract keywords from symptom manifestations based on NLP keyword extraction technology to obtain manifestation word segmentation;

[0127] Step S201.2: Compare the performance segmentation with the symptom database, and mark the performance segmentation that exists in the symptom database as a single symptom;

[0128] Step S202: Analyze the symptom correlations between different individual symptoms within the symptom presentation;

[0129] Step S202 includes the following sub-steps:

[0130] Step S202.1, the symptoms and the confirmed disease constitute a case data, and the number of individual symptoms included in the symptoms of different case data is different.

[0131] Step S202.2: For any single symptom, mark it as the target symptom, find all case data containing the target symptom, and mark them as target data;

[0132] Step S202.3: Mark the number of target symptoms in the target data as the target quantity. For any single symptom in the target data, count its quantity and mark it as the quantity to be associated. The target quantity and the quantity to be associated are collectively referred to as the symptom quantity.

[0133] Step S202.4: Sort and number the quantities to be associated in ascending order, using the symbol P. n This represents a sequence where n is a non-zero natural number and n is the index of P. The maximum value of n is incremented by one and marked as N. This is then processed through P. N Indicates the target quantity;

[0134] Step S202.5: Establish a Cartesian coordinate system with the numbers 1 and 2 as the X-axis and the number of symptoms as the Y-axis, and name it the Correlation Proportion Analysis Chart. Then, set the coordinate point (2, P) as the X-axis. N ) and coordinates (1, P) n Enter the correlation ratio analysis chart;

[0135] Step S202.6, set the coordinates of point (2, P) N Name it the target point, and set the coordinates (1, P) as the target point. n The target point is named the point to be associated. Connect the target point to each of the different points to be associated using straight line segments to obtain the associated line segments. Then, assign the coordinates (1, P) to the target point. n ) and coordinates (2, P) N The associated line segment is marked as L. n ;

[0136] Step S202.7, obtain L n The slope, denoted as K n K in ascending order n Number them using the symbol T i Let i represent the sequence number of T, where i is a non-zero natural number and i is the index of T;

[0137] Step S202.8, with i as the horizontal axis, T i Establish a Cartesian coordinate system for the vertical axis, name it the Effective Correlation Analysis Chart, and set T... i Enter the valid association analysis graph according to i;

[0138] Step S202.9: Analyze the symptoms to be associated with the target symptoms based on the effective association analysis graph;

[0139] Step S202.9 includes the following sub-steps:

[0140] Step S202.9.a: Perform multinomial regression on the effective correlation analysis graph and name the curve corresponding to the regression function as the effective correlation curve;

[0141] Step S202.9.b: Connect the two ends of the effective associated curve with a straight line segment, and name the connected straight line segment as the effective associated line;

[0142] Step S202.9.c: Draw auxiliary lines. The auxiliary lines are straight lines that extend infinitely and are perpendicular to the effective associated lines.

[0143] Step S202.9.d: Obtain the intersection points of the auxiliary line with the effective associated straight line and the effective associated curve, and name them as the first intersection point and the second intersection point, respectively;

[0144] Step S202.9.e: Calculate the distance between the first intersection point and the second intersection point, named the regression drop distance. Move the auxiliary line along the horizontal axis in ascending order and calculate the regression drop distance corresponding to each value of the horizontal axis in real time.

[0145] Step S202.9.f: Find the maximum value in the regression drop distance, mark it as the maximum drop distance, and mark the value of the vertical axis of the second intersection point corresponding to the maximum drop distance as the association segmentation threshold;

[0146] Step S202.9.g, K n Compare with the associated segmentation threshold, if K n If the value is less than or equal to the correlation segmentation threshold, the output indicates the presence of a correlation signal; otherwise, the output indicates the absence of a correlation signal.

[0147] In step S202.9.h, if a correlated signal exists in the output, then K... nThe corresponding symptoms to be associated are marked as related symptoms of the target symptoms. The ratio of the number of related symptoms to be associated to the number of target symptoms is the degree of association.

[0148] Step S3, based on the initial symptoms input by the user, combines symptom correlation analysis to predict the user's condition; Step S3 includes the following sub-steps:

[0149] Step S301: Obtain the initial symptoms input by the user, and analyze the related symptoms based on the symptom correlation analysis;

[0150] Step S301 includes the following sub-steps:

[0151] Step S301.1: Obtain the existing symptoms input by the user, mark them as initial symptoms, and assign numbers to the initial symptoms using the symbol D. j This indicates that j is a non-zero natural number and j is the index of D;

[0152] Step S301.2: Find the related symptoms of the initial symptoms, for any D j , for D j The corresponding related symptoms are sorted and numbered in descending order of their degree of correlation, denoted by the symbol G(j,h), where h is a non-zero natural number and (j,h) is the index of G, and G(j,h) represents D. j The related symptom ranked hth in the middle;

[0153] Step S301.3, obtain each D j If there are identical related symptoms, they are marked as co-related symptoms. The sum of h in G(j,h) corresponding to co-related symptoms is obtained and marked as the symptom matching value.

[0154] Step S301.4: Number the associated symptoms according to their symptom matching values ​​in ascending order, and use the symbol Q. c It is represented as follows, where c is a non-zero natural number and c is the index of Q;

[0155] Step S302: Based on the initial symptoms and related symptoms, make an accurate prediction of the user's actual condition.

[0156] Step S302 includes the following sub-steps:

[0157] Step S302.1, starting with c=1, set Q c The initial symptoms are included, and the initial symptoms are compared with the case data. If there are case data with symptoms that are exactly the same as the initial symptoms, a diagnosis signal is output; otherwise, a diagnosis verification signal is output.

[0158] Step S302.2: If a symptom verification signal is output, then obtain the case data that contains all initial symptoms and has the fewest number of individual symptoms, and mark it as data to be verified.

[0159] Step S302.3: Obtain a single symptom other than the initial symptom in the data to be tested, mark it as the symptom to be tested, and check whether all the symptom to be tested are within the same related symptom. If so, output the diagnosis signal; otherwise, output the exclusion signal.

[0160] Step S302.4: If a symptom exclusion signal is output, then Q... c Remove from the initial symptoms and increment c by one, then add Q again. c Incorporate initial symptoms and repeatedly compare them with case data, repeating the process until a diagnosis signal is output;

[0161] Step S302.5: If a disease diagnosis signal is output, then the diagnosed disease in the corresponding case data is marked as the user's predicted disease.

[0162] Example 3: This invention discloses an electronic device, which may include a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The memory stores computer-readable instructions, and the processor can call these instructions. When the processor executes a computer-readable instruction, it performs steps such as those in an event association analysis method based on large model graph rules to achieve the following functions: establishing a data connection with medical big data to obtain publicly available case data; extracting symptom manifestations from the case data and analyzing the symptom correlations between different individual symptoms; and analyzing the user's predicted symptoms based on the initial symptoms input by the user and the symptom correlations.

[0163] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0164] Example 4: This application also provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it performs the steps of the event association analysis method based on large model graph rules as described above to achieve the following functions: establishing a data connection with medical big data to obtain publicly available case data; extracting symptom manifestations from the case data and analyzing the symptom correlations between different single symptoms within the symptom manifestations; and analyzing the user's predicted symptoms based on the initial symptoms input by the user and the symptom correlations.

[0165] Based on the above description of the embodiments, the embodiments of the present invention can be provided as methods, systems, or computer program products. Based on this understanding, the technical solutions described above, or the parts that contribute to the prior art, can be embodied in the form of software products. These computer software products can be stored in computer-readable storage media, such as ROM / RAM, magnetic disks, optical disks, etc., and include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or certain parts of the embodiments.

[0166] In the embodiments provided in this application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules or units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interfaces. The indirect coupling or communication connection between systems, modules, and units may be electrical, mechanical, or other forms.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for event correlation analysis based on large model graph rules, characterized in that, The method comprises the following steps: establishing a data connection with medical big data to obtain publicly available case data; extracting symptom manifestations in the case data and analyzing symptom correlations between different single symptoms in the symptom manifestations; analyzing the user's estimated disease based on the initial symptom input by the user and combining the symptom correlation; analyzing the symptom correlations between different single symptoms in the symptom manifestations comprises the following sub-steps: The symptom manifestations and the diagnosed diseases constitute a piece of case data, and the number of single symptoms contained in the symptom manifestations of different case data is not the same; For any single symptom, mark it as a target symptom, find all case data containing the target symptom, and mark it as target data; Mark the number of target symptoms in the target data as the target number, and for any single symptom in the target data, count the number and mark it as the to-be-associated number. The target number and the to-be-associated number are collectively referred to as the symptom number; The number of to-be-associated quantities is sorted and numbered in ascending order of the number of to-be-associated quantities, and is represented by a symbol P n , where n is a non-zero natural number and n is the serial number of P. The maximum value of n plus one is marked as N, and the target quantity is represented by P N . A plane rectangular coordinate system is established with the numbers 1 and 2 as the X axis and the number of symptoms as the Y axis, and is named as a correlation proportion analysis graph. The coordinate point (2, P N ) and the coordinate point (1, P n ) are recorded in the correlation proportion analysis graph. The coordinate point (2, P N ) is named as a target point, the coordinate point (1, P n ) is named as a point to be associated, the target point is connected with different points to be associated respectively by straight line segments, the associated line segments are obtained, and the associated line segments of the coordinate point (1, P n ) and the coordinate point (2, P N ) are marked as L n ; Acquisition L n of the slope, marked as K n , is numbered in ascending order of K n , denoted by symbol T i , where i is a non-zero natural number and i is the serial number of T; A plane rectangular coordinate system is established with i as the horizontal axis and T i as the vertical axis, and is named as the effective correlation analysis graph. T i is recorded in the effective correlation analysis graph according to i; Based on the effective correlation analysis graph, analyze the to-be-associated symptoms related to the target symptom. 2.The large model graph rule-based incident correlation analysis method according to claim 1, characterized in that, Establishing a data connection with medical big data to obtain publicly available case data comprises the following sub-steps: establishing a data connection with medical big data; obtaining publicly available case data; The case data includes symptom manifestations and diagnosed diseases. 3.The large model graph rule-based incident correlation analysis method according to claim 2, characterized in that, Extracting symptom manifestations in the case data and analyzing symptom correlations between different single symptoms in the symptom manifestations comprises the following sub-steps: Perform keyword extraction on the symptom manifestations to obtain single symptoms; analyze the symptom correlations between different single symptoms in the symptom manifestations. 4.The large model graph rule-based incident correlation analysis method according to claim 3, characterized in that, Perform keyword extraction on the symptom manifestations to obtain single symptoms comprises the following sub-steps: Perform keyword extraction on the symptom manifestations based on NLP keyword extraction technology to obtain performance segmentation; Compare the performance segmentation with the symptom database, and mark the performance segmentation existing in the symptom database as a single symptom. 5.The large model graph rule-based incident correlation analysis method according to claim 4, characterized in that, Based on the effective correlation analysis graph, analyze the to-be-associated symptoms related to the target symptom, which comprises the following sub-steps: Perform polynomial regression on the effective correlation analysis graph, and name the curve corresponding to the regression function as the effective correlation curve; Connect the two ends of the effective correlation curve by a straight line segment, and name the connected straight line segment as the effective correlation straight line; Draw an auxiliary line, which is an infinitely long straight line, and the auxiliary line is perpendicular to the effective correlation straight line; Obtain the intersection points of the auxiliary line and the effective correlation straight line and the effective correlation curve, and name them as the first intersection point and the second intersection point respectively; Calculate the distance between the first intersection point and the second intersection point, and name it as the regression gap distance. Move the auxiliary line along the horizontal axis from small to large, and calculate the regression gap distance corresponding to each value of the horizontal axis in real time; Find the maximum value in the regression gap distance, mark it as the maximum gap distance, and mark the value of the vertical axis of the second intersection point corresponding to the maximum gap distance as the correlation segmentation threshold; K n is compared with an associated segmentation threshold, if K n is less than or equal to the associated segmentation threshold, an associated signal is output, otherwise an unassociated signal is output. If the output exists the association signal, K n The corresponding symptom to be associated is marked as the association symptom of the target symptom, and the ratio of the number of symptoms to be associated and the number of target symptoms is calculated, which is the association degree. 6.The large model graph rule-based incident correlation analysis method according to claim 5, wherein, Based on the initial symptom input by the user, combining the symptom correlation to analyze the user's estimated disease comprises the following sub-steps: Obtain the initial symptom input by the user, and analyze the same associated symptoms of the initial symptom based on the symptom correlation; Based on the initial symptom and the same associated symptoms, accurately estimate the user's actual disease.

7. The large model graph rule-based incident correlation analysis method according to claim 6, characterized in that, Obtaining an initial symptom input by a user, analyzing a same-associated symptom of the initial symptom based on symptom association includes the following sub-steps: Obtaining the existing symptoms input by the user, marked as initial symptoms, numbering the initial symptoms, through the symbol D j denotes, wherein j is a non-zero natural number and j is the serial number of D; Find the associated symptoms of the initial symptoms, for any D j , D j , the corresponding associated symptoms are sorted and numbered in descending order of relevance, represented by the symbol G(j, h), where h is a non-zero natural number and (j, h) is the serial number of G, G(j, h) represents the associated symptoms ranked h in D j ; Obtain the same correlation symptom of each D j Mark the same correlation symptom as the same correlation symptom, obtain the sum of h in G(j, h) corresponding to the same correlation symptom, and mark it as the symptom matching value; The same associated symptoms are numbered in order of the symptom matching value from small to large, by the symbol Q c is indicated, where c is a non-zero natural number and c is the order number of Q. 8.The large model graph rule-based incident correlation analysis method according to claim 7, characterized in that, Based on the initial symptom and the same-associated symptom, accurately estimating the actual disease of the user includes the following sub-steps: With c = 1 as the starting point, Q c If the initial symptom is included in the case data, the case data is compared with the initial symptom, and if there is case data that is exactly the same as the initial symptom, a disease diagnosis signal is output, otherwise a disease verification signal is output. If the disease verification signal is output, obtaining the case data containing all the initial symptoms and the least number of single symptoms, and marking it as the data to be verified; Obtaining the single symptom in the data to be verified except the initial symptom, marking it as the symptom to be verified, and checking whether the symptom to be verified is in the same-associated symptom, if yes, outputting the disease diagnosis signal, otherwise outputting the disease exclusion signal; If an illness exclusion signal is output, Q is incremented by one and the process is repeated. c The initial symptom is eliminated and c is incremented by one, and Q is incremented by one and the process is repeated. c The initial symptom is included and the comparison with the case data is repeated, and the process is repeated until an illness confirmation signal is output. If the disease diagnosis signal is output, marking the corresponding case data as the estimated disease of the user.

9. A large model graph rule-based incident correlation analysis system for implementing the large model graph rule-based incident correlation analysis method of any one of claims 1-8, characterized in that, It comprises a data acquisition module, a symptom association analysis module and a disease estimation module; the data acquisition module and the disease estimation module are respectively connected with the symptom association analysis module; The data acquisition module is used to establish data connection with medical big data and obtain publicly available case data; The symptom association analysis module is used to extract the symptom performance in the case data and analyze the symptom association between different single symptoms in the symptom performance; The disease estimation module is used to estimate the disease of the user based on the initial symptom input by the user and the symptom association.

Images