Reimbursement information auditing method and device, reimbursement information training method and device, equipment and storage medium

Abstract

The invention relates to the technical field of reimbursement auditing intellectualization, can be applied to an artificial intelligence business scene, a financial business scene and a medical health business scene, and provides a reimbursement information auditing method, training method, device and equipment and a storage medium, and the method comprises the steps: obtaining preset reimbursement information; obtaining feature information corresponding to the preset reimbursement information based on a feature extraction sub-model of the reimbursement auditing model; determining risk indexes of the preset reimbursement information under a plurality of preset reimbursement dimensions based on the risk assessment sub-model; and based on the auditing sub-model, auditing the preset reimbursement information to obtain an auditing result corresponding to the preset reimbursement information so as to improve the auditing convenience of the reimbursement information. For a financial business scene, the preset reimbursement information comprises invoices, contracts and the like related to business activities, and can be used for financial reimbursement auditing. For the medical health service scene, the preset reimbursement information comprises a doctor-seeing fee payment voucher, an electronic medical record and the like, and can be used for medical reimbursement auditing.

Description

Technical Field

[0001] The present application relates to the field of intelligent reimbursement review technology, and can be applied to artificial intelligence business scenarios, financial business scenarios, and medical and health business scenarios. In particular, it relates to a review method, training method, device, equipment, and storage medium for reimbursement information. Background Art

[0002] Currently, reviewing financial expense reimbursements is a crucial task across all industries. For example, in financial business scenarios, employees may engage in work-related consumer activities such as business and shopping during their workdays. When engaging in these activities, employees can access relevant reimbursement information, such as invoices and contracts, for financial reimbursement. In healthcare scenarios, patients can access reimbursement information during their visits, including payment receipts for medical expenses, medications, and electronic medical records. This reimbursement information can be used for appropriate reimbursement review. Traditional reimbursement review methods often rely on manual labor, resulting in low review efficiency and an inability to effectively identify potential risks, making the review of reimbursement information inconvenient. Summary of the Invention

[0003] The main purpose of this application is to provide a reimbursement information review method, training method, device, equipment and storage medium, aiming to improve the convenience of reviewing reimbursement information.

[0004] In a first aspect, the present application provides a method for reviewing reimbursement information, comprising:

[0005] Obtain preset reimbursement information pending review;

[0006] Based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information;

[0007] Based on the risk assessment sub-model of the reimbursement review model, and according to the characteristic information, determining the risk index of the preset reimbursement information under multiple preset reimbursement dimensions;

[0008] Based on the audit sub-model of the reimbursement audit model, the preset reimbursement information is audited according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain an audit result corresponding to the preset reimbursement information.

[0009] In a second aspect, the present application also provides a training method for a reimbursement review model, comprising:

[0010] Obtaining a training sample set, the training sample set including a plurality of preset reimbursement information and an audit result label corresponding to each of the preset reimbursement information;

[0011] Based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information;

[0012] Based on the risk assessment sub-model of the reimbursement review model, and according to the characteristic information, determining the risk index of the preset reimbursement information under multiple preset reimbursement dimensions;

[0013] Based on the audit sub-model of the reimbursement audit model, the preset reimbursement information is audited according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain an audit result corresponding to the preset reimbursement information;

[0014] Determining model optimization parameters of the reimbursement review model according to the review result and the review result label;

[0015] According to the model optimization parameters, the model parameters of the reimbursement review model are adjusted.

[0016] In a third aspect, the present application further provides a reimbursement information review device, the review device comprising:

[0017] An information acquisition module is used to obtain preset reimbursement information to be reviewed;

[0018] A feature extraction module, configured to perform feature extraction processing on the preset reimbursement information based on a feature extraction sub-model of a reimbursement review model to obtain feature information corresponding to the preset reimbursement information;

[0019] A risk assessment module, configured to determine, based on the risk assessment sub-model of the reimbursement review model and according to the characteristic information, the risk index of the preset reimbursement information under a plurality of preset reimbursement dimensions;

[0020] The audit module is used to audit the preset reimbursement information based on the audit sub-model of the reimbursement audit model and the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain the audit results corresponding to the preset reimbursement information.

[0021] In a fourth aspect, the present application further provides a computer device, comprising a memory and a processor;

[0022] The memory is used to store computer programs;

[0023] The processor is used to execute the computer program and implement the steps of the above-mentioned reimbursement information review method when executing the computer program.

[0024] In a fifth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned reimbursement information review method are implemented.

[0025] The present application provides a reimbursement information review method, training method, device, equipment and storage medium. The reimbursement information review method includes: obtaining preset reimbursement information to be reviewed; based on the feature extraction sub-model of the reimbursement review model, performing feature extraction processing on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information; based on the risk assessment sub-model of the reimbursement review model, determining the risk index of the preset reimbursement information under multiple preset reimbursement dimensions according to the feature information; based on the review sub-model of the reimbursement review model, reviewing the preset reimbursement information according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, and obtaining the review result corresponding to the preset reimbursement information.

[0026] Based on the setting of the reimbursement review model, the computer device can determine the risk index of the preset reimbursement information under multiple preset reimbursement dimensions through the reimbursement review model, and determine the review result corresponding to the preset reimbursement information by comprehensively considering the risk index of the preset reimbursement information under multiple preset reimbursement dimensions. In the process of reviewing the results corresponding to the preset reimbursement information, there is no need for reviewers to evaluate the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, and to determine the review results corresponding to the preset reimbursement information, which is conducive to saving manpower and thus conducive to improving the convenience of reviewing the reimbursement information. Accordingly, the reimbursement review model can identify the risks existing in the preset reimbursement information based on the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, which is conducive to improving the review effect of the reimbursement information. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0028] Figure 1 This is a schematic diagram of an application environment of a method for reviewing reimbursement information in an embodiment of the present application;

[0029] Figure 2 This is a flow chart of a method for reviewing reimbursement information in one embodiment of the present application;

[0030] Figure 3 This is a flowchart of a training method for a reimbursement review model in one embodiment of the present application;

[0031] Figure 4 This is a reimbursement review flow chart involved in the reimbursement review model in one embodiment of the present application;

[0032] Figure 5 This is a structural diagram of a device for reviewing reimbursement information in one embodiment of the present application;

[0033] Figure 6 is a structural diagram of a computer device in one embodiment of the present application;

[0034] Figure 7 It is another structural diagram of a computer device in one embodiment of the present application. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0036] The method for reviewing reimbursement information provided in the embodiment of the present application can be applied to Figure 1In an application environment, the client communicates with the server through a network. The server can obtain the preset reimbursement information to be reviewed through the client, and based on the feature extraction sub-model of the reimbursement review model, perform feature extraction processing on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information; based on the risk assessment sub-model of the reimbursement review model, determine the risk index of the preset reimbursement information under multiple preset reimbursement dimensions according to the feature information; based on the review sub-model of the reimbursement review model, review the preset reimbursement information according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, obtain the review result corresponding to the preset reimbursement information, and then feed back the review result corresponding to the preset reimbursement information to the client. In this application, the reimbursement review model is an artificial intelligence model to perform intelligent review of the preset reimbursement information. The reimbursement review model can be used to review reimbursement information related to financial business scenarios, such as invoices, contracts, reimbursement documents, payment vouchers, etc., obtained by corporate employees when participating in business activities, shopping activities and other work-related consumption activities, which is conducive to improving the convenience of reviewing reimbursement information. The reimbursement review model can also be used to review reimbursement information related to medical and health business scenarios obtained by patients during the treatment process, such as medical fee payment vouchers, drug fee payment vouchers, electronic medical records, etc., so as to facilitate the subsequent medical reimbursement of patients, which is conducive to improving the convenience of reviewing reimbursement information. Among them, the client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented as an independent server or a server cluster consisting of multiple servers. The following is a detailed description of this application through specific embodiments.

[0037] See also Figure 2 As shown, Figure 2 A flowchart of a method for reviewing reimbursement information provided in an embodiment of the present application includes the following steps:

[0038] S101: Obtain preset reimbursement information to be reviewed.

[0039] In the case where the reimbursement information needs to be reviewed, the computer device can obtain the preset reimbursement information to be reviewed.

[0040] For example, the pre-set reimbursement information can be determined based on a pre-set reimbursement request submitted by a reimbursement recipient. For example, in a financial business scenario, the reimbursement recipient may include a company employee. When submitting a pre-set reimbursement request, the company employee may upload pre-set reimbursement documents related to the pre-set reimbursement request, such as invoices, contracts, payment receipts, etc. In a healthcare business scenario, the reimbursement recipient may include a patient. When submitting a pre-set reimbursement request, the patient may upload pre-set reimbursement documents related to the pre-set reimbursement request, such as medical fee payment receipts, drug fee payment receipts, electronic medical records, etc. The computer device may determine the pre-set reimbursement information based on the pre-set reimbursement request and the pre-set reimbursement documents included in the pre-set reimbursement request. The pre-set reimbursement information may include at least one of the pre-set reimbursement request and the pre-set reimbursement documents. The pre-set reimbursement documents may include invoices, contracts, payment receipts, medical fee payment receipts, drug fee payment receipts, electronic medical records, etc. Of course, the pre-set reimbursement information is not limited to these and is not intended to be limiting herein.

[0041] When the preset reimbursement information to be reviewed is obtained, the computer device may input the obtained preset reimbursement information into the reimbursement review model, so that the reimbursement review model reviews the preset reimbursement information. For example, when the reimbursement review model receives the preset reimbursement request and the preset reimbursement document included in the preset reimbursement information, it may subsequently review the preset reimbursement request and the preset reimbursement document included in the preset reimbursement information and obtain a corresponding review result.

[0042] The preset reimbursement information can be reviewed by the reimbursement review model to subsequently determine the review results corresponding to the preset reimbursement information, thereby promoting the intelligent review of the preset reimbursement information, which will be beneficial to subsequently improve the convenience of reviewing the preset reimbursement information.

[0043] S102: Based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information.

[0044] For example, when the reimbursement review model obtains the preset reimbursement information, it can use the feature extraction sub-model to perform feature extraction processing on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information.

[0045] In some embodiments, at least one of the preset reimbursement request and preset reimbursement document included in the preset reimbursement information is input into the feature sub-model of the reimbursement review model to obtain feature information corresponding to the preset reimbursement information. For example, the feature information includes at least one of the reimbursement item, reimbursement amount, reimbursement type, reimbursement time, reimbursement requester, and historical reimbursement information corresponding to the reimbursement requester. Of course, the feature information is not limited to this and is not intended to be limiting herein.

[0046] For example, the preset reimbursement information can be in the form of text, pictures, etc. Taking the example of the preset reimbursement information including invoices, the invoice can be a picture obtained by taking a picture of a physical invoice, or it can be an electronic invoice. The invoice can include text. Taking the example of the preset reimbursement information including electronic medical records, the electronic medical records can include medical images taken during the diagnosis process, the doctor's diagnosis text, etc., which can be used as the basis for medical reimbursement. In the process of performing feature extraction processing on the preset reimbursement information, the feature extraction sub-model can use a natural language processing algorithm to perform text analysis processing on the preset reimbursement information to determine the corresponding feature information. Of course, this is not limited to this. The feature extraction sub-model can also use an image recognition algorithm to perform image recognition processing on the preset reimbursement information to determine the corresponding feature information. This is not limited here.

[0047] For example, when the person making the reimbursement request submits a preset reimbursement request, they can specify text indicating information such as the reimbursement items and the reimbursement amount. The feature extraction sub-model can perform text analysis on the text in the preset reimbursement request to extract feature information corresponding to the preset reimbursement request, such as the reimbursement items and reimbursement amount corresponding to the preset reimbursement request. The feature extraction sub-model can also perform text analysis on the reimbursement items to determine the reimbursement type to which the reimbursement items belong. The feature extraction sub-model can also determine the reimbursement person based on the person who submitted the preset reimbursement request. The feature extraction sub-model can also obtain the historical reimbursement information corresponding to the determined reimbursement person based on the determined reimbursement person.

[0048] For another example, when the person making the reimbursement provides a preset reimbursement document, the preset reimbursement document may be scanned, photographed, or otherwise processed to provide the preset reimbursement document in the form of a picture. The feature extraction sub-model may perform image recognition processing on the preset reimbursement document in the form of a picture to extract feature information corresponding to the preset reimbursement document, such as the reimbursement item, reimbursement amount, and reimbursement person corresponding to the preset reimbursement document. Accordingly, the feature extraction sub-model may also perform text analysis processing on the reimbursement reason to determine the reimbursement type to which the reimbursement reason belongs. The feature extraction sub-model may also obtain historical reimbursement information corresponding to the determined reimbursement person based on the determined reimbursement person. And so on.

[0049] When the characteristic information corresponding to the preset reimbursement information is determined, the characteristic information corresponding to the preset reimbursement information can be used to subsequently determine the audit result corresponding to the preset reimbursement information, which is conducive to subsequently improving the convenience of auditing the preset reimbursement information.

[0050] S103: Based on the risk assessment sub-model of the reimbursement review model, determine the risk index of the preset reimbursement information under multiple preset reimbursement dimensions according to the characteristic information.

[0051] For example, after determining the characteristic information corresponding to the pre-set reimbursement information, the reimbursement review model can use the risk assessment sub-model to perform a multi-dimensional risk assessment on the characteristic information to determine the risk index of the pre-set reimbursement information across multiple pre-set reimbursement dimensions. The risk index of the pre-set reimbursement information across the pre-set reimbursement dimensions can be obtained by quantitatively assessing whether the pre-set reimbursement information has risks or potential risks within the pre-set reimbursement dimensions.

[0052] The number of characteristic information corresponding to the preset reimbursement information may include multiple. The risk assessment sub-model can determine the risk index of the same preset reimbursement information under different preset reimbursement dimensions based on the same characteristic information. The risk assessment sub-model can also determine the risk index of the same preset reimbursement information under different preset reimbursement dimensions based on different characteristic information. The risk assessment sub-model can also determine the risk index of the same preset reimbursement information under different preset reimbursement dimensions based on partially the same characteristic information, which is not limited here.

[0053] For example, the preset reimbursement dimensions may include at least two of the following: a reimbursement information risk assessment dimension, a reimbursement person risk assessment dimension, an organizational risk assessment dimension, and a historical reimbursement trend assessment dimension. Of course, the preset reimbursement dimensions are not limited thereto and are not intended to be limiting herein.

[0054] The reimbursement information risk assessment dimension can be used to assess the risks or potential risks associated with at least one of the reimbursement amount, reimbursement item, and reimbursement type included in the pre-set reimbursement information. For example, in the case of pre-set reimbursement information including the reimbursement amount and reimbursement item determined based on an invoice, the risk assessment sub-model can assess whether the reimbursement amount indicated on the invoice falls within the budget corresponding to the reimbursement item indicated on the invoice to determine the risk index of the pre-set reimbursement information under the reimbursement information risk assessment dimension. This is, of course, not a limitation and is not intended to be construed herein.

[0055] The reimbursement person risk assessment dimension can be used to assess the risks or potential risks of at least one of the reimbursement person and the historical reimbursement information corresponding to the reimbursement person included in the preset reimbursement information. For example, the risk assessment sub-model can determine at least one of the reimbursement person's reimbursement frequency and the reimbursement person's historical reimbursement behavior based on the historical reimbursement information corresponding to the reimbursement person, and determine the risk index of the preset reimbursement information under the reimbursement person risk assessment dimension based on at least one of the reimbursement person's reimbursement frequency and the reimbursement person's historical reimbursement behavior. Historical reimbursement behavior can be used to indicate whether the reimbursement person has committed illegal reimbursement behavior within a historical time period. Illegal reimbursement behavior may include false reimbursement behavior, repeated reimbursement behavior, etc., which are not limited here.

[0056] The organizational risk assessment dimension can be used to assess the risks or potential risks of the organization to which the reimbursement claimant belongs, as included in the preset reimbursement information. The organization to which the reimbursement claimant belongs can include at least one of the following: the group, department, or enterprise to which the reimbursement claimant belongs, but is not limited thereto and is not intended to be limiting herein. For example, the organization to which the reimbursement claimant belongs can include the reimbursement claimant and other reimbursement claimants other than the reimbursement claimant. The risk assessment sub-model can comprehensively consider the reimbursement behavior of the reimbursement claimant and other reimbursement claimants within a preset time period to determine the risk index of the preset reimbursement information under the organizational risk assessment dimension.

[0057] The historical reimbursement trend assessment dimension can be used to assess the risks or potential risks of the changing trends corresponding to the preset reimbursement information. In the case where the volatility of the changing trends corresponding to the preset reimbursement information is large, the risk index of the preset reimbursement information under the historical reimbursement trend assessment dimension is correspondingly higher. For example, the changing trends corresponding to the preset reimbursement information may include the changing trends corresponding to the reimbursement amounts. The risk assessment sub-model can compare the reimbursement amounts and the historical reimbursement amounts corresponding to the same reimbursement item to determine whether the changing trends corresponding to the reimbursement amounts are consistent with the historical reimbursement amount trends corresponding to the same reimbursement item, and then determine the risk index of the preset reimbursement information under the historical reimbursement trend assessment dimension.

[0058] If the risk assessment sub-model can determine the risk index of the pre-set reimbursement information under multiple pre-set reimbursement dimensions based on the characteristic information, the risk index of the pre-set reimbursement information under multiple pre-set reimbursement dimensions can be used to indicate the risk or potential risk of the pre-set reimbursement information under multiple pre-set reimbursement dimensions, thereby facilitating a more convenient and comprehensive risk assessment of the pre-set reimbursement information. Accordingly, the risk index of the pre-set reimbursement information under multiple pre-set reimbursement dimensions can be used to subsequently determine the audit results corresponding to the pre-set reimbursement information, thereby facilitating a more convenient audit of the pre-set reimbursement information.

[0059] S104: Based on the audit sub-model of the reimbursement audit model, the preset reimbursement information is audited according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain an audit result corresponding to the preset reimbursement information.

[0060] For example, when the reimbursement audit model determines the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, it can use the audit sub-model to combine the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to audit the preset reimbursement information and obtain the corresponding audit results.

[0061] In some embodiments, the audit sub-model can determine the audit sub-results for the pre-set reimbursement information under each pre-set reimbursement dimension based on the risk index under each pre-set reimbursement dimension. Accordingly, the audit sub-model can combine the audit sub-results under each pre-set reimbursement dimension to determine the audit result corresponding to the pre-set reimbursement information.

[0062] In other embodiments, the audit sub-model can jointly determine a risk score corresponding to the preset reimbursement information based on risk indices under multiple preset reimbursement dimensions. Accordingly, the audit sub-model can determine an audit result corresponding to the preset reimbursement information based on the risk score corresponding to the preset reimbursement information.

[0063] Illustratively, the audit result corresponding to the preset reimbursement information may include at least one of passed, pending manual review, and prompt information.

[0064] In the case where the reimbursement review model determines the review result corresponding to the preset reimbursement information, the reimburser can determine whether it is necessary to perform subsequent operations on the preset reimbursement information based on the instructions of the review result. Take the reimburser as an example, including the reimburser and the reviewer. In the case where the review result corresponding to the preset reimbursement information is approved, the reimburser and the reviewer do not need to perform subsequent operations on the preset reimbursement information. In the case where the review result corresponding to the preset reimbursement information is pending manual review, the reviewer needs to review the preset reimbursement information. In the case where the review result corresponding to the preset reimbursement information includes prompt information, if the prompt information is used to indicate the key review dimension corresponding to the preset reimbursement information, the reviewer can focus on the key review dimension corresponding to the preset reimbursement information in accordance with the prompt information. If the prompt information is used to indicate the information supplement dimension corresponding to the preset reimbursement information, the reimburser can supplement the relevant reimbursement information for the information supplement dimension corresponding to the preset reimbursement information in accordance with the prompt information.

[0065] For example, for those who claim reimbursement, such as corporate employees or patients, the reimbursement review model can be used to review the preset reimbursement information that they can currently submit and obtain the corresponding review results. Accordingly, based on the review results, the reimbursement person can determine whether they still need to submit other preset reimbursement information. The reimbursement review model can be used to assist the reimbursement person in advancing the reimbursement review process, thereby improving the reimbursement person's experience of using the reimbursement review model. For auditors, the reimbursement review model can be used to review the preset reimbursement information provided by the reimbursement person and obtain the corresponding review results. Accordingly, based on the review results, the auditor can determine whether they still need to review the preset reimbursement information. The reimbursement review model can be used to reduce the auditor's workload when reviewing the preset reimbursement information, thereby improving the auditor's experience of using the reimbursement review model. The reimbursement review model can be used to promote the intelligence and automation of the reimbursement review process.

[0066] If the reimbursement review model can determine the review result corresponding to the preset reimbursement information by combining the risk index of the preset reimbursement information under multiple preset reimbursement dimensions through the review sub-model, it is no longer necessary for the reviewer to evaluate the risk index of the preset reimbursement information under multiple preset reimbursement dimensions and determine the review result corresponding to the preset reimbursement information, which is conducive to improving the convenience of reviewing reimbursement information. Accordingly, the reimbursement review model can identify risks in the preset reimbursement information based on the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, which is conducive to improving the effectiveness of the review of reimbursement information.

[0067] In some embodiments, multiple initial reimbursement information to be reviewed is obtained; based on the correlation evaluation sub-model of the reimbursement review model, information correlation evaluation is performed on the multiple initial reimbursement information to obtain corresponding preset reimbursement information; the preset reimbursement information includes at least one initial reimbursement information, and there is a correlation relationship between the multiple initial reimbursement information included in the preset reimbursement information.

[0068] For example, the initial reimbursement information can be determined based on at least one of an initial reimbursement request filed by the claimant and an initial reimbursement receipt uploaded by the claimant. The initial reimbursement information can include, without limitation, invoices, contracts, payment receipts, medical fee payment receipts, drug fee payment receipts, and electronic medical records.

[0069] For example, if different reimbursers submit different initial reimbursement requests, or if the same reimburser submits different initial reimbursement requests, there may be correlations between the different initial reimbursement requests. Accordingly, if different reimbursers upload different initial reimbursement invoices, or if the same reimburser uploads different initial reimbursement invoices, there may be correlations between the different initial reimbursement invoices. Based on this, the reimbursement review model can use the correlation assessment sub-model to perform an information correlation assessment on multiple initial reimbursement information, such as multiple initial reimbursement requests and multiple initial reimbursement invoices, to obtain the corresponding preset reimbursement information.

[0070] For example, when the relevance evaluation sub-model determines that at least two of the multiple initial reimbursement requests are associated with each other, the at least two associated initial reimbursement requests may be determined as the same preset reimbursement request, which may be used to determine preset reimbursement information.

[0071] For another example, when the relevance assessment sub-model determines that at least two of the multiple initial reimbursement documents are associated with each other, the at least two associated initial reimbursement documents may be determined as the same preset reimbursement document. The preset reimbursement document may be used to determine the preset reimbursement information.

[0072] For another example, after determining at least one pre-set reimbursement request and at least one pre-set reimbursement document, the relevance assessment sub-model may further perform an information relevance assessment on the pre-set reimbursement request and the pre-set reimbursement document. If a relevance relationship is determined between the pre-set reimbursement request and the pre-set reimbursement document, the relevance assessment sub-model may jointly determine the pre-set reimbursement information based on the relevance relationship between the pre-set reimbursement request and the pre-set reimbursement document.

[0073] Based on this, the preset reimbursement information may include at least one initial reimbursement information. Accordingly, when the preset reimbursement information includes multiple initial reimbursement information, there is an association relationship between the multiple initial reimbursement information.

[0074] When the reimbursement review model determines preset reimbursement information based on one or more initial reimbursement information, the preset reimbursement information can be used to subsequently determine the review results corresponding to the preset reimbursement information, which is conducive to improving the convenience of the subsequent review of the preset reimbursement information. Accordingly, when the preset reimbursement information includes multiple initial reimbursement information, the reimbursement review model can simultaneously review multiple initial reimbursement information that have a correlation, which is conducive to improving the efficiency of the reimbursement review model for reviewing multiple initial reimbursement information. Furthermore, the reimbursement review model can integrate multiple initial reimbursement information to determine the review results corresponding to the preset reimbursement information, which is conducive to improving the accuracy of the reimbursement review model for reviewing the preset reimbursement information.

[0075] Of course, this isn't limited to this. Once the reimbursement information is obtained, the relevance assessment sub-model can also verify its integrity. For example, if the initial reimbursement information includes invoices, the relevance assessment sub-model can connect to the tax platform to verify the authenticity of the invoices, and then determine authentic invoices as the pre-set reimbursement information. Based on this, the reimbursement review model can ensure that subsequent reimbursement review processes are based on authentic and trustworthy pre-set reimbursement information, which helps improve the accuracy of the review of pre-set reimbursement information.

[0076] In some embodiments, semantic reasoning is performed on the initial reimbursement information to determine the semantic information of the initial reimbursement information; when it is determined based on the semantic information of at least two initial reimbursement information that there is at least one of a reimbursement spatiotemporal association relationship, a reimbursement matter association relationship, and a reimbursement amount association relationship between the at least two initial reimbursement information, the at least two initial reimbursement information are determined to be the same preset reimbursement information.

[0077] For example, the relevance assessment sub-model of the reimbursement review model can have large language model capabilities, such as semantic understanding and reasoning capabilities. When the relevance assessment sub-model obtains multiple initial reimbursement information, it can perform semantic reasoning on each initial reimbursement information to obtain the semantic information of each initial reimbursement information.

[0078] Exemplarily, the relevance evaluation sub-model can extract implicit features from the unstructured information included in the initial reimbursement information, and then determine the semantic information of the initial reimbursement information in combination with the implicit features. Accordingly, the semantic information of the initial reimbursement information can be associated with one or more of the corresponding reimbursement time and space, the corresponding reimbursement items, and the corresponding reimbursement amount. When the relevance evaluation sub-model determines that different initial reimbursement information is respectively associated with the same reimbursement time and space based on the semantic information of different initial reimbursement information, it can be determined that there is a reimbursement time and space association relationship between different initial reimbursement information. When the relevance evaluation sub-model determines that different initial reimbursement information is respectively associated with the same reimbursement item based on the semantic information of different initial reimbursement information, it can be determined that there is a reimbursement item association relationship between different initial reimbursement information. When the relevance evaluation sub-model determines that different initial reimbursement information is respectively associated with the same reimbursement amount based on the semantic information of different initial reimbursement information, it can be determined that there is a reimbursement amount association relationship between different initial reimbursement information.

[0079] For example, consider the initial reimbursement information that includes invoices A and B. The relevance assessment sub-model can, for example, extract implicit features of invoice A based on invoice notes, consumption details, and other information that may be written on invoice A. If, for example, the implicit features of invoice A include a restaurant invoice and a meeting location, the relevance assessment sub-model can determine that the reimbursement item associated with invoice A is a meeting item. Accordingly, if the reimbursement item associated with invoice B is also a meeting item, the reimbursement assessment sub-model can determine that a reimbursement item association exists between invoices A and B, thereby grouping invoices A and B into the same pre-set reimbursement information.

[0080] For example, consider the case where the initial reimbursement information includes payment vouchers C and D, both uploaded by the same reimburser. Based on the semantic information of payment voucher C, the relevance assessment sub-model can determine that the time of payment voucher C is time 1 and the location of payment voucher C is location 1. Based on the semantic information of payment voucher D, the relevance assessment sub-model can determine that the time of payment voucher D is time 2 and the location of payment voucher C is location 2. If time 1 falls within the preset time period corresponding to time 2, and location 1 falls within the preset spatial range corresponding to location 2, the relevance assessment sub-model can determine that a spatiotemporal reimbursement relationship exists between payment vouchers C and D, and thus group payment vouchers C and D together as the same preset reimbursement information. For example, the preset time period can include the five days before and after time 2, and the preset spatial range can include the same city as location 2. Of course, this is not limiting; the preset time period and preset spatial range can be pre-set or user-defined, and are not intended to be limiting here.

[0081] Take the initial reimbursement information including invoice E as an example. The reimbursement review model can identify the reimbursement amount indicated by invoice E. When the reimbursement review model determines that the reimbursement amount indicated by invoice E exceeds the preset reimbursement amount threshold, the reimbursement review model can determine that invoice E is a large-amount reimbursement, and then the amount of invoice E can be split. For example, the reimbursement review model can split invoice E into multiple sub-invoices and generate association numbers for each of the multiple sub-invoices. The sum of the reimbursement amounts of the multiple sub-invoices is equal to the reimbursement amount of invoice E. The semantic information of the sub-invoice determined by the association evaluation sub-model can be used to indicate the association number of the sub-invoice. The association evaluation sub-model can then use the association number of the sub-invoice to determine that there is a reimbursement amount association relationship between the multiple sub-invoices corresponding to invoice E, and then group the multiple sub-invoices corresponding to invoice E into the same preset reimbursement information.

[0082] Of course, this is not limited to this. When the correlation evaluation sub-model determines that a certain initial reimbursement information has no correlation with other initial reimbursement information, the initial reimbursement information can be directly determined as the preset reimbursement information, which is not limited here.

[0083] The relevance assessment sub-model allows the reimbursement review model to identify the relationships between different initial reimbursement information, which helps improve the reimbursement review model's convenience in assessing the relevance of initial reimbursement information. If the reimbursement review model can assess the relevance of multiple initial reimbursement information and obtain the corresponding preset reimbursement information, the reimbursement review model can simultaneously review multiple initial reimbursement information with related relationships, which helps improve the efficiency and accuracy of the reimbursement review model's review of multiple initial reimbursement information.

[0084] In some embodiments, the characteristic information is compared with the preset reimbursement conditions corresponding to the preset reimbursement dimension to obtain a preset comparison result of the preset reimbursement information under the preset reimbursement dimension; based on the preset comparison result, the reimbursement risk corresponding to the preset reimbursement information in the preset reimbursement dimension is evaluated to obtain a risk index of the preset reimbursement information under the preset reimbursement dimension.

[0085] For example, the multiple preset reimbursement dimensions may include at least two of a reimbursement information risk assessment dimension, a reimbursement person risk assessment dimension, an organizational risk assessment dimension, and a historical reimbursement trend assessment dimension. Preset reimbursement conditions corresponding to different preset reimbursement dimensions may be different.

[0086] In the case where the preset reimbursement dimension includes a reimbursement information risk assessment dimension, the preset reimbursement conditions corresponding to the reimbursement information risk assessment dimension may include: the reimbursement amount risk assessment condition may include at least one of the following: the reimbursement amount is within the budget range corresponding to the reimbursement item, the reimbursement amount is less than or equal to the average historical reimbursement amount corresponding to the reimbursement item, the reimbursement amount is less than or equal to the preset reimbursement amount threshold, the application fee corresponding to the reimbursement item matches the budget item, and the application fee corresponding to the reimbursement item is within the budget item corresponding to the reimbursement item, so as to evaluate whether there is risk or potential risk in at least one of the reimbursement amount and reimbursement items included in the preset reimbursement information.

[0087] In an exemplary embodiment, if the preset comparison result indicates that the reimbursement amount included in the preset reimbursement information is not within the budget range corresponding to the reimbursement item, the greater the amount by which the reimbursement amount exceeds the budget range, the higher the risk index of the preset reimbursement information under the reimbursement information risk assessment dimension. Accordingly, if the preset comparison result indicates that the application fee corresponding to the reimbursement item included in the preset reimbursement information does not match the budget item, the higher the application fee corresponding to the reimbursement item, the higher the risk index of the preset reimbursement information under the reimbursement information risk assessment dimension. And so on.

[0088] If the preset reimbursement dimensions include a reimbursement recipient risk assessment dimension, the preset reimbursement conditions corresponding to the reimbursement recipient risk assessment dimension may include at least one of the following: the reimbursement recipient's reimbursement frequency is less than or equal to a preset reimbursement frequency threshold, and the reimbursement recipient's historical reimbursement behavior does not contain any illegal reimbursement behavior, to assess whether the reimbursement recipient corresponding to the preset reimbursement information is at risk or potentially at risk. The reimbursement recipient's reimbursement frequency may be determined based on the number of reimbursements made by the reimbursement recipient within a preset time period corresponding to the current time.

[0089] In an exemplary embodiment, if the preset comparison result indicates that the reimbursement frequency of the reimbursement claimant is greater than a preset reimbursement frequency threshold, the higher the reimbursement frequency, the higher the risk index of the preset reimbursement information under the reimbursement claimant risk assessment dimension. If the preset comparison result indicates that the reimbursement claimant's historical reimbursement behavior contains irregularities in reimbursement, the greater the number of irregularities in the historical reimbursement behavior, the higher the risk index of the preset reimbursement information under the reimbursement claimant risk assessment dimension.

[0090] If the preset reimbursement dimension includes an organizational risk assessment dimension, the preset reimbursement condition corresponding to the organizational risk assessment dimension may include that the organizational reimbursement amount corresponding to the reimbursement amount is within the organizational budget range of the organization, thereby assessing whether the organizational reimbursement amount included in the preset reimbursement information is at risk or potentially at risk. The organizational reimbursement amount corresponding to the reimbursement amount may be calculated by adding the reimbursement amount to the organization's current reimbursement amount to determine whether the organizational reimbursement amount exceeds the organizational budget range.

[0091] In an exemplary embodiment, if the preset comparison result indicates that the organization's reimbursement amount does not exceed the organization's budget range, the risk index of the preset reimbursement information under the organizational risk assessment dimension may be determined to be low. If the preset comparison result indicates that the organization's reimbursement amount exceeds the organization's budget range, the risk index of the preset reimbursement information under the organizational risk assessment dimension may be determined to be high.

[0092] In the case where the preset reimbursement dimension includes a historical reimbursement trend evaluation dimension, the preset reimbursement condition corresponding to the historical reimbursement trend evaluation dimension may include that the change trend corresponding to the reimbursement amount is consistent with the historical reimbursement amount trend corresponding to the same reimbursement item, so as to evaluate whether there is a risk or potential risk in the change trend corresponding to the reimbursement amount included in the preset reimbursement information.

[0093] In one exemplary embodiment, if a preset comparison result indicates that the trend in the reimbursement amount does not conform to the historical reimbursement amount trend for the same reimbursement item, the greater the deviation between the trend and the historical reimbursement amount trend, the higher the risk index of the preset reimbursement information under the historical reimbursement trend assessment dimension. For example, if the reimbursement amount for a certain reimbursement item in a certain month increases by more than 50% of the historical reimbursement threshold, the risk assessment sub-model may determine that the trend in the reimbursement amount does not conform to the historical reimbursement amount trend. And so on.

[0094] The risk assessment sub-model can be used by the reimbursement review model to detect the risks or potential risks of the preset reimbursement information under multiple preset reimbursement dimensions, and then determine the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, which is conducive to improving the convenience and accuracy of risk identification of the preset reimbursement information.

[0095] In some embodiments, the risk score corresponding to the preset reimbursement information is determined based on the risk index of the preset reimbursement information under multiple preset reimbursement dimensions and the weight coefficients corresponding to each of the multiple preset reimbursement dimensions; when the risk score is less than the preset score threshold, the audit result corresponding to the preset reimbursement information is determined to be approved; when the risk score is greater than or equal to the preset score threshold, the audit result corresponding to the preset reimbursement information is determined to be pending manual review.

[0096] For example, the weight coefficients corresponding to different preset reimbursement dimensions may be different, the same, or partially the same, and there is no limitation here.

[0097] In an exemplary embodiment, the multiple preset reimbursement dimensions may include a reimbursement information risk assessment dimension, a reimbursement person risk assessment dimension, an organizational risk assessment dimension, and a historical reimbursement trend assessment dimension. For example, the weight coefficient corresponding to the reimbursement information risk assessment dimension is 0.4, the weight coefficient corresponding to the reimbursement person risk assessment dimension is 0.2, the weight coefficient corresponding to the organizational risk assessment dimension is 0.2, and the weight coefficient corresponding to the historical reimbursement trend assessment dimension is 0.2. Of course, this is not limited to this, and the weight coefficients may be pre-set, user-defined, or user-adjustable, without limitation here.

[0098] When the audit sub-model of the reimbursement audit model obtains the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, it can multiply the risk index under the preset reimbursement dimension with the corresponding weight coefficient, and determine the risk score corresponding to the preset reimbursement information based on the sum of the products of the risk index and the weight coefficient under multiple preset reimbursement dimensions.

[0099] The audit sub-model can compare the numerical value of the risk score with the preset score threshold. The preset score threshold can be pre-set, set by the user, or adjusted by the user, and there is no restriction here. In the case that the risk score is less than the preset score threshold, the audit sub-model can determine that the risk or potential risk of the preset reimbursement information is low, and the audit result corresponding to the preset reimbursement information can be determined to be approved. In the case that the risk score is greater than or equal to the preset score threshold, the audit sub-model can determine that the risk or potential risk of the preset reimbursement information is high, and the audit result corresponding to the preset reimbursement information can be determined to be pending manual review.

[0100] Of course, it is not limited to this. The preset scoring threshold can be further divided into a first scoring threshold and a second scoring threshold. The first scoring threshold is less than the second scoring threshold. When the risk score is less than the first scoring threshold, it can be determined that the audit result corresponding to the preset reimbursement information is approved. When the risk score is greater than or equal to the first scoring threshold, and less than the second scoring threshold, it can be determined that the audit result corresponding to the preset reimbursement information is pending manual review. When the risk score is greater than or equal to the second scoring threshold, it can be determined that the audit result corresponding to the preset reimbursement information is pending manual review, and a key review mark is added to the preset reimbursement information, so that when the auditor reviews the preset reimbursement information to be manually reviewed, he can focus on the preset reimbursement information and follow up inquiries based on the key review mark carried by the preset reimbursement information. There is no limitation here.

[0101] When the reimbursement review model can determine the review results corresponding to the preset reimbursement information through the review sub-model, it is helpful to improve the convenience of reviewing the preset reimbursement information.

[0102] In some embodiments, the audit result includes prompt information; when the risk index of the preset reimbursement information under the preset reimbursement dimension is greater than or equal to the preset risk index threshold, the prompt information is output, and the prompt information is used to indicate that the preset reimbursement dimension with a risk index greater than or equal to the preset risk index threshold is at least one of the key audit dimension corresponding to the preset reimbursement information and the information supplementary dimension corresponding to the preset reimbursement information.

[0103] For example, when the risk index of the preset reimbursement information under the preset reimbursement dimension is greater than or equal to the preset risk index threshold, the audit sub-model of the reimbursement audit model can determine that the preset reimbursement information has a high possibility of risk or potential risk under the preset reimbursement dimension. Then, when the audit sub-model outputs the audit result, it can output corresponding prompt information, so that when the auditor manually reviews the preset reimbursement information, the preset reimbursement dimension with a risk index greater than or equal to the preset risk index threshold can be focused on, that is, the key audit dimension; or when the reimburser supplements the preset reimbursement information, the information can be supplemented based on the preset reimbursement dimension with a risk index greater than or equal to the preset risk index threshold, that is, the information supplement dimension.

[0104] For example, if the pre-set reimbursement information includes invoices and the pre-set reimbursement dimensions include the reimbursement information risk assessment dimension, when the reimbursement amount indicated on the invoice exceeds the budget range corresponding to the reimbursement item, the review sub-model can output a corresponding prompt message to prompt the reviewer to focus on the reimbursement amount exceeding the budget range during the manual review process.

[0105] For example, if the pre-set reimbursement information includes a payment voucher and the pre-set reimbursement dimension includes a historical reimbursement trend assessment dimension, if the hotel expense indicated on the payment voucher does not conform to the historical reimbursement amount trend, such as an increase of more than 20% compared to the historical hotel expenses of the previous few months, the review sub-model can output a corresponding prompt message, prompting the reimbursement requester to provide supporting documentation of the increase in hotel expenses, such as an official hotel price list, to assist the review sub-model in reviewing the payment voucher.

[0106] Of course, it is not limited to this and no limitation is made here.

[0107] When the audit sub-model of the reimbursement audit model outputs prompt information based on the risk index of the preset reimbursement information under the preset reimbursement dimension, the prompt information can be used as a basis for the auditor to manually review the preset reimbursement information, and as a basis for the reimbursement person to supplement the preset reimbursement information, which is conducive to promoting the intelligence and automation of the reimbursement audit process, and thereby improving the convenience of auditing the preset reimbursement information.

[0108] It can be seen that in the above embodiments, for entities under the reimbursement review business, such as a platform under a digital financial system for reimbursement review of expenses incurred in connection with work-related consumption activities, a platform under a digital medical system for reimbursement review of expenses incurred in connection with medical treatment, etc., the preset reimbursement information to be reviewed can be obtained; based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information; based on the risk assessment sub-model of the reimbursement review model, the risk index of the preset reimbursement information under multiple preset reimbursement dimensions is determined according to the feature information; based on the review sub-model of the reimbursement review model, the preset reimbursement information is reviewed according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain the review result corresponding to the preset reimbursement information.

[0109] Based on the setting of the reimbursement review model, the computer device can determine the risk index of the preset reimbursement information under multiple preset reimbursement dimensions through the reimbursement review model, and determine the review result corresponding to the preset reimbursement information by comprehensively considering the risk index of the preset reimbursement information under multiple preset reimbursement dimensions. In the process of reviewing the results corresponding to the preset reimbursement information, there is no need for reviewers to evaluate the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, and to determine the review results corresponding to the preset reimbursement information, which is conducive to saving manpower and thus conducive to improving the convenience of reviewing the reimbursement information. Accordingly, the reimbursement review model can identify the risks existing in the preset reimbursement information based on the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, which is conducive to improving the review effect of the reimbursement information.

[0110] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean 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.

[0111] For example, a computer device can use a large model framework and deep learning to build a reimbursement review model. Multiple pre-set reimbursement information is input into the reimbursement review model for training. By adjusting the model parameters and optimizing the algorithm, the accuracy and generalization of the reimbursement review model for the pre-set reimbursement information can be improved. Cross-validation and regularization techniques can be used to prevent overfitting of the reimbursement review model and ensure its stability and reliability.

[0112] See also Figure 3 As shown, Figure 3 A flowchart of a method for training a reimbursement review model provided in an embodiment of the present application includes the following steps:

[0113] S201: Obtain a training sample set, where the training sample set includes a plurality of preset reimbursement information and an audit result label corresponding to each preset reimbursement information.

[0114] For example, when training a reimbursement review model, a training sample set can be obtained. The training sample set can include multiple preset reimbursement information. Each preset reimbursement information in the training sample set can be assigned a corresponding review result label. The training sample set can be used for subsequent training of the reimbursement review model. The preset reimbursement information can be determined based on invoices, contracts, payment receipts, medical fee payment receipts, drug fee payment receipts, electronic medical records, etc., without limitation.

[0115] In some embodiments, the preset reimbursement information may be determined based on one or more initial reimbursement information. When multiple initial reimbursement information is obtained, the reimbursement review model may use a relevance assessment sub-model to assess the relevance of the multiple initial reimbursement information to obtain the corresponding preset reimbursement information. The preset reimbursement information includes at least one initial reimbursement information, and there is a correlation between the multiple initial reimbursement information included in the preset reimbursement information. Accordingly, the training sample set may include multiple preset reimbursement information and the corresponding review result labels for the preset reimbursement information, to facilitate subsequent training of the reimbursement review model.

[0116] In some implementations, when preset reimbursement information is obtained, the audit result label corresponding to the preset reimbursement information can be determined by a preset marking method, which is not limited here.

[0117] In this way, when the preset reimbursement information and the audit result label corresponding to the preset reimbursement information are determined, the reimbursement audit model can be trained subsequently.

[0118] S202: Based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information.

[0119] For example, during training, the reimbursement review model needs to learn how to determine the feature information corresponding to the pre-set reimbursement information. Based on this, the pre-set reimbursement information can be input into the feature extraction sub-model of the reimbursement review model, which then extracts the features of the pre-set reimbursement information and obtains the feature information corresponding to the pre-set reimbursement information.

[0120] In some embodiments, the preset reimbursement information may be in the form of text, images, etc. The preset reimbursement information may include at least one of a reimbursement request and a reimbursement document. When the preset reimbursement information is input into the feature extraction sub-model, the feature extraction sub-model may learn how to determine the feature information corresponding to the preset reimbursement information based on at least one of the reimbursement request and the reimbursement document. Accordingly, the feature information corresponding to the preset reimbursement information may include at least one of the reimbursement item, reimbursement amount, reimbursement type, reimbursement time, reimbursement person, and historical reimbursement information corresponding to the reimbursement person. Of course, the feature information is not limited to this and is not limited here.

[0121] In this way, when the characteristic information corresponding to the preset reimbursement information is determined, the reimbursement review model can optimize the model parameters of the reimbursement review model based on the characteristic information to improve the training effect of the reimbursement review model.

[0122] S203: Based on the risk assessment sub-model of the reimbursement review model, determine the risk index of the preset reimbursement information under multiple preset reimbursement dimensions according to the characteristic information.

[0123] For example, during training, the reimbursement review model needs to learn how to determine the risk index of pre-set reimbursement information across multiple pre-set reimbursement dimensions. Based on this, feature information can be input into the risk assessment sub-model of the reimbursement review model. This sub-model then predicts the risk or potential risk of the pre-set reimbursement information across multiple pre-set reimbursement dimensions based on the feature information, thereby obtaining the risk index for the pre-set reimbursement information across these dimensions.

[0124] For example, the preset reimbursement dimensions may include at least two of the following: a reimbursement information risk assessment dimension, a reimbursement person risk assessment dimension, an organizational risk assessment dimension, and a historical reimbursement trend assessment dimension. Of course, the preset reimbursement dimensions are not limited thereto and are not intended to be limiting herein.

[0125] In this way, when the risk index of the preset reimbursement information under multiple preset reimbursement dimensions is determined, the reimbursement review model can comprehensively consider the risk index of the preset reimbursement information under multiple preset reimbursement dimensions and optimize the model parameters of the reimbursement review model to improve the training effect of the reimbursement review model.

[0126] S204: Based on the audit sub-model of the reimbursement audit model, the preset reimbursement information is audited according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain an audit result corresponding to the preset reimbursement information.

[0127] For example, during training, the reimbursement review model needs to learn how to determine the review results for pre-set reimbursement information. To this end, the risk index of the pre-set reimbursement information across multiple pre-set reimbursement dimensions can be input into the review sub-model. Based on this, the review sub-model then determines the review results for the pre-set reimbursement information by combining the risk indexes of the pre-set reimbursement information across multiple pre-set reimbursement dimensions.

[0128] In this way, when the audit results corresponding to the preset reimbursement information are determined, the reimbursement audit model can optimize the model parameters of the reimbursement audit model based on the audit results corresponding to the preset reimbursement information to improve the training effect of the reimbursement audit model.

[0129] S205: Determine the model optimization parameters of the reimbursement audit model according to the audit result and the audit result label.

[0130] For example, the audit result determined by the reimbursement audit model is equivalent to the model's predicted audit result for the preset reimbursement information. The audit result label corresponding to the preset reimbursement information is equivalent to the actual audit result determined by the reimbursement audit model for the preset reimbursement information. Based on the audit result and the audit result label, the loss value between the actual result and the predicted result when the reimbursement audit model determined the audit result for the preset reimbursement information can be determined.

[0131] In some embodiments, when training the reimbursement review model, a loss function can be used to measure the difference between the predicted result and the actual result, i.e., the loss value. The loss function can be expressed as:

[0132]

[0133] y i Used to indicate the audit result label corresponding to the preset reimbursement information. This value indicates the audit result corresponding to the preset reimbursement information determined by the reimbursement review model, and n indicates the number of preset reimbursement information in the training sample set. MSE measures the difference between the audit result label and the audit result. This is calculated by averaging the squared difference between the audit result label and the audit result. A smaller MSE indicates a closer match between the audit result label and the audit result, indicating better performance of the reimbursement review model.

[0134] The reimbursement review model can determine its optimization parameters based on the loss value. These optimization parameters can be used to subsequently adjust the model parameters of the reimbursement review model, enabling the model to more accurately align its predicted audit results for pre-set reimbursement information with the actual results, thereby achieving the training objective.

[0135] S206: Adjust the model parameters of the reimbursement review model according to the model optimization parameters.

[0136] For example, the reimbursement review model can adjust model parameters of the reimbursement review model based on model optimization parameters. For example, the reimbursement review model can adjust model parameters of at least one of the feature extraction sub-model, the risk assessment sub-model, and the review sub-model based on the model optimization parameters. Of course, this is not limited to this. For example, the model parameters of the reimbursement review model's relevance assessment sub-model can also be adjusted based on the model optimization parameters, which is not a limitation here.

[0137] Accordingly, accuracy indicators can be used to quantitatively evaluate the performance of the reimbursement review model, such as the training effect of the reimbursement review model.

[0138] The accuracy of the reimbursement review model may be determined based on the ratio of the number of true positives to the total number of true positives and false positives.

[0139] For example, if there are 100 pre-set reimbursement information in the training sample set and the corresponding audit results are "pending manual review", if 80 of the 100 pre-set reimbursement information have the audit result label "pending manual review", then the accuracy of the reimbursement audit model can be determined to be 80%. And so on.

[0140] In the process of adjusting the model parameters of the reimbursement review model, a real-time feedback mechanism can also be used to adjust the model parameters of the reimbursement review model. For example, the preset reimbursement information determined by the reimbursement review model can be manually reviewed to determine the review result corresponding to the preset reimbursement information. The reimbursement review model can monitor the review result corresponding to the preset reimbursement information in real time to determine whether the reimbursement review model has misjudgments and missed judgments based on the review result corresponding to the preset reimbursement information determined by itself and the preset review result determined by manual review. For example, when the review result corresponding to the preset reimbursement information is approved, and the review result is pending manual review, the reimbursement review model can mark the preset reimbursement information as having a misjudgment. For another example, when the reimbursement review model has not determined the review result corresponding to the preset reimbursement information, and the preset reimbursement information has a corresponding review result, the reimbursement review model can mark the preset reimbursement information as having missed judgments. Accordingly, the reimbursement review model can continue to add the preset reimbursement information marked as misjudgment or omission to the training sample set of the reimbursement review model, and dynamically update the model parameters of the reimbursement review model in combination with the online learning mechanism to continuously optimize and update the reimbursement review model, thereby improving the reimbursement review accuracy of the reimbursement review model for the preset reimbursement information.

[0141] Accordingly, the training sample set obtained by the reimbursement review model can also be updated regularly. For example, the reimbursement review model can collect new preset reimbursement information at preset time intervals to combine historical review result labels for incremental training of the reimbursement review model. This is not limited here.

[0142] By adjusting the model parameters of the reimbursement review model, the reimbursement review model can more accurately move the predicted results of the audit results corresponding to the preset reimbursement information closer to the actual results, so as to achieve the purpose of training, thereby helping to improve the accuracy of the reimbursement review model in determining the audit results corresponding to the preset reimbursement information.

[0143] In an exemplary embodiment, Figure 4 As shown, the reimbursement review model involves a reimbursement review process including: obtaining multiple preset reimbursement information to be reviewed; determining the review results corresponding to each preset reimbursement information through the reimbursement review model; judging whether there is preset reimbursement information whose review results are to be manually reviewed; if not, the corresponding reimbursement process can be executed for the preset reimbursement information whose review results are approved; if so, the corresponding prompt information is output for the reviewer to manually review, and / or the reimburser to supplement the preset reimbursement information; in the case of manual review of the preset reimbursement information, the reimbursement review model can determine whether to execute the reimbursement process corresponding to the preset reimbursement information based on the manual review results; in the case of executing the reimbursement process corresponding to the preset reimbursement information, the reimbursement of the preset reimbursement information is completed and recorded, and a corresponding reimbursement review report is generated; in the case of not executing the reimbursement process corresponding to the preset reimbursement information, a reimbursement rejection prompt is output and the reason for the reimbursement rejection is recorded; the reimbursement review model can continuously optimize the reimbursement review model based on the reimbursement review report.

[0144] In one embodiment, a device for reviewing reimbursement information is provided, which corresponds to the method for reviewing reimbursement information in the above embodiment. Figure 5 As shown, the reimbursement information review device includes an information acquisition module 110, a feature extraction module 120, a risk assessment module 130 and a review module 140. The functional modules are described in detail as follows:

[0145] The information acquisition module 110 is used to obtain preset reimbursement information to be reviewed;

[0146] A feature extraction module 120 is configured to perform feature extraction processing on the preset reimbursement information based on a feature extraction sub-model of the reimbursement review model to obtain feature information corresponding to the preset reimbursement information;

[0147] A risk assessment module 130 is configured to determine, based on the risk assessment sub-model of the reimbursement review model and the characteristic information, the risk index of the preset reimbursement information under multiple preset reimbursement dimensions;

[0148] The audit module 140 is used to audit the preset reimbursement information based on the audit sub-model of the reimbursement audit model and the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain the audit result corresponding to the preset reimbursement information.

[0149] In one embodiment, the information acquisition module 110 is configured to:

[0150] Obtain multiple initial reimbursement information for review;

[0151] Based on the correlation evaluation sub-model of the reimbursement review model, information correlation evaluation is performed on the multiple initial reimbursement information to obtain corresponding preset reimbursement information; the preset reimbursement information includes at least one initial reimbursement information, and there is a correlation relationship between the multiple initial reimbursement information included in the preset reimbursement information.

[0152] In one embodiment, the information acquisition module 110 is configured to:

[0153] Performing semantic reasoning on the initial reimbursement information to determine semantic information of the initial reimbursement information;

[0154] When it is determined based on the semantic information of at least two of the initial reimbursement information that there is at least one of a reimbursement time-space association relationship, a reimbursement matter association relationship, and a reimbursement amount association relationship between the at least two of the initial reimbursement information, the at least two of the initial reimbursement information are determined to be the same preset reimbursement information.

[0155] In one embodiment, the risk assessment module 130 is configured to:

[0156] Comparing the characteristic information with the preset reimbursement condition corresponding to the preset reimbursement dimension to obtain a preset comparison result of the preset reimbursement information under the preset reimbursement dimension;

[0157] According to the preset comparison result, the reimbursement risk corresponding to the preset reimbursement information in the preset reimbursement dimension is evaluated to obtain a risk index of the preset reimbursement information in the preset reimbursement dimension.

[0158] In one embodiment, the audit module 140 is configured to:

[0159] Determining a risk score corresponding to the preset reimbursement information according to the risk indexes of the preset reimbursement information under the plurality of preset reimbursement dimensions and the weight coefficients corresponding to the plurality of preset reimbursement dimensions;

[0160] When the risk score is less than a preset score threshold, determining that the audit result corresponding to the preset reimbursement information is audit passed;

[0161] When the risk score is greater than or equal to a preset score threshold, the audit result corresponding to the preset reimbursement information is determined to be pending manual review.

[0162] In one embodiment, the audit result includes prompt information; the audit module 140 is used to:

[0163] When the risk index of the preset reimbursement information under the preset reimbursement dimension is greater than or equal to the preset risk index threshold, a prompt message is output, and the prompt message is used to prompt that the preset reimbursement dimension with a risk index greater than or equal to the preset risk index threshold is at least one of the key review dimension corresponding to the preset reimbursement information and the information supplementary dimension corresponding to the preset reimbursement information.

[0164] The present application provides a reimbursement information review device, which obtains preset reimbursement information to be reviewed; performs feature extraction processing on the preset reimbursement information based on a feature extraction sub-model of a reimbursement review model to obtain feature information corresponding to the preset reimbursement information; determines the risk index of the preset reimbursement information under multiple preset reimbursement dimensions based on the feature information based on a risk assessment sub-model of the reimbursement review model; and reviews the preset reimbursement information based on the risk index of the preset reimbursement information under multiple preset reimbursement dimensions based on an audit sub-model of the reimbursement review model to obtain an audit result corresponding to the preset reimbursement information.

[0165] Based on the setting of the reimbursement review model, the computer device can determine the risk index of the preset reimbursement information under multiple preset reimbursement dimensions through the reimbursement review model, and determine the review result corresponding to the preset reimbursement information by comprehensively considering the risk index of the preset reimbursement information under multiple preset reimbursement dimensions. In the process of reviewing the results corresponding to the preset reimbursement information, there is no need for reviewers to evaluate the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, and to determine the review results corresponding to the preset reimbursement information, which is conducive to saving manpower and thus conducive to improving the convenience of reviewing the reimbursement information. Accordingly, the reimbursement review model can identify the risks existing in the preset reimbursement information based on the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, which is conducive to improving the review effect of the reimbursement information.

[0166] For example, the reimbursement review model can be used in different business scenarios to review pre-set reimbursement information in different business scenarios. For example, it can be used in financial business scenarios to review reimbursement information generated by business activities and procurement activities involved in financial business scenarios. Another example is that it can be used in healthcare business scenarios to review reimbursement information such as medical expense reimbursement, treatment expense reimbursement, and electronic medical records involved in healthcare business scenarios. Based on this, the reimbursement review model can be used to improve the convenience of reimbursement review of corresponding reimbursement information in financial business scenarios or healthcare business scenarios.

[0167] The specific definition of the reimbursement information review device can be found in the definition of the reimbursement information review method above and will not be repeated here. Each module in the aforementioned reimbursement information review device can be implemented in whole or in part through software, hardware, or a combination thereof. Each of the aforementioned modules can be embedded in or independent of a processor in a computer device in hardware form, or can be stored in a computer device memory in software form, so that the processor can call and execute the corresponding operations of each of the aforementioned modules.

[0168] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a service-side method for reviewing reimbursement information, or implements the functions or steps of a service-side method for training a reimbursement review model.

[0169] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 7As shown. The computer device includes a processor, memory, network interface, display screen and input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a client side of a method for reviewing reimbursement information, or implements the functions or steps of a client side of a method for training a reimbursement review model.

[0170] 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, the following steps are performed:

[0171] Obtain preset reimbursement information pending review;

[0172] Based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information;

[0173] Based on the risk assessment sub-model of the reimbursement review model, and according to the characteristic information, determining the risk index of the preset reimbursement information under multiple preset reimbursement dimensions;

[0174] Based on the audit sub-model of the reimbursement audit model, the preset reimbursement information is audited according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain an audit result corresponding to the preset reimbursement information.

[0175] In another embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are performed:

[0176] Obtaining a training sample set, the training sample set including a plurality of preset reimbursement information and an audit result label corresponding to each of the preset reimbursement information;

[0177] Based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information;

[0178] Based on the risk assessment sub-model of the reimbursement review model, and according to the characteristic information, determining the risk index of the preset reimbursement information under multiple preset reimbursement dimensions;

[0179] Based on the audit sub-model of the reimbursement audit model, the preset reimbursement information is audited according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain an audit result corresponding to the preset reimbursement information;

[0180] Determining model optimization parameters of the reimbursement review model according to the review result and the review result label;

[0181] According to the model optimization parameters, the model parameters of the reimbursement review model are adjusted.

[0182] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0183] Obtain preset reimbursement information pending review;

[0184] Based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information;

[0185] Based on the risk assessment sub-model of the reimbursement review model, and according to the characteristic information, determining the risk index of the preset reimbursement information under multiple preset reimbursement dimensions;

[0186] Based on the audit sub-model of the reimbursement audit model, the preset reimbursement information is audited according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain an audit result corresponding to the preset reimbursement information.

[0187] In another embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0188] Obtaining a training sample set, the training sample set including a plurality of preset reimbursement information and an audit result label corresponding to each of the preset reimbursement information;

[0189] Based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information;

[0190] Based on the risk assessment sub-model of the reimbursement review model, and according to the characteristic information, determining the risk index of the preset reimbursement information under multiple preset reimbursement dimensions;

[0191] Based on the audit sub-model of the reimbursement audit model, the preset reimbursement information is audited according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain an audit result corresponding to the preset reimbursement information;

[0192] Determining model optimization parameters of the reimbursement review model according to the review result and the review result label;

[0193] According to the model optimization parameters, the model parameters of the reimbursement review model are adjusted.

[0194] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can be found in the relevant descriptions of the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0195] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database 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), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0196] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by 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.

[0197] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A method for reviewing reimbursement information, characterized in that: include: Obtain preset reimbursement information pending review; Based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information; Based on the risk assessment sub-model of the reimbursement review model, and according to the characteristic information, determining the risk index of the preset reimbursement information under multiple preset reimbursement dimensions; Based on the audit sub-model of the reimbursement audit model, the preset reimbursement information is audited according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain an audit result corresponding to the preset reimbursement information.

2. The audit method according to claim 1, characterized in that: The obtaining of preset reimbursement information to be reviewed includes: Obtain multiple initial reimbursement information for review; Based on the correlation evaluation sub-model of the reimbursement review model, information correlation evaluation is performed on the multiple initial reimbursement information to obtain corresponding preset reimbursement information; the preset reimbursement information includes at least one initial reimbursement information, and there is a correlation relationship between the multiple initial reimbursement information included in the preset reimbursement information.

3. The audit method according to claim 2, characterized in that: The relevance evaluation sub-model based on the reimbursement review model performs information relevance evaluation on the plurality of initial reimbursement information to obtain corresponding preset reimbursement information, including: Performing semantic reasoning on the initial reimbursement information to determine semantic information of the initial reimbursement information; When it is determined based on the semantic information of at least two of the initial reimbursement information that there is at least one of a reimbursement time-space association relationship, a reimbursement matter association relationship, and a reimbursement amount association relationship between the at least two of the initial reimbursement information, the at least two of the initial reimbursement information are determined to be the same preset reimbursement information.

4. The audit method according to any one of claims 1 to 3, characterized in that: The risk assessment sub-model based on the reimbursement review model determines, according to the characteristic information, the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, including: Comparing the characteristic information with the preset reimbursement condition corresponding to the preset reimbursement dimension to obtain a preset comparison result of the preset reimbursement information under the preset reimbursement dimension; According to the preset comparison result, the reimbursement risk corresponding to the preset reimbursement information in the preset reimbursement dimension is evaluated to obtain a risk index of the preset reimbursement information in the preset reimbursement dimension.

5. The audit method according to any one of claims 1 to 3, characterized in that: The audit sub-model based on the reimbursement audit model audits the preset reimbursement information according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, and obtains the audit result corresponding to the preset reimbursement information, including: Determining a risk score corresponding to the preset reimbursement information according to the risk indexes of the preset reimbursement information under the plurality of preset reimbursement dimensions and the weight coefficients corresponding to the plurality of preset reimbursement dimensions; When the risk score is less than a preset score threshold, determining that the audit result corresponding to the preset reimbursement information is audit passed; When the risk score is greater than or equal to a preset score threshold, the audit result corresponding to the preset reimbursement information is determined to be pending manual review.

6. The audit method according to any one of claims 1 to 3, characterized in that: The audit result includes prompt information; The audit sub-model based on the reimbursement audit model audits the preset reimbursement information according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions, and obtains the audit result corresponding to the preset reimbursement information, including: When the risk index of the preset reimbursement information under the preset reimbursement dimension is greater than or equal to the preset risk index threshold, a prompt message is output, and the prompt message is used to prompt that the preset reimbursement dimension with a risk index greater than or equal to the preset risk index threshold is at least one of the key review dimension corresponding to the preset reimbursement information and the information supplementary dimension corresponding to the preset reimbursement information.

7. A training method for a reimbursement review model, characterized in that: The training method comprises: Obtaining a training sample set, the training sample set including a plurality of preset reimbursement information and an audit result label corresponding to each of the preset reimbursement information; Based on the feature extraction sub-model of the reimbursement review model, feature extraction processing is performed on the preset reimbursement information to obtain feature information corresponding to the preset reimbursement information; Based on the risk assessment sub-model of the reimbursement review model, and according to the characteristic information, determining the risk index of the preset reimbursement information under multiple preset reimbursement dimensions; Based on the audit sub-model of the reimbursement audit model, the preset reimbursement information is audited according to the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain an audit result corresponding to the preset reimbursement information; Determining model optimization parameters of the reimbursement review model according to the review result and the review result label; According to the model optimization parameters, the model parameters of the reimbursement review model are adjusted.

8. A device for reviewing reimbursement information, characterized in that: The audit device includes: An information acquisition module is used to obtain preset reimbursement information to be reviewed; A feature extraction module, configured to perform feature extraction processing on the preset reimbursement information based on a feature extraction sub-model of a reimbursement review model to obtain feature information corresponding to the preset reimbursement information; A risk assessment module, configured to determine, based on the risk assessment sub-model of the reimbursement review model and according to the characteristic information, the risk index of the preset reimbursement information under a plurality of preset reimbursement dimensions; The audit module is used to audit the preset reimbursement information based on the audit sub-model of the reimbursement audit model and the risk index of the preset reimbursement information under multiple preset reimbursement dimensions to obtain the audit results corresponding to the preset reimbursement information.

9. A computer device, characterized in that: The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is used to execute the computer program and implement the reimbursement information review method as described in any one of claims 1 to 6, or implement the reimbursement review model training method as described in claim 7 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for reviewing reimbursement information according to any one of claims 1 to 6 are implemented, or the steps of the method for training a reimbursement review model according to claim 7 are implemented.

Images