Intelligent claim settlement method and device, electronic equipment, storage medium and program product

Abstract

The invention provides an intelligent claim settlement method and device, electronic equipment, a storage medium and a program product. The method comprises the steps of receiving claim settlement request information of a to-be-claimed case; determining a compliance verification result of the claim settlement request information based on a preset rule engine; inputting the claim settlement request information into a pre-trained risk assessment model to obtain an assessment result in response to the fact that the compliance verification result is passed; according to the evaluation result, determining a risk score of the case to be subjected to claim settlement; and according to the risk score, allocating a processing channel of the to-be-claimed case, and determining a processing result of the to-be-claimed case. The claim settlement efficiency can be remarkably improved, the claim settlement risk is reduced, and the user experience is improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an intelligent claims processing method, device, electronic device, storage medium, and program product. Background Technology

[0002] As the health insurance industry accelerates its digital transformation, the traditional manual claims review model is gradually revealing its efficiency bottlenecks. The rapid increase in the number of claims, the increasing complexity of risk scenarios, and consumers' demands for efficient services are driving the industry towards intelligent transformation.

[0003] However, the current claims process still faces many challenges: on the one hand, although some processes have been automated, there are significant shortcomings in terms of rule flexibility, accuracy of risk assessment, and integration of multi-source data; on the other hand, the rigid design of traditional rule engines is difficult to adapt to dynamic market changes and the emergence of new fraud methods, resulting in increased misjudgment rates and frequent risk control loopholes. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide an intelligent claims processing method, device, electronic device, storage medium, and program product.

[0005] As one aspect of this application, an intelligent claims settlement method is provided, comprising:

[0006] Receive claim request information for pending claims;

[0007] Based on a preset rule engine, the compliance verification result of the claim request information is determined;

[0008] In response to the compliance verification result being passed, the claim request information is input into the pre-trained risk assessment model to obtain the assessment result;

[0009] Based on the assessment results, a risk score is determined for the pending claims.

[0010] Based on the risk score, the processing channel for the pending claims is assigned, and the processing result for the pending claims is determined.

[0011] Optionally, allocating processing channels for pending claims based on the risk score includes:

[0012] Based on the risk score, the risk label of the pending claim case is determined; the risk label includes Level 1, Level 2, Level 3, and Level 4.

[0013] In response to the risk label being Level 1, an automatic claims processing operation is performed on the pending claims case;

[0014] In response to the risk label being Level 2, an automatic claims processing operation is performed on the pending claims, and the claims result is sent to the first user for review;

[0015] In response to the risk label being Level 3, the pending claim is sent to the first user so that the first user can review and process the claim.

[0016] In response to the risk label being Level 4, the pending claim is sent to the second user so that the second user can review and process the claim; wherein, the second user has higher privileges than the first user.

[0017] Optionally, determining the risk label of the claim case based on the risk score includes:

[0018] In response to the risk score falling within a first threshold range, the risk label of the claim case is Level 1;

[0019] In response to the risk score falling within a second threshold range, the risk label of the pending claim case is Level 2; any value within the second threshold range is greater than any value within the first threshold range;

[0020] In response to the risk score falling within the third threshold range, the risk label for the pending claim is Level 3; any value within the third threshold range is greater than any value within the second threshold range;

[0021] In response to the risk score falling within the fourth threshold range, the risk label for the pending claim is Level 4; any value within the fourth threshold range is greater than any value within the third threshold range.

[0022] Optionally, the pre-trained risk assessment model is trained using the following method:

[0023] Obtain historical claims data; the historical claims data includes data from financial, credit, and compliance dimensions;

[0024] Determine the weighting coefficients for the financial dimension, the credit dimension, and the compliance dimension;

[0025] The risk assessment model is trained based on the historical claims data and the weighting coefficients to obtain the pre-trained risk assessment model.

[0026] Optionally, after determining the processing result of the claim case, the method further includes:

[0027] Based on the processing results, calculate the prediction accuracy of the risk assessment model;

[0028] In response to the prediction accuracy being less than a preset accuracy threshold, the claim request with an error in the processing result is determined to be a negative sample, and the risk assessment model is optimized.

[0029] Optionally, receiving the claim request information for pending claims includes:

[0030] Receive initial claim request;

[0031] Based on the initial claim request, key fields are extracted and data verification is performed to obtain the claim request information.

[0032] As a second aspect of this application, an intelligent claims processing device is provided, comprising: a receiving module, a determining module, and a processing module;

[0033] The receiving module is used to receive claim request information for cases pending claims.

[0034] The determining module is used to determine the compliance verification result of the claim request information based on a preset rule engine.

[0035] The processing module is used to, in response to the compliance verification result being passed, input the claim request information into a pre-trained risk assessment model to obtain an assessment result;

[0036] The determining module is further configured to determine the risk score of the claim case based on the assessment results;

[0037] The determining module is also used to allocate processing channels for the pending claims cases based on the risk score, and determine the processing results of the pending claims cases.

[0038] Optionally, the determining module is specifically used to determine the risk label of the claim case based on the risk score; the risk label includes a first level, a second level, a third level, and a fourth level;

[0039] In response to the risk label being Level 1, an automatic claims processing operation is performed on the pending claims case;

[0040] In response to the risk label being Level 2, an automatic claims processing operation is performed on the pending claims, and the claims result is sent to the first user for review;

[0041] In response to the risk label being Level 3, the pending claim is sent to the first user so that the first user can review and process the claim.

[0042] In response to the risk label being Level 4, the pending claim is sent to the second user so that the second user can review and process the claim; wherein, the second user has higher privileges than the first user.

[0043] Optionally, the determining module is further configured to, in response to the risk score being within a first threshold range, assign the risk label of the claim case to a first level;

[0044] In response to the risk score falling within a second threshold range, the risk label of the pending claim case is Level 2; any value within the second threshold range is greater than any value within the first threshold range;

[0045] In response to the risk score falling within the third threshold range, the risk label for the pending claim is Level 3; any value within the third threshold range is greater than any value within the second threshold range;

[0046] In response to the risk score falling within the fourth threshold range, the risk label for the pending claim is Level 4; any value within the fourth threshold range is greater than any value within the third threshold range.

[0047] Optionally, the pre-trained risk assessment model is trained using the following method:

[0048] Obtain historical claims data; the historical claims data includes data from financial, credit, and compliance dimensions;

[0049] Determine the weighting coefficients for the financial dimension, the credit dimension, and the compliance dimension;

[0050] The risk assessment model is trained based on the historical claims data and the weighting coefficients to obtain the pre-trained risk assessment model.

[0051] Optionally, after determining the processing result of the claim case, the intelligent claims device further includes a calculation module;

[0052] The calculation module is used to calculate the prediction accuracy of the risk assessment model based on the processing results.

[0053] The processing module is further configured to, in response to the prediction accuracy being less than a preset accuracy threshold, determine that the claim request with an error in the processing result is a negative sample, and optimize the risk assessment model.

[0054] Optionally, the receiving module is specifically used to receive an initial claim request;

[0055] Based on the initial claim request, key fields are extracted and data verification is performed to obtain the claim request information.

[0056] As a third aspect of this application, an electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, implements the intelligent claims method as described above.

[0057] As a fourth aspect of this application, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the above-described intelligent claims settlement method provided in this application.

[0058] As a fifth aspect of this application, a computer program product is provided, including computer program instructions that, when executed on a computer, cause the computer to perform the intelligent claims settlement method described above.

[0059] As can be seen from the above, the intelligent claims processing method, device, electronic equipment, storage medium, and program products provided in this application, by receiving claims request information and performing compliance verification according to a preset rule engine, eliminate the need for manual review of initial compliance, significantly shortening the initial processing time of claims requests, improving overall claims processing efficiency, and enabling claims to proceed to subsequent processing stages more quickly. Then, a pre-trained risk assessment model is used to evaluate the claims request information that has passed compliance verification, and a risk score is determined based on the results. Compared to traditional manual judgment or simple rule-based risk assessment, this method can more comprehensively and accurately consider various factors to assess the risk level of claims. Simultaneously, it can dynamically adjust the assessment standards and methods for claims risk, reducing misjudgments and risk control loopholes caused by rigid rules. Furthermore, processing channels for pending claims are allocated based on risk scores, achieving precise triage of claims. High-risk cases can be assigned to a professional manual review team for in-depth verification, while low-risk cases can be quickly approved or processed using simplified procedures. This resource allocation method avoids the tediousness and uncertainty of manual case-by-case judgment and processing, ensuring that limited claims processing resources can be used rationally according to the degree of risk of the case, further improving the efficiency of the claims process and speeding up the claims cycle. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in this application or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 A schematic diagram of a risk control and review system based on static rules provided in an embodiment of this application;

[0062] Figure 2 A flowchart illustrating an intelligent claims settlement method provided in this application embodiment;

[0063] Figure 3 A flowchart illustrating another intelligent claims settlement method provided in this application embodiment;

[0064] Figure 4 A flowchart illustrating yet another intelligent claims settlement method provided in this application embodiment;

[0065] Figure 5 A flowchart illustrating yet another intelligent claims settlement method provided in this application embodiment;

[0066] Figure 6 This is a schematic diagram illustrating the composition of an intelligent claims processing device provided in an embodiment of this application;

[0067] Figure 7 This is a schematic diagram of the composition of an electronic device provided in an embodiment of this application. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and accompanying drawings. Obviously, the described embodiments are merely some, not all, of the embodiments described in this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0069] Before using the technical solutions disclosed in the embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained. For example, in response to receiving a user's active request, a prompt message should be sent to the user to clearly inform the user that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware such as electronic devices, applications, servers, or storage media that perform the operations of the technical solutions of this application, based on the prompt message.

[0070] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device. It is understood that the above notification and user authorization process is merely illustrative and does not limit the implementation of this application; other methods that comply with relevant laws and regulations may also be applied to the implementation of this application. It is understood that the user personal information data involved in this technical solution (including but not limited to the data itself, its acquisition, storage, and use) shall comply with the requirements of relevant laws and regulations and shall not violate public order and good morals.

[0071] It should be noted that in the embodiments of this application, the words "exemplarily" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplarily" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of words such as "exemplarily" or "for example" is intended to present the relevant concepts in a specific manner. Unless otherwise defined, the technical or scientific terms used in the embodiments of this application should have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. Invention Overview

[0073] In related technologies, such as Figure 1 As shown, a risk control review system based on static rules is disclosed. The core architecture of the system includes three main modules: a data acquisition module is responsible for extracting key information such as customer insurance application records, claims records, and blacklists from the insurance company's internal database, providing a data foundation for subsequent risk assessment; a rule matching module uses a set of predefined fixed rules (e.g., "medical expenses exceeding the deductible will be subject to manual review") to initially screen submitted claims, aiming to identify and eliminate obviously non-compliant applications; and a manual review module receives cases marked as high-risk by the rule matching module and submits them to human reviewers for detailed processing, while cases not marked (i.e., initially determined to be low-risk) are automatically approved by the system. The typical process is as follows: after a claim application is submitted, the system first retrieves relevant customer historical data; then, it performs initial screening based on preset static rules (such as liability scope limits and maximum compensation amounts); finally, cases that meet all rules enter the automatic calculation stage, while cases that do not meet the rules are transferred to the manual review queue.

[0074] The inventors of this application have discovered that existing technical solutions have significant limitations. First, their core rules are static and fixed, unable to dynamically adjust or learn based on changes in the business environment (such as the emergence of new risk patterns, adjustments to claims strategies, market dynamics, etc.), leading to an outdated rule base and affecting the accuracy and adaptability of risk identification. Second, the system lacks a refined risk threshold management mechanism, simply categorizing cases into "passed" or "manually reviewed." This binary approach results in a large number of cases with actually low risk levels, which could be processed quickly and automatically, being forcibly transferred to the manual review process simply because a fixed rule has been triggered. This significantly increases the workload and operating costs of manual review, reducing overall processing efficiency.

[0075] To address the aforementioned issues, this application provides an intelligent claims processing method. By receiving claims request information and performing compliance verification according to a pre-set rule engine, it eliminates the need for manual review of initial compliance, significantly shortening the initial processing time and improving overall claims efficiency, allowing claims to proceed more quickly to subsequent processing stages. Then, a pre-trained risk assessment model evaluates the compliance-verified claims information to obtain a risk score. Compared to traditional manual judgment or simple rule-based risk assessment, this method comprehensively and accurately considers various factors to assess the risk level of claims. Simultaneously, it dynamically adjusts the assessment standards and methods for claims risk, reducing misjudgments and risk control loopholes caused by rigid rules. Furthermore, processing channels for pending claims are allocated based on risk scores, achieving precise triage of claims. High-risk cases can be assigned to a professional manual review team for in-depth verification, while low-risk cases can be quickly approved or processed using a simplified process. This resource allocation method avoids the tediousness and uncertainty of manual case-by-case judgment and processing, ensuring that limited claims processing resources can be used rationally according to the degree of risk of the case, further improving the efficiency of the claims process and speeding up the claims cycle.

[0076] After introducing the basic principles of this application, the various non-limiting embodiments of this application will be described in detail below.

[0077] Figure 2 This is a flowchart illustrating an intelligent claims processing method provided in an embodiment of this application. Figure 2 As shown, the intelligent claims processing method provided in this application specifically includes the following steps:

[0078] S201. Receive claim request information for pending claims.

[0079] In some embodiments, claim request information refers to a series of relevant documents and data submitted by an insured customer to the insurance company when filing a claim after an insured event has occurred. This information includes the customer's basic personal information (such as name, ID number, contact information, etc.), the detailed contents of the insurance contract (such as the type of insurance, the sum insured, the insurance period, etc.), a detailed description of the time, place, cause, and course of the accident, and various supporting documents related to the accident (such as medical diagnoses, hospitalization records, expense invoices, accident scene photos, traffic police liability determination letters, etc.).

[0080] In some embodiments, such as Figure 3 As shown, S201 can be specifically implemented as follows: S2011-S2012:

[0081] S2011, Receive initial claim request.

[0082] In some embodiments, after a claim application is submitted online or offline, the system receives the user's original claim materials, i.e., the initial claim request. This can be received through multiple channels, such as the insurance company's own application, third-party partner platforms, and offline branches.

[0083] S2012. Based on the initial claim request, extract key fields and perform data validation to obtain the claim request information.

[0084] In some embodiments, upon receiving an initial claim request, the claims system initiates a preprocessing procedure. First, it automatically extracts key fields from the submitted materials using large-scale model intelligence technology, such as the insurance contract number, accident time, location, accident type, and loss amount. Second, it performs basic data validation to ensure completeness, accuracy, and reasonableness, including format validation (e.g., whether the date format is correct), consistency validation (e.g., whether the accident time is within the insurance contract's validity period), logical validation (e.g., whether the loss amount is obviously unreasonable), and basic compliance reviews such as material completeness checks. Minor formatting errors (e.g., extra zeros after the decimal point in the amount) are also standardized. Only the validated key field data can be combined to form complete claim request information that can be used in subsequent claims processes.

[0085] For example, Optical Character Recognition (OCR) can be used to identify key information such as amount, date, and hospital stamp in medical invoices, automatically comparing them against anti-tampering measures (e.g., detecting invoice watermarks) and dates; audio recordings of police reports can be transcribed into text to extract information such as "accident location" and "contact person"; Natural Language Processing (NLP) technology or pre-trained models (such as BERT-NER) can identify key entities in the text. For instance, fields such as "disease name (acute myocardial infarction)" and "surgery type (coronary artery stent implantation)" can be extracted from medical diagnostic reports.

[0086] It should be understood that for information indicating that data validation fails, an error code will be returned and a revision suggestion will be provided.

[0087] S202. Based on the preset rule engine, determine the compliance verification result of the claim request information.

[0088] In some embodiments, the preset rule engine optimizes the traditional static rule system, organically combining human experience with machine learning to construct a continuously evolving risk control system for determining the compliance verification results of claims request information. In terms of technical architecture, a layered design concept is adopted. The top layer is the interactive interface for claims personnel, employing an intuitive visual operation method that allows non-technical personnel to construct complex compliance verification rules through drag-and-drop. The middle layer is the rule execution engine, using a distributed computing framework to ensure high-performance processing capabilities, enabling rapid matching and judgment of compliance verification rules for large amounts of claims request information. The bottom layer is the data analysis and machine learning layer, continuously monitoring the rule execution effect and providing optimization suggestions, allowing the compliance verification rules to be continuously adjusted and optimized based on actual operating conditions.

[0089] The compliance verification rules include compliance rules, normative rules, and reasonableness rules. Compliance rules are the first line of defense in risk management, mainly including material compliance, timeliness compliance, and entity compliance. Normative rules focus on the standardization of business materials, mainly including material specifications, process specifications, and data specifications, reducing human error. Reasonableness rules require a deep integration of professional knowledge and artificial intelligence (AI) technology, mainly including medical reasonableness, accident reasonableness, and behavioral reasonableness. This in-depth analysis can improve the accuracy of fraud detection.

[0090] In some embodiments, once a claim case enters the system, the preset rule engine executes the relevant compliance verification rules in parallel, recording the triggering conditions and execution results of each rule in detail. This execution data constitutes the training material for the machine learning model, which automatically analyzes the actual effect of each rule. For example, if the false positive rate of a certain medical invoice verification rule continues to rise, the engine will automatically reduce its weight or suggest modifying the judgment conditions, thereby improving the accuracy and reliability of compliance verification. The rule optimization process demonstrates the adaptive capability of the preset rule engine. By integrating multiple machine learning algorithms, it can identify potential changes in fraud patterns. For example, when it is discovered that fraudsters are starting to use new methods such as dispersed medical visits and split invoices, the preset rule engine will automatically generate new combined rule suggestions, which can be put into use after confirmation by claims personnel for compliance verification of claim request information. This closed-loop mechanism of "discovery-learning-adaptation" ensures that risk control management remains sensitive to new fraud methods during the compliance verification process, ensuring that only compliant claim requests can pass the compliance verification.

[0091] It's important to note that all rule modifications within the default rule engine are tested in a sandbox environment. This simulates the processing effects of compliance checks on historical cases, preventing any adverse impact on existing business operations. Access control ensures that personnel at different levels can only adjust compliance rules within their authorized scope. A comprehensive version control mechanism maintains a record of every rule change, allowing for rapid rollback when necessary, ensuring the controllability and traceability of the rule adjustment process.

[0092] In practical applications, the pre-defined rule engine has proven highly effective in the compliance verification process. In implementation cases, after its application, the pre-defined rule engine accurately identified various new fraud methods, and the rule base was automatically updated to adapt to new compliance verification requirements. The average processing time for compliant claims cases was reduced by 40%, improving operational efficiency while ensuring risk control, thus achieving a dual promotion of risk management and business development.

[0093] It should be understood that by analyzing historical case data in real time, new risk patterns are automatically identified and corresponding defense rules are generated. For example, verification rules are automatically created to address the frequent occurrence of medical invoice fraud. The optimization process employs reinforcement learning algorithms to evaluate the effectiveness of the rules and adjust parameters, ensuring that the risk control strategy maintains an accuracy rate of over 90%.

[0094] In some embodiments, the preset rule engine can be replaced by a deep learning-based dynamic rule generation system. This system utilizes neural network models in deep learning to automatically learn from a large amount of historical claims data, uncovering key features and risk patterns. This historical data includes various types of case information, such as basic customer information, accident details, claims application materials, and past claims records. Through its multi-layered neuron structure, the neural network can capture complex nonlinear relationships and hidden patterns in the data, thereby more accurately identifying risk factors. Based on the learned features and patterns, the system can dynamically generate rules adapted to the current risk environment. This approach is better suited for handling unstructured data, such as text descriptions and image data in claims applications, while traditional machine learning methods (such as decision trees) may be limited in processing such complex data.

[0095] S203. In response to the compliance verification result being passed, the claim request information is input into the pre-trained risk assessment model to obtain the assessment result.

[0096] In some embodiments, after compliance verification, the claim request information is input into a pre-trained risk assessment model. This pre-trained model utilizes extensive historical claims data and relevant risk characteristics, trained in advance using machine learning or deep learning algorithms. It automatically analyzes and assesses the potential risk level of the case based on the input claim request information. This model integrates numerous factors influencing claim risk, such as the customer's historical claims records, suspicious characteristics of the accident (e.g., multiple accidents involving the same vehicle within a short period), and abnormal fluctuations in medical expenses. Through complex algorithmic calculations, it derives a quantitative assessment result representing the likelihood of risk in the claim case. The assessment result includes multi-dimensional risk indicators output by the risk assessment model, such as fraud probability and case complexity coefficient.

[0097] In some embodiments, such as Figure 4 As shown, the pre-trained risk assessment model was trained using the following method:

[0098] S301. Obtain historical claims data; historical claims data includes data from financial, credit, and compliance dimensions.

[0099] In some embodiments, at the risk assessment level, a comprehensive risk factor evaluation system is constructed from historical claims data. Key factors are extracted from three dimensions: first, the financial dimension, including quantitative data such as the ratio of claim amount to insured amount and historical claim frequency; second, the credit dimension, integrating behavioral data such as customer credit scores and insurance history records; and third, the compliance dimension, through analyzing professional indicators such as the authenticity of medical invoices and the rationality of treatment plans.

[0100] S302. Determine the weighting coefficients for the financial dimension, credit dimension, and compliance dimension.

[0101] In some embodiments, the weighting coefficients are determined based on the importance of each dimension to claims risk, combining the experience of insurance business experts with data analysis results. For example, after analysis and evaluation, the financial dimension has a greater impact on claims risk because it is directly related to the insurance company's economic expenditures and cost control, so it is given a high weight, such as 50%. The credit dimension is next, because a customer's credit status can affect the authenticity of their claims behavior to some extent, so it is weighted at 30%. The compliance dimension has a weight of 20%, although compliance is also important, but its impact is relatively smaller than the first two dimensions in this case.

[0102] S303. Train the risk assessment model based on historical claims data and weighting coefficients to obtain a pre-trained risk assessment model.

[0103] In some embodiments, the risk assessment model is trained using acquired historical claims data and predetermined weighting coefficients, employing appropriate machine learning algorithms (such as logistic regression, decision trees, random forests, neural networks, etc.). The risk assessment model may include a gradient boosting tree (XGBoost) as the base model and a Transformer deep learning model. The model learns the relationship patterns between information in each dimension of historical data and claims risk, adjusting the influence of each dimension's data on the model according to the weighting coefficients, thereby constructing a model structure and parameters capable of accurately assessing claims risk.

[0104] For example, taking the logistic regression algorithm as an example, during the training process, the risk assessment model will take the financial, credit, and compliance data of each case in the historical claims data as input features, and the corresponding claims risk results (such as whether it belongs to a high-risk case, whether there is a risk of fraud, etc.) as target variables. By continuously adjusting the parameters of the risk assessment model through optimization algorithms, the error between the prediction results of the risk assessment model and the actual results is minimized, and finally a pre-trained risk assessment model is obtained. This risk assessment model can effectively assess the risk level of new claims cases based on the input data of each dimension.

[0105] S204. Based on the assessment results, determine the risk score of the cases pending claims.

[0106] In some embodiments, the risk factors in S301 above are weighted and calculated using a risk assessment model, and the final assessment result is mapped to a score range of 0-100, with a higher score indicating a higher risk. Determining the risk score can help claims personnel understand the risk status of a case more intuitively, thereby providing a basis for subsequent handling decisions.

[0107] S205. Based on the risk score, allocate processing channels for pending claims and determine the processing results for pending claims.

[0108] In some embodiments, risk labels for pending claims are determined based on risk scores; risk labels include Level 1, Level 2, Level 3, and Level 4.

[0109] It should be noted that, in response to a risk score falling within the first threshold range, the risk label for a claim is Level 1; in response to a risk score falling within the second threshold range, the risk label for a claim is Level 2, provided that any value in the second threshold range is greater than any value in the first threshold range; in response to a risk score falling within the third threshold range, the risk label for a claim is Level 3, provided that any value in the third threshold range is greater than any value in the second threshold range; in response to a risk score falling within the fourth threshold range, the risk label for a claim is Level 4, provided that any value in the fourth threshold range is greater than any value in the third threshold range. For example, 0-20 points is Level 1 (no risk), 21-50 points is Level 2 (low risk), 51-80 points is Level 3 (medium risk), and 81-100 points is Level 4 (high risk).

[0110] In some embodiments, in response to a risk label of Level 1, an automatic claims processing operation is performed on the pending claims; in response to a risk label of Level 2, an automatic claims processing operation is performed on the pending claims, and the claims result is sent to a first user for review; in response to a risk label of Level 3, the pending claims are sent to the first user so that the first user can review and process the claims; in response to a risk label of Level 4, the pending claims are sent to a second user so that the second user can review and process the claims; wherein, the second user has higher permissions than the first user.

[0111] For example, Level 1 is a fully automated channel with no risk. The system automatically verifies the authenticity of invoices, determines liability, and calculates compensation without any manual intervention. Level 2 is an automatic notification channel with minor risks, but no impact on claims processing. The system automatically completes the claims and sends the results to claims adjusters for reference. Levels 1 and 2 account for approximately 70% of claims. Level 3 involves some risk, accounting for approximately 25% of claims. The system generates a risk assessment report, highlighting 3-5 key points of suspicion, providing similar case references, and suggesting necessary supplementary materials. Claims adjusters then make quick decisions based on this intelligent assistance and execute claims. Level 4 involves higher risk, accounting for approximately 5% of claims. The system automatically links to historical suspicious cases, marks potential fraud characteristics, and generates a detailed investigation list. Department heads then organize in-depth investigations, review the cases, and process the claims.

[0112] The intelligent claims processing method provided in this application receives claims request information and performs compliance verification according to a preset rule engine. This eliminates the need for manual review of initial compliance, significantly shortening the initial processing time and improving overall claims efficiency, allowing claims to proceed more quickly to subsequent processing stages. Then, a pre-trained risk assessment model evaluates the compliance-verified claims information to determine a risk score. Compared to traditional manual judgment or simple rule-based risk assessment, this method comprehensively and accurately considers various factors to assess the risk level of claims. It also dynamically adjusts the assessment standards and methods for claims risk, reducing misjudgments and risk control loopholes caused by rigid rules. Furthermore, processing channels for pending claims are allocated based on the risk score, achieving precise triage of claims. High-risk cases can be assigned to a professional manual review team for in-depth verification, while low-risk cases can be quickly approved or processed using a simplified process. This resource allocation method avoids the tediousness and uncertainty of manual case-by-case judgment and processing, ensuring that limited claims processing resources can be used rationally according to the degree of risk of the case, further improving the efficiency of the claims process and speeding up the claims cycle.

[0113] In some embodiments, such as Figure 5 As shown, after S105, the intelligent claims processing method provided in this application embodiment further includes the following S401-S402:

[0114] S401. Based on the processing results, calculate the prediction accuracy of the risk assessment model.

[0115] In some embodiments, user complaint records are synchronized from customer service systems (telephone / online chat), APP feedback portals, and regulatory complaint platforms. A certain percentage or all of the completed claim requests are then manually reviewed by experienced claims personnel. During the manual review process, the number of claim requests with incorrect processing results is recorded. The type of error can also be recorded (e.g., low-risk cases mistakenly classified as high-risk and denied payment, errors in compensation amount calculation, etc.). This data can serve as an important indicator for measuring model accuracy and the quality of claims review. Statistical analysis of the review results allows for a more accurate determination of the number of claim requests with incorrect processing results. The prediction accuracy of the risk assessment model is then obtained by dividing the number of correctly predicted claim cases by the total number of assessed claim cases.

[0116] S402. In response to a prediction accuracy rate that is less than a preset accuracy rate threshold, determine that the claim request with an erroneous processing result is a negative sample, and optimize the risk assessment model.

[0117] In some embodiments, when the prediction accuracy is less than a preset accuracy threshold, the performance of the risk assessment model can be considered unsatisfactory. In this case, the claim requests with erroneous processing results identified in S401 are used as negative samples to optimize the risk assessment model. Optimization methods may include adjusting model parameters, adding new feature variables, and improving the algorithm. For example, if the risk assessment model is found to have a large error in predicting a new type of fraud, more feature information related to this fraud can be collected and added to the risk assessment model as new feature variables. Simultaneously, the model can be retrained, and parameters adjusted to improve its ability to identify this type of fraud. In this way, the risk assessment model can continuously adapt to new risk situations and business changes, improving its prediction accuracy and risk assessment capabilities, thereby better supporting claims review and reducing the risk losses of insurance companies.

[0118] In some embodiments, this application can also construct a customer-case-medical institution relationship graph, with each entity (customer, case, medical institution, etc.) as nodes in the graph, and the relationships between entities (such as the association between a customer and a case, the association between a case and a medical institution, etc.) as edges. In this way, the complex relationship network between various relevant factors in a claim case can be comprehensively displayed and analyzed. For example, a customer may seek treatment at multiple medical institutions and submit multiple claim cases; the relationship graph clearly shows the connections between these cases and medical institutions. Furthermore, graph embedding technology is used to map the nodes and edges in the graph to a low-dimensional vector space, preserving the semantic information of the nodes and edges. By analyzing these embedded vectors, potential fraud networks can be discovered. For example, if there are abnormal association patterns among multiple customers, such as submitting similar claim cases in a concentrated period of time, and these cases all involve the same medical institution, this may indicate the existence of a fraud ring. Graph neural networks can learn the embedded representations of nodes and edges in the graph to uncover these hidden fraud patterns and risk associations, thereby providing a more comprehensive and in-depth perspective for risk assessment.

[0119] Furthermore, this application can also be applied to e-commerce platforms, where intelligent risk control technology can significantly improve the efficiency and security of the return and refund process. By analyzing historical return and refund data and user behavior data from multiple dimensions, a complete review mechanism is constructed. An automated decision engine combining rule engines and machine learning is used to process return requests. For low-risk orders, such as those with good user credit, no abnormalities in the purchased goods, and reasonable return reasons, automatic review and approval of returns and refunds can be completed quickly, reducing manual intervention and improving processing speed. For medium-risk orders, such as those with some return risk but not yet reaching the high-risk standard, supplementary materials are required. For example, users are required to provide clearer photos of the returned goods, unboxing videos, and other evidence to further verify the authenticity of the return. For high-risk orders, such as those with obvious signs of fraud (e.g., frequent returns, discrepancies between purchased and returned goods), the order is transferred to a manual review queue, and a review checklist containing key verification items is automatically generated for detailed review by professional reviewers. Continuous monitoring of new fraud methods, such as the recently emerging "cash on delivery" refund fraud, is also implemented. After identifying this pattern, the platform's intelligent risk control system can promptly update its review rules, increasing the risk weight of such orders. For example, it implements stricter reviews for "cash on delivery" orders, adding steps such as verifying the recipient's identity and checking proof of delivery, thereby effectively preventing fraud risks, protecting the interests of the platform and merchants, maintaining normal transaction order, and improving the fairness and impartiality of the user experience.

[0120] It should be noted that the method in this embodiment can be executed by a single device, such as a computer or server. The method can also be applied in a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method in this embodiment, and the multiple devices will interact with each other to complete the method described.

[0121] It should be noted that the above description describes some embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0122] Based on the same inventive concept, corresponding to any of the above embodiments, this application also provides an intelligent claims processing device.

[0123] refer to Figure 6 The intelligent claims processing device includes: a receiving module 601, a determining module 602, and a processing module 603;

[0124] The receiving module 601 is used to receive claim request information for cases pending claims.

[0125] The determining module 602 is used to determine the compliance verification result of the claim request information based on a preset rule engine;

[0126] The processing module 603 is used to, in response to the compliance verification result being passed, input the claim request information into a pre-trained risk assessment model to obtain an assessment result;

[0127] The determining module 602 is further configured to determine the risk score of the claim case based on the assessment results;

[0128] The determining module 602 is further configured to allocate processing channels for the pending claims cases based on the risk score, and determine the processing results of the pending claims cases.

[0129] In some embodiments, the determining module 602 is specifically used to determine the risk label of the claim case based on the risk score; the risk label includes a first level, a second level, a third level, and a fourth level;

[0130] In response to the risk label being Level 1, an automatic claims processing operation is performed on the pending claims case;

[0131] In response to the risk label being Level 2, an automatic claims processing operation is performed on the pending claims, and the claims result is sent to the first user for review;

[0132] In response to the risk label being Level 3, the pending claim is sent to the first user so that the first user can review and process the claim.

[0133] In response to the risk label being Level 4, the pending claim is sent to the second user so that the second user can review and process the claim; wherein, the second user has higher privileges than the first user.

[0134] In some embodiments, the determining module 602 is further configured to, in response to the risk score being within a first threshold range, assign a risk label of Level 1 to the claim case;

[0135] In response to the risk score falling within a second threshold range, the risk label of the pending claim case is Level 2; any value within the second threshold range is greater than any value within the first threshold range;

[0136] In response to the risk score falling within the third threshold range, the risk label for the pending claim is Level 3; any value within the third threshold range is greater than any value within the second threshold range;

[0137] In response to the risk score falling within the fourth threshold range, the risk label for the pending claim is Level 4; any value within the fourth threshold range is greater than any value within the third threshold range.

[0138] In some embodiments, the pre-trained risk assessment model is trained using the following methods:

[0139] Obtain historical claims data; the historical claims data includes data from financial, credit, and compliance dimensions;

[0140] Determine the weighting coefficients for the financial dimension, the credit dimension, and the compliance dimension;

[0141] The risk assessment model is trained based on the historical claims data and the weighting coefficients to obtain the pre-trained risk assessment model.

[0142] In some embodiments, after determining the processing result of the claim case, the intelligent claims device further includes a calculation module 604;

[0143] The calculation module 604 is used to calculate the prediction accuracy of the risk assessment model based on the processing result.

[0144] The processing module 603 is further configured to, in response to the prediction accuracy being less than a preset accuracy threshold, determine that the claim request with an error in the processing result is a negative sample, and optimize the risk assessment model.

[0145] In some embodiments, the receiving module 601 is specifically used to receive an initial claim request;

[0146] Based on the initial claim request, key fields are extracted and data verification is performed to obtain the claim request information.

[0147] For ease of description, the above devices are described in terms of function, divided into various modules. Of course, in implementing this application, the functions of each module can be implemented in one or more software and / or hardware.

[0148] The apparatus described above is used to implement the corresponding intelligent claims method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0149] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides an electronic 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 intelligent claims method described in any of the above embodiments.

[0150] Figure 7 This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1060. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1060.

[0151] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0152] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0153] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0154] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0155] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0156] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0157] The electronic devices described above are used to implement the corresponding intelligent claims method in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0158] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides a non-transitory computer-readable storage medium that stores computer instructions for causing the computer to execute the intelligent claims processing method as described in any of the above embodiments.

[0159] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0160] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the intelligent claims method as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0161] Based on the same inventive concept, corresponding to the intelligent claims settlement method described in any of the above embodiments, this disclosure also provides a computer program product, which includes a computer program. In some embodiments, the computer program is executable by one or more processors to cause the processors to execute the intelligent claims settlement method. Corresponding to the execution entity for each step in each embodiment of the intelligent claims settlement method, the processor executing the corresponding step may belong to the corresponding execution entity.

[0162] The computer program product of the above embodiments is used to cause the computer and / or the processor to execute the intelligent claims method as described in any of the above embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0163] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this application (including the claims) is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this application as described above, which are not provided in the details for the sake of brevity.

[0164] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this application, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this application, and this also takes into account the fact that the details of the implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this application will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this application, it will be apparent to those skilled in the art that the embodiments of this application can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0165] Although this application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0166] The embodiments of this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this application.

Claims

1. An intelligent claims settlement method, characterized in that, The method includes: Receive claim request information for pending claims; Based on a preset rule engine, the compliance verification result of the claim request information is determined; In response to the compliance verification result being passed, the claim request information is input into the pre-trained risk assessment model to obtain the assessment result; Based on the assessment results, a risk score is determined for the pending claims. Based on the risk score, the processing channel for the pending claims is assigned, and the processing result for the pending claims is determined.

2. The method according to claim 1, characterized in that, The process of allocating processing channels for pending claims based on the risk score includes: Based on the risk score, the risk label of the pending claim case is determined; the risk label includes Level 1, Level 2, Level 3, and Level 4. In response to the risk label being Level 1, an automatic claims processing operation is performed on the pending claims case; In response to the risk label being Level 2, an automatic claims processing operation is performed on the pending claims, and the claims result is sent to the first user for review; In response to the risk label being Level 3, the pending claim is sent to the first user so that the first user can review and process the claim. In response to the risk label being Level 4, the pending claim is sent to the second user so that the second user can review and process the claim; wherein, the second user has higher privileges than the first user.

3. The method according to claim 2, characterized in that, The process of determining the risk label of the pending claim case based on the risk score includes: In response to the risk score falling within a first threshold range, the risk label of the claim case is Level 1; In response to the risk score falling within a second threshold range, the risk label of the pending claim case is Level 2; any value within the second threshold range is greater than any value within the first threshold range; In response to the risk score falling within the third threshold range, the risk label for the pending claim is Level 3; any value within the third threshold range is greater than any value within the second threshold range; In response to the risk score falling within the fourth threshold range, the risk label for the pending claim is Level 4; any value within the fourth threshold range is greater than any value within the third threshold range.

4. The method according to claim 1, characterized in that, The pre-trained risk assessment model was trained using the following method: Obtain historical claims data; the historical claims data includes data from financial, credit, and compliance dimensions; Determine the weighting coefficients for the financial dimension, the credit dimension, and the compliance dimension; The risk assessment model is trained based on the historical claims data and the weighting coefficients to obtain the pre-trained risk assessment model.

5. The method according to claim 1, characterized in that, After determining the outcome of the claim case, the method further includes: Based on the processing results, calculate the prediction accuracy of the risk assessment model; In response to the prediction accuracy being less than a preset accuracy threshold, the claim request with an error in the processing result is determined to be a negative sample, and the risk assessment model is optimized.

6. The method according to claim 1, characterized in that, The receipt of claim request information for pending claims includes: Receive initial claim request; Based on the initial claim request, key fields are extracted and data verification is performed to obtain the claim request information.

7. An intelligent claims processing device, characterized in that, The device includes: a receiving module, a determining module, and a processing module; The receiving module is used to receive claim request information for cases pending claims. The determining module is used to determine the compliance verification result of the claim request information based on a preset rule engine. The processing module is used to, in response to the compliance verification result being passed, input the claim request information into a pre-trained risk assessment model to obtain an assessment result; The determining module is further configured to determine the risk score of the claim case based on the assessment results; The determining module is also used to allocate processing channels for the pending claims cases based on the risk score, and determine the processing results of the pending claims cases.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method as claimed in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method of any one of claims 1 to 6.

10. A computer program product comprising computer program instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1 to 6.

Images