Knowledge graph-based risk control methods and related equipment
By constructing a knowledge graph-based risk control method, a claims risk control knowledge graph and decision tree model are generated, which solves the problem that traditional models cannot determine the risk of gray list users, achieves more accurate risk control and claims result judgment, and improves the automated risk management capability of insurance business.
Patent Information
- Application Number
- CN202310293468.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-03-23
- Publication Date
- 2025-12-02
- Estimated Expiration
- 2043-03-23
AI Technical Summary
Traditional insurance risk control models cannot effectively determine the claims risk control results of some reporting users, making it difficult to identify users on the gray list and accurately identify them as blacklisted or whitelisted, resulting in incomplete risk control.
A knowledge graph-based risk control method is constructed. By acquiring historical insurance case data and business knowledge data, a claims risk control knowledge graph ontology is generated. Combined with a secondary diagnostic identification model and a decision tree model, a target knowledge graph claims risk control model is constructed to identify information data of gray list users and to make risk judgments through graph relationships.
It improves the automated risk control capabilities of insurance business, can accurately identify the risks of gray-list users, reduce the risk of false disclosure, and reduce claims losses.
Smart Images
Figure CN116308823B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, and in particular to risk control methods and related equipment based on knowledge graphs. Background Technology
[0002] In the health insurance business of the insurance industry, most of the external expenditures of insurance companies are claims expenditures. That is, when a user purchases health insurance and goes to the hospital for treatment after falling ill, the user claims the medical expenses. Therefore, risk control for health insurance claims is a very important aspect. Avoiding risks, reducing claims and increasing efficiency are the ultimate goals of automated risk control.
[0003] Traditional insurance risk control models are misrepresentation risk control models, primarily used to determine whether a claimant has made a false disclosure. However, in addition to outputting blacklisted and whitelisted claimants, these models also output a gray list of claimants whose claims cannot be determined. This means that traditional insurance risk control models cannot provide the corresponding claims risk control results for some claimants. Summary of the Invention
[0004] This application provides a knowledge graph-based risk control method, apparatus, computer equipment, and storage medium to address the problem that traditional insurance risk control models cannot obtain the claims risk control results for some reporting users.
[0005] The first aspect of this application provides a risk control method based on knowledge graphs, including:
[0006] Obtain data on historical insurance cases associated with the target health insurance business, as well as first knowledge graph data on business knowledge data associated with the target health insurance business, and use the data on historical insurance cases and the first knowledge graph data as data to be processed;
[0007] Construct a target claims risk control knowledge graph ontology based on ontology construction requirements, and obtain ontology field data contained in the target claims risk control knowledge graph ontology and ontology association relationships between the target claims risk control knowledge graph ontology from the data to be processed.
[0008] The disease description data of the target claims risk control knowledge graph ontology is input into the trained sub-diagnosis entity recognition model, and the corresponding sub-diagnosis recognition result is output. The sub-diagnosis recognition result includes disease entities, and / or examination entities, and / or treatment entities.
[0009] Import the target claims risk control knowledge graph ontology, the ontology associations, and the sub-diagnosis identification results into a preset first graph database to obtain the target claims risk control knowledge graph;
[0010] Construct a target knowledge graph-based claims risk control model based on the target claims risk control knowledge graph and decision tree model;
[0011] Information data of gray-list users identified by the false disclosure risk control model is obtained as data to be judged. The false disclosure risk control model is an artificial intelligence claims risk control model built on XGBoost. The information data of gray-list users cannot be identified by the false disclosure risk control model as black-list user information data or white-list user information data.
[0012] Input the information data of the gray list users into the target knowledge graph claims risk control model, and output the target claims risk judgment result corresponding to the gray list users.
[0013] A second aspect of this application provides a knowledge graph-based risk control device, comprising:
[0014] The first data acquisition module is used to acquire data of historical insurance cases associated with the target health insurance business, as well as the first knowledge graph data of business knowledge data associated with the target health insurance business, and to use the data of historical insurance cases and the first knowledge graph data as data to be processed.
[0015] The second data acquisition module is used to construct the target claims risk control knowledge graph ontology according to the ontology construction requirements, and to obtain the ontology field data contained in the target claims risk control knowledge graph ontology and the ontology association relationship between the target claims risk control knowledge graph ontology from the data to be processed.
[0016] The secondary diagnosis identification module is used to input the disease description content data of the target claims risk control knowledge graph ontology into the trained secondary diagnosis entity identification model and output the corresponding secondary diagnosis identification result, wherein the secondary diagnosis identification result includes disease entity, and / or examination entity, and / or treatment entity;
[0017] The claims risk control knowledge graph module is used to import the target claims risk control knowledge graph ontology, the ontology association relationship and the sub-diagnosis identification result into the preset first graph database to obtain the target claims risk control knowledge graph;
[0018] The claims risk control model module is used to construct a target knowledge graph claims risk control model based on the target claims risk control knowledge graph and decision tree model.
[0019] The third data acquisition module is used to acquire information data of gray-list users identified by the false disclosure risk control model as data to be judged. The false disclosure risk control model is an artificial intelligence claims risk control model built on XGBoost. The information data of gray-list users cannot be identified by the false disclosure risk control model as black-list user information data or white-list user information data.
[0020] The claims risk assessment module is used to input the information data of the gray list users into the target knowledge graph claims risk control model and output the target claims risk assessment results corresponding to the gray list users.
[0021] A third aspect of this application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the knowledge graph-based risk control method described above.
[0022] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the knowledge graph-based risk control method described above.
[0023] The aforementioned knowledge graph-based risk control method, device, computer equipment, and storage medium acquire and process historical insurance case data and related business knowledge data associated with the target health insurance business to generate a corresponding target claims risk control knowledge graph. A target knowledge graph claims risk control model is then constructed using this knowledge graph and a decision tree model, and this model is used to output claims risk control results. This not only solves the problem that traditional insurance risk control models cannot obtain claims risk control results for some reporting users, but also improves the automated risk control capabilities of insurance business, further mitigating risks, reducing claims, and increasing efficiency. Attached Figure Description
[0024] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0025] Figure 1 This is a schematic diagram of an application environment for a knowledge graph-based risk control method in one embodiment of this application;
[0026] Figure 2 This is a flowchart of a knowledge graph-based risk control method in one embodiment of this application;
[0027] Figure 3 This is a schematic diagram of the structure of a knowledge graph-based risk control device in one embodiment of this application;
[0028] Figure 4 This is a schematic diagram of a computer device according to one embodiment of this application. Detailed Implementation
[0029] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0030] The knowledge graph-based risk control method provided in this application can be applied to, for example... Figure 1 In this application environment, the computer equipment can be, but is not limited to, various personal computers and laptops. The computer equipment can also be a server, which can be a standalone server or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. This is understandable. Figure 1 The number of computer devices shown is merely illustrative and can be expanded in any number according to actual needs.
[0031] In one embodiment, such as Figure 2 As shown, a knowledge graph-based risk control method is provided, which is then applied to... Figure 1 The following steps, S101 to S107, are used as an example of computer equipment in the example:
[0032] S101. Obtain data of historical insurance cases associated with the target health insurance business, and first knowledge graph data of business knowledge data associated with the target health insurance business, and use the data of historical insurance cases and the first knowledge graph data as data to be processed.
[0033] S102. Construct the target claims risk control knowledge graph ontology according to the ontology construction requirements, and obtain the ontology field data contained in the target claims risk control knowledge graph ontology and the ontology association relationships between the target claims risk control knowledge graph ontology from the data to be processed.
[0034] For example, the health insurance business of a certain fintech platform generates a target claims risk control knowledge graph ontology containing fields such as: insured number, case number, bill number, disease item, disease description, case closure time, consultation end time, false disclosure model prediction result, whether the adjustment is genuine, product name, policy effective time, whether it is renewed, outpatient or inpatient.
[0035] S103. Input the disease description content data of the target claims risk control knowledge graph ontology into the trained sub-diagnosis entity recognition model, and output the corresponding sub-diagnosis recognition result, wherein the sub-diagnosis recognition result includes disease entity, and / or examination entity, and / or treatment entity.
[0036] S104. Import the target claims risk control knowledge graph ontology, the ontology association relationship, and the sub-diagnosis identification result into the preset first graph database to obtain the target claims risk control knowledge graph.
[0037] Graph databases (GDBs) are non-relational databases that use graph structures for semantic queries. They use ontology, relations, and attributes to represent and store data. A graph database directly associates stored data items with the data ontology and the set of edges representing relations between the ontology. Furthermore, in this embodiment, the first graph database may be, but is not limited to, Neo4j, OrientDB, ArangoDB, JanusGraph, Dgraph, HugeGraph, etc.
[0038] S105. Construct a target knowledge graph claims risk control model based on the target claims risk control knowledge graph and decision tree model.
[0039] The target claims risk control knowledge graph contains historical insurance case data and business knowledge data, while the decision tree model includes processes for data processing, calculation, and analysis. Specifically, it includes data processing methods (e.g., word segmentation of disease description text), specific algorithms for data calculation (e.g., algorithms for calculating disease correlation), analysis steps for data analysis, and judgment rules for data judgment. During operation, the decision tree model receives data from both the current case data and the target claims risk control knowledge graph. The target claims risk control knowledge graph provides data support for the decision tree model's processing of the current case. Therefore, the target claims risk control knowledge graph and the decision tree model are used together to construct the target knowledge graph claims risk control model.
[0040] S106. Obtain information data of gray-list users identified by the false disclosure risk control model as data to be judged. The false disclosure risk control model is an artificial intelligence claims risk control model built on XGBoost. The information data of gray-list users cannot be identified by the false disclosure risk control model as black-list user information data or white-list user information data.
[0041] Furthermore, after obtaining the gray list user information data identified by the false disclosure risk control model as the data to be judged, the process further includes: first, obtaining the corresponding historical case data based on the unique user code contained in the gray list user information data. Then, if the historical case data is empty, the gray list user information data and the corresponding gray list judgment result are directly output, and the gray list user information data is not input into the target knowledge graph claims risk control model.
[0042] S107. Input the information data of the gray list users into the target knowledge graph claims risk control model, and output the target claims risk judgment result corresponding to the gray list users.
[0043] Specifically, the step of inputting the information data of the gray-listed users into the target knowledge graph claims risk control model and outputting the target claims risk judgment result corresponding to the gray-listed users includes: First, obtaining the historical case data of the corresponding gray-listed users based on their unique user codes. Then, determining the first graph relationship between the current claim content of the gray-listed users and the historical case data of the gray-listed users based on the target claims risk knowledge graph. Finally, the decision tree model calculates the target claims risk judgment result of the gray-listed users based on the first graph relationship and the target claims risk control knowledge graph.
[0044] Specifically, the step of determining the first graph relationship between the current claim content of the gray-listed user and the historical case data of the gray-listed user based on the target claims risk knowledge graph includes: First, obtaining a first judgment result regarding whether the disease items and the secondary diagnosis identification results in the corresponding historical case data of the gray-listed user information data are the same or contain the same characters. Second, obtaining a second judgment result regarding whether the current claim content of the gray-listed user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data have an inclusion relationship. Third, obtaining a third judgment result regarding whether the current claim content of the gray-listed user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data have a sibling relationship. Fourth, obtaining a fourth judgment result regarding whether the current claim content of the gray-listed user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data have an association relationship. Fifth, obtaining a fifth judgment result regarding whether the current claim content of the gray-listed user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data have a similarity relationship. Finally, the first judgment result, the second judgment result, the third judgment result, the fourth judgment result, and the fifth judgment result are added to the first graph relationship.
[0045] In a more specific embodiment, firstly, the first identifier variable corresponding to the first judgment result is set to `old_flag`, the second identifier variable corresponding to the second judgment result is set to `included_flag`, the third identifier variable corresponding to the third judgment result is set to `brother_flag`, the fourth identifier variable corresponding to the fourth judgment result is set to `related_flag`, and the fifth identifier variable corresponding to the fifth judgment result is set to `alike_flag`. The values of the five identifier variables—`old_flag`, `included_flag`, `brother_flag`, `related_flag`, and `alike_flag`—are 1, 2, or 3. A value of 1 indicates a judgment result of "promotion", a value of 2 indicates a judgment result of "pass", and a value of 3 indicates a judgment result of "cannot be judged". Further, the initial values of the first, second, third, fourth, and fifth identifier variables are all 0.
[0046] Further, in the more specific embodiment, a sixth identifier variable is set as `old_get_flag`, which indicates whether the first judgment result is empty; a seventh identifier variable is set as `old_2_num`, which indicates the first number of times a historical case is judged as invalid in the first judgment result; an eighth identifier variable is set as `included_get_flag`, which indicates whether the second judgment result is empty; a ninth identifier variable is set as `included_2_num`, which indicates the second number of times a historical case is judged as invalid in the second judgment result; a tenth identifier variable is set as `brother_get_flag`, which indicates whether the third judgment result is empty; and an eleventh identifier variable is set as `brother_2_num`, which indicates the third number of times a historical case is judged as invalid in the third judgment result. The twelfth identifier variable is set to `related_get_flag`, which indicates whether the fourth judgment result is empty. The thirteenth identifier variable is set to `related_2_num`, which indicates the fourth number of times a historical case was judged as invalid in the fourth judgment result. The fourteenth identifier variable is set to `alike_get_flag`, which indicates whether the fifth judgment result is empty. The fifteenth identifier variable is set to `alike_2_num`, which indicates the fifth number of times a historical case was judged as invalid in the fifth judgment result. Further, the initial values of the sixth, seventh, eighth, ninth, tenth, eleventh, twelfth, thirteenth, fourteenth, and fifteenth identifier variables are all 0.
[0047] Furthermore, in the more specific embodiment, the process of data processing, calculation, and judgment by the decision tree model includes the following steps S201 to S222:
[0048] S201. Traverse the historical case data of the graylisted users and determine whether the first determination result is empty. If not, assign the value of old_get_flag to 1.
[0049] S202. Determine whether the end time of the consultation for the currently traversed historical case is greater than the consultation time for the current incident. If yes, set the old_flag value to 1. If no, proceed to step S203.
[0050] S203. Determine whether the insurance product of the historical case being traversed is the same as the insurance product of the current claim. If not, proceed to step S204; if yes, proceed to step S205.
[0051] S204. Determine whether the currently traversed historical case has been transferred. If not, and the old_flag is not equal to 1, then assign the old_flag value to 3; if yes, then assign the old_flag value to 1, obtain the historical cases that have not been traversed from the gray list user historical case data to update the currently traversed historical case, and execute step S202.
[0052] S205. Determine whether the effective date of the policy of the historical case being traversed is less than the effective date of the current claim. If yes, proceed to step S206; otherwise, proceed to step S214.
[0053] S206. Determine whether the old_flag is equal to 1. If yes, proceed to step S207. If yes, obtain the historical cases that have not been traversed from the historical case data of the gray list users and update the currently traversed historical cases, then proceed to step S202.
[0054] S207. Determine whether the currently traversed historical cases are renewed. If yes, proceed to step S208. If no, assign the value of old_flag to 1, and obtain the historical cases that have not been traversed from the gray list user historical case data to update the currently traversed historical cases, and proceed to step S202.
[0055] S208. Determine whether the currently traversed historical case has been transferred. If transferred, proceed to step S210. If not transferred, increment old_2_num by 1. Determine whether the time difference between the policy effective date of the currently traversed historical case and the policy effective date of the current claim is less than 1 year, and the time difference between the medical treatment end date of the currently traversed historical case and the effective date of the current claim is less than 1 year, and old_flag is not equal to 3. If it is less than 1 year, assign old_flag the value of 2, and obtain the untraversed historical cases from the gray list user historical case data to update the currently traversed historical case, and proceed to step S202. If it is not less than 1 year, proceed to step S209.
[0056] S209. Determine whether old_2_num is greater than 2. If yes, assign old_flag the value 2, and obtain the historical cases that have not been traversed from the historical case data of the gray list users to update the currently traversed historical cases, and execute step S202; if no, assign old_flag the value 3.
[0057] S210. Subtract 1 from old_2_num, and then determine whether old_2_num is greater than or equal to 2. If yes, assign the value of old_flag to 2, and obtain the historical cases that have not been traversed from the historical case data of the gray list users to update the currently traversed historical cases, and execute step S202; if no, assign the value of old_flag to 3, and execute step S211.
[0058] S211. If, in step S205, the policy effective time of the currently traversed historical case is equal to the effective time of the current claim, determine whether the old_flag is equal to 1. If yes, then obtain the historical cases that have not been traversed from the gray list user's historical case data to update the currently traversed historical case, and execute step S202; otherwise, execute step S212.
[0059] S212. If no historical case is currently being traversed, increment old_2_num by 1; determine whether the time difference between the end time of the consultation for the currently traversed historical case and the policy effective time of the current claim is less than 1 year and old_flag is not equal to 3. If it is less than 1 year, assign old_flag the value of 2, and obtain untraversed historical cases from the gray list user historical case data to update the currently traversed historical case, and execute step S202; if it is not less than 1 year, execute step S213.
[0060] S213. If no historical case is currently being traversed, then the old_flag value is set to 3, and then step S214 is executed.
[0061] S214. Determine whether the effective date of the policy of the current historical case being traversed is less than the effective date of the current claim. If not, assign the value of old_flag to 1, and then execute step S215.
[0062] S215. If the end time of the medical treatment for the currently traversed historical case is equal to the medical treatment time of the current claim, then determine whether the insurance product of the currently traversed historical case is the same as the insurance product of the current claim. If not, proceed to step S216.
[0063] S216. Determine whether the currently traversed historical case has been transferred. If yes, set the old_flag value to 1, and obtain the historical cases that have not been traversed from the gray list user historical case data to update the currently traversed historical case, and execute step S202; if no, set the old_flag value to 3, and then execute step S217.
[0064] S217. If the insurance product of the currently traversed historical case is the same as the insurance product of the current claim, then further determine whether old_flag is equal to 1. If yes, then obtain the historical cases that have not been traversed from the gray list user's historical case data to update the currently traversed historical case, and execute step S202; if no, then execute step S218.
[0065] S218. Determine whether the currently traversed historical case has been transferred. If yes, assign the value of old_flag to 1; otherwise, increment old_2_num by 1, and then execute step S219.
[0066] S219. Determine whether the time difference between the end time of the current medical visit of the historical case being traversed and the effective time of the policy of the current claim is less than 1 year, and whether old_flag is not equal to 3. If yes, then assign old_flag the value of 2, and obtain the historical cases that have not been traversed from the historical case data of the gray list users to update the historical cases being traversed, and execute step S202; if no, then execute step S220.
[0067] S220. Continue to traverse the historical case data of the gray list users until all historical cases included in the historical case data of the gray list users have been judged. Before each traversal, if the old_flag is equal to 3, then reset the old_flag to 0. If the old_flag is equal to 0 and the old_get_flag is equal to 1, then reset the old_flag to 2.
[0068] S221. Based on the above steps, continue processing the second judgment result, the third judgment result, and the fourth judgment result, wherein old_get_flag corresponds to included_get_flag, brother_get_flag, related_get_flag, and alike_get_flag; old_flag corresponds to included_flag, brother_flag, related_flag, and alike_flag; and old_2_num corresponds to included_2_num, brother_2_num, related_2_num, and alike_2_num.
[0069] S222. Count the number of values of old_flag, included_flag, brother_flag, related_flag, and alike_flag that are 1 as black_num; count the number of values of old_flag, included_flag, brother_flag, related_flag, and alike_flag that are 2 as white_num; and count the number of values of old_flag, included_flag, brother_flag, related_flag, and alike_flag that are 0 as grey_num.
[0070] S223. Set the judgment result as result_flag. If old_flag equals 2, then assign the value 2 to result_flag; if black_num equals 1 and brother_flag is 1 or alike_flag is 1, then assign the value 0 to result_flag; otherwise, assign the value 1 to result_flag; if black_num is greater than 1, then assign the value 1 to result_flag; if grey_num equals 5, then assign the value 0 to result_flag; if black_num equals 0, or if white_num is 1, or if brother_flag or alike_flag is 2, then assign the value 0 to result_flag; otherwise, assign the value 2 to result_flag. Wherein, a judgment result_flag of 0 indicates gray, a judgment result_flag of 1 indicates black, and a judgment result_flag of 2 indicates white.
[0071] Furthermore, after inputting the information data of the gray-listed users into the target knowledge graph claims risk control model and outputting the target claims risk judgment result corresponding to the gray-listed users, the process further includes: first, obtaining the case closure result data of the claims cases corresponding to the target claims risk judgment result, and importing the case closure result data into the first graph database to update the target claims risk control knowledge graph. Then, obtaining the comparison result data between the target claims risk judgment result and the case closure result data, and optimizing the parameters of the misrepresentation risk control model and the decision tree model based on the comparison result data.
[0072] Further, after obtaining the case closure data of the claims corresponding to the target claims risk judgment result, the method further includes: First, calculating the first statistical result of the output results of the false disclosure risk control model and the target knowledge graph claims risk control model. The first statistical result includes the black-mark hit rate, black-mark accuracy rate, black-mark coverage rate, white-mark hit rate, white-mark accuracy rate, and white-mark coverage rate. The hit rate is the ratio of claims judged as black or white to the total number of historical insurance cases; the accuracy rate is the ratio of claims judged as black or white (which is the correct result) to the total number of claims judged as black or white; and the coverage rate is the ratio of claims judged as black or white (which is the correct result) to the total number of claims judged as black or white. Then, setting a first statistical result threshold range corresponding to the first statistical result, creating a monitoring thread to monitor whether the first statistical result exceeds the first statistical result threshold range, and if so, generating a warning message containing the first statistical result and the corresponding first statistical result threshold range.
[0073] The knowledge graph-based risk control method provided in this embodiment acquires and processes historical insurance case data and related business knowledge data associated with the target health insurance business to generate a corresponding target claims risk control knowledge graph. A target knowledge graph claims risk control model is then constructed using this knowledge graph and a decision tree model, and this model is used to output claims risk control results. This not only solves the problem that traditional insurance risk control models cannot obtain claims risk control results for some reporting users, but also improves the automated risk control capabilities of insurance business, further mitigating risks and reducing claims while increasing efficiency.
[0074] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.
[0075] In one embodiment, a knowledge graph-based risk control device 100 is provided, which corresponds one-to-one with the knowledge graph-based risk control methods described in the above embodiments. For example... Figure 3 As shown, the knowledge graph-based risk control device 100 includes a first data acquisition module 11, a second data acquisition module 12, a secondary diagnostic identification module 13, a claims risk control knowledge graph module 14, a claims risk control model module 15, a third data acquisition module 16, and a claims risk assessment module 17. Detailed descriptions of each functional module are as follows:
[0076] The first data acquisition module 11 is used to acquire data of historical insurance cases associated with the target health insurance business, and the first knowledge graph data of business knowledge data associated with the target health insurance business, and to use the data of historical insurance cases and the first knowledge graph data as data to be processed.
[0077] The second data acquisition module 12 is used to construct a target claims risk control knowledge graph ontology according to the ontology construction requirements, and to obtain ontology field data contained in the target claims risk control knowledge graph ontology and ontology association relationships between the target claims risk control knowledge graph ontology from the data to be processed.
[0078] The secondary diagnosis identification module 13 is used to input the disease description content data of the target claims risk control knowledge graph ontology into the trained secondary diagnosis entity identification model and output the corresponding secondary diagnosis identification result, wherein the secondary diagnosis identification result includes disease entity, and / or examination entity, and / or treatment entity;
[0079] The claims risk control knowledge graph module 14 is used to import the target claims risk control knowledge graph ontology, the ontology association relationship and the sub-diagnosis identification result into the preset first graph database to obtain the target claims risk control knowledge graph;
[0080] Claims risk control model module 15 is used to construct a target knowledge graph claims risk control model based on the target claims risk control knowledge graph and decision tree model;
[0081] The third data acquisition module 16 is used to acquire information data of gray list users identified by the false disclosure risk control model as data to be judged. The false disclosure risk control model is an artificial intelligence claims risk control model built on XGBoost. The information data of gray list users cannot be identified by the false disclosure risk control model as black list user information data or white list user information data.
[0082] The claims risk assessment module 17 is used to input the information data of the gray list users into the target knowledge graph claims risk control model and output the target claims risk assessment result corresponding to the gray list users.
[0083] Furthermore, the claims risk assessment module 17 also includes:
[0084] The gray list user data submodule is used to obtain the corresponding gray list user's historical case data based on the gray list user's unique user code.
[0085] The first graph relationship submodule is used to determine the first graph relationship between the current claim content of the gray list user and the historical case data of the gray list user based on the target claim risk knowledge graph.
[0086] The claims risk calculation submodule is used by the decision tree model to calculate the target claims risk judgment result of the gray list user based on the first graph relationship and the target claims risk control knowledge graph.
[0087] Furthermore, the first graph relation submodule also includes:
[0088] The first judgment result subunit is used to obtain the first judgment result of whether the current accident content corresponding to the gray list user information data is the same as or contains the sub-diagnosis identification result in the corresponding historical case data;
[0089] The second judgment result subunit is used to obtain a second judgment result on whether the current accident content corresponding to the gray list user information data and the disease items and the sub-diagnosis identification results in the corresponding historical case data are inclusive;
[0090] The third judgment result subunit is used to obtain a third judgment result on whether there is a sibling relationship between the current accident content corresponding to the gray list user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data;
[0091] The fourth judgment result subunit is used to obtain a fourth judgment result on whether there is a correlation between the current accident content corresponding to the gray list user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data;
[0092] The fifth judgment result subunit is used to obtain a fifth judgment result on whether there is a similarity between the current accident content corresponding to the gray list user information data and the disease items and the sub-diagnosis identification results in the corresponding historical case data;
[0093] The first graph relation generation subunit is used to add the first judgment result, the second judgment result, the third judgment result, the fourth judgment result, and the fifth judgment result to the first graph relation.
[0094] Furthermore, the third data acquisition module 16 also includes:
[0095] The fourth data acquisition submodule is used to obtain the corresponding historical case data based on the unique user code contained in the gray list user information data;
[0096] The historical data void detection submodule is used to directly output the gray list user information data and the corresponding gray list detection result if the historical case data is empty, without inputting the gray list user information data into the target knowledge graph claims risk control model.
[0097] Furthermore, the claims risk assessment module 17 also includes:
[0098] The knowledge graph update submodule is used to obtain the case closure result data of the claims case corresponding to the target claims risk judgment result, and import the case closure result data into the first graph database to update the target claims risk control knowledge graph;
[0099] The parameter optimization submodule is used to obtain the comparison result data between the target claim risk assessment result and the case closure result data, and optimize the parameters of the false disclosure risk control model and the decision tree model based on the comparison result data.
[0100] Furthermore, the knowledge graph update submodule also includes:
[0101] The first statistical results submodule is used to calculate the first statistical results of the output results of the false disclosure risk control model and the target knowledge graph claims risk control model. The first statistical results include the black-marking hit rate, black-marking accuracy rate, black-marking coverage rate, white-marking hit rate, white-marking accuracy rate, and white-marking coverage rate. Among them, the hit rate is the ratio of claims cases judged as black or white to the historical insurance cases, the accuracy rate is the ratio of claims cases judged as black or white (which is the correct result) to the total number of claims cases judged as black or white, and the coverage rate is the ratio of claims cases judged as black or white (which is the correct result) to the total number of claims cases judged as black or white.
[0102] The warning information generation submodule is used to set the threshold range of the first statistical result corresponding to the first statistical result, create a monitoring thread to monitor whether the first statistical result exceeds the threshold range of the first statistical result, and if so, generate a warning information containing the first statistical result and the corresponding threshold range of the first statistical result.
[0103] The terms "first" and "second" in the above-mentioned modules / units are only used to distinguish different modules / units and are not intended to specify which module / unit has a higher priority or any other limiting meaning. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or modules is not necessarily limited to those steps or modules explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. The module divisions appearing in this application are merely logical divisions; in actual applications, different division methods may be used.
[0104] For specific limitations regarding knowledge graph-based risk control devices, please refer to the limitations of knowledge graph-based risk control methods mentioned above, which will not be repeated here. Each module in the aforementioned knowledge graph-based risk control device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.
[0105] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data involved in a knowledge graph-based risk control method. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a knowledge graph-based risk control method.
[0106] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the knowledge graph-based risk control method described in the above embodiments, for example... Figure 2 The steps S101 to S107 shown, as well as other extensions and related steps of the method, are examples. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit of the knowledge graph-based risk control device in the above embodiments, for example... Figure 3 The functions of modules 11 to 17 are shown. To avoid repetition, they will not be described again here.
[0107] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the computer device, connecting various parts of the computer device via various interfaces and lines.
[0108] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the mobile phone (such as audio data, video data, etc.).
[0109] The memory can be integrated into the processor or it can be set up separately from the processor.
[0110] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When executed by a processor, the computer program implements the steps of the knowledge graph-based risk control method described in the above embodiments, for example... Figure 2 The steps S101 to S107 shown, as well as other extensions and related steps of the method, are examples. Alternatively, when a computer program is executed by a processor, it implements the functions of each module / unit of the knowledge graph-based risk control device in the above embodiments, for example... Figure 3 The functions of modules 11 to 17 are shown. To avoid repetition, they will not be described again here.
[0111] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0112] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0113] The above-described 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, and should all be included within the protection scope of this application.
Claims
1. A risk control method based on knowledge graphs, characterized in that, include: Obtain data on historical insurance cases associated with the target health insurance business, as well as first knowledge graph data on business knowledge data associated with the target health insurance business, and use the data on historical insurance cases and the first knowledge graph data as data to be processed; Construct a target claims risk control knowledge graph ontology based on ontology construction requirements, and obtain ontology field data contained in the target claims risk control knowledge graph ontology and ontology association relationships between the target claims risk control knowledge graph ontology from the data to be processed. The disease description data of the target claims risk control knowledge graph ontology is input into the trained sub-diagnosis entity recognition model, and the corresponding sub-diagnosis recognition result is output. The sub-diagnosis recognition result includes disease entities, and / or examination entities, and / or treatment entities. Import the target claims risk control knowledge graph ontology, the ontology associations, and the sub-diagnosis identification results into a preset first graph database to obtain the target claims risk control knowledge graph; A target knowledge graph claims risk control model is constructed based on the target claims risk control knowledge graph and decision tree model. The target claims risk control knowledge graph contains data on historical insurance cases and business knowledge data. The decision tree model includes processing methods for data processing, specific algorithms for data calculation, analysis steps for data analysis, and judgment rules for data judgment. During the operation of the decision tree model, the data comes from the input data of the current case to be judged and the target claims risk control knowledge graph. The target claims risk control knowledge graph provides data support for the decision tree model to process the current case to be judged. Information data of gray-list users identified by the false disclosure risk control model is obtained as data to be judged. The false disclosure risk control model is an artificial intelligence claims risk control model built on XGBoost. The information data of gray-list users cannot be identified by the false disclosure risk control model as black-list user information data or white-list user information data. Input the information data of the gray list users into the target knowledge graph claims risk control model, and output the target claims risk judgment result corresponding to the gray list users; The process of inputting the information data of the gray-listed users into the target knowledge graph claims risk control model and outputting the target claims risk judgment result corresponding to the gray-listed users includes: obtaining the corresponding historical case data of the gray-listed users based on their unique user codes; determining the first graph relationship between the current claims of the gray-listed users and their historical case data based on the target claims risk knowledge graph; and the decision tree model calculating the target claims risk judgment result of the gray-listed users based on the first graph relationship and the target claims risk control knowledge graph. The step of determining the first graph relationship between the current claim content of the gray-listed user and the historical case data of the gray-listed user based on the target claims risk knowledge graph includes: obtaining a first judgment result on whether the disease items and the secondary diagnosis identification results in the corresponding historical case data of the gray-listed user information data are the same or contain the same characters; obtaining a second judgment result on whether the current claim content of the gray-listed user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data have an inclusion relationship; obtaining a third judgment result on whether the current claim content of the gray-listed user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data have a sibling relationship; obtaining a fourth judgment result on whether the current claim content of the gray-listed user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data have an association relationship; obtaining a fifth judgment result on whether the current claim content of the gray-listed user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data have a similarity relationship; and adding the first judgment result, the second judgment result, the third judgment result, the fourth judgment result, and the fifth judgment result to the first graph relationship.
2. The risk control method based on knowledge graphs according to claim 1, characterized in that, After obtaining the gray list user information data identified by the false disclosure risk control model as the data to be judged, the method further includes: The corresponding historical case data is obtained based on the unique user code contained in the gray list user information data. If the historical case data is empty, the gray list user information data and the corresponding gray list judgment result are directly output, and the gray list user information data is not input into the target knowledge graph claims risk control model.
3. The risk control method based on knowledge graphs according to claim 1, characterized in that, After inputting the information data of the gray list users into the target knowledge graph claims risk control model and outputting the target claims risk judgment result corresponding to the gray list users, the method further includes: Obtain the case closure result data of the claims cases corresponding to the target claims risk assessment result, and import the case closure result data into the first graph database to update the target claims risk control knowledge graph; Obtain the comparison results between the target claim risk assessment results and the case closure results data, and optimize the parameters of the false disclosure risk control model and the decision tree model based on the comparison results data.
4. The knowledge graph-based risk control method according to claim 3, characterized in that, After obtaining the case closure result data of the claim cases corresponding to the target claim risk assessment result, the method further includes: The first statistical result of the output results of the false disclosure risk control model and the target knowledge graph claims risk control model includes the black detection hit rate, black detection accuracy, black detection coverage, white detection hit rate, white detection accuracy, and white detection coverage. Among them, the hit rate is the ratio of claims judged as black or white to the total number of historical insurance cases, the accuracy rate is the ratio of claims judged as black or white (which is the correct result) to the total number of claims judged as black or white, and the coverage rate is the ratio of claims judged as black or white (which is the correct result) to the total number of claims judged as black or white (which is the actual result). Set the threshold range of the first statistical result corresponding to the first statistical result, create a monitoring thread to monitor whether the first statistical result exceeds the threshold range of the first statistical result, and if so, generate a warning message containing the first statistical result and the corresponding threshold range of the first statistical result.
5. A risk control device based on a knowledge graph, characterized in that, include: The first data acquisition module is used to acquire data of historical insurance cases associated with the target health insurance business, as well as the first knowledge graph data of business knowledge data associated with the target health insurance business, and to use the data of historical insurance cases and the first knowledge graph data as data to be processed. The second data acquisition module is used to construct the target claims risk control knowledge graph ontology according to the ontology construction requirements, and to obtain the ontology field data contained in the target claims risk control knowledge graph ontology and the ontology association relationship between the target claims risk control knowledge graph ontology from the data to be processed. The secondary diagnosis identification module is used to input the disease description content data of the target claims risk control knowledge graph ontology into the trained secondary diagnosis entity identification model and output the corresponding secondary diagnosis identification result, wherein the secondary diagnosis identification result includes disease entity, and / or examination entity, and / or treatment entity; The claims risk control knowledge graph module is used to import the target claims risk control knowledge graph ontology, the ontology association relationship and the sub-diagnosis identification result into the preset first graph database to obtain the target claims risk control knowledge graph; The claims risk control model module is used to construct a target knowledge graph claims risk control model based on the target claims risk control knowledge graph and decision tree model. The third data acquisition module is used to acquire information data of gray-list users identified by the false disclosure risk control model as data to be judged. The false disclosure risk control model is an artificial intelligence claims risk control model built on XGBoost. The information data of gray-list users cannot be identified by the false disclosure risk control model as black-list user information data or white-list user information data. The claims risk assessment module is used to input the information data of the gray list users into the target knowledge graph claims risk control model and output the target claims risk assessment result corresponding to the gray list users; The claims risk assessment module further includes: a gray-list user data submodule, used to obtain the corresponding historical case data of the gray-list user based on the user's unique code; a first graph relationship submodule, used to determine the first graph relationship between the current claim content of the gray-list user and the historical case data of the gray-list user based on the target claims risk knowledge graph; and a claims risk calculation submodule, used by the decision tree model to calculate the target claims risk assessment result of the gray-list user based on the first graph relationship and the target claims risk control knowledge graph. The first graph relationship submodule further includes: a first judgment result subunit, used to obtain a first judgment result regarding whether the current incident content corresponding to the gray list user information data is the same as or contains the disease items and the sub-diagnosis identification results in the corresponding historical case data; a second judgment result subunit, used to obtain a second judgment result regarding whether the current incident content corresponding to the gray list user information data is contained in the disease items and the sub-diagnosis identification results in the corresponding historical case data; and a third judgment result subunit, used to obtain whether the current incident content corresponding to the gray list user information data is contained in the disease items and the sub-diagnosis identification results in the corresponding historical case data. The system includes: a third judgment result for determining sibling relationships; a fourth judgment result subunit for determining whether there is a correlation between the current incident content corresponding to the gray list user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data; a fifth judgment result subunit for determining whether there is a similarity relationship between the current incident content corresponding to the gray list user information data and the disease items and the secondary diagnosis identification results in the corresponding historical case data; and a first graph relationship generation subunit for adding the first judgment result, the second judgment result, the third judgment result, the fourth judgment result, and the fifth judgment result to the first graph relationship.
6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the knowledge graph-based risk control method as described in any one of claims 1 to 4.
7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the knowledge graph-based risk control method as described in any one of claims 1 to 4.