Claim settlement case impairment method and device, computer equipment and storage medium
By using a loss reduction prediction model in claims cases, the scale of loss reduction is automatically determined and a notification message is sent, which solves the problem of imbalance in loss reduction assessment and achieves efficient and low-cost claims processing.
Patent Information
- Application Number
- CN202511011108.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-21
- Publication Date
- 2025-11-07
AI Technical Summary
The current claims process suffers from an imbalance in the assessment of loss reduction, resulting in inaccurate claims amounts, increased costs and reduced efficiency, and reliance on professional personnel prevents the full utilization of human resources.
By inputting claims data into a pre-trained loss reduction prediction model, the model can determine the extent of loss reduction, send alerts, and mark case status, thereby reducing human intervention.
It improves the efficiency of assessing the extent of loss reduction, saves labor costs, makes full use of human resources, and improves claims efficiency.
Smart Images
Figure CN120912341A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of artificial intelligence, and is suitable for a financial scenario, in particular to a claim case damage reduction method and device, computer equipment and a storage medium. BACKGROUND
[0002] With the rapid development of insurance business, the processing efficiency and accuracy of claim business have become the focus of insurance companies. The post-damage link of claim is to reduce the claim amount as much as possible to achieve the purpose of damage reduction. The part of the claim amount reduced is the damage scale. The current damage scale of the case is mainly determined by professional claim personnel based on experience. However, most claim personnel do not have the ability to accurately judge the damage scale, which causes the imbalance of damage scale judgment in claim cases, resulting in inaccurate claim amount and increased cost. In addition, if the damage scale is entirely dependent on professional claim personnel for judgment, it cannot fully utilize all human resources, which not only increases the cost of human resources, but also affects the efficiency of claim. SUMMARY
[0003] The present application provides a claim case damage reduction method, device, computer equipment and storage medium to solve the technical problems of high compensation and human cost and low claim efficiency of the existing claim case damage reduction method.
[0004] In a first aspect, a claim case damage reduction method is provided, which comprises:
[0005] Obtaining claim data of a to-be-audited claim case, the claim data comprising a claim amount, product information and customer information;
[0006] Inputting the claim data into a pre-trained damage scale prediction model to obtain a damage scale of the to-be-audited claim case;
[0007] If the damage scale is greater than a preset threshold, marking the to-be-audited claim case as an audit failure case and sending a prompt message to a claim personnel, the prompt message comprising the damage scale.
[0008] In a second aspect, a claim case damage reduction device is provided, which comprises:
[0009] A claim data acquisition module for acquiring claim data of a to-be-audited claim case, the claim data comprising a claim amount, product information and customer information;
[0010] A damage scale acquisition module for inputting the claim data into a pre-trained damage scale prediction model to obtain a damage scale of the to-be-audited claim case;
[0011] The auditing judgment module is configured to, if the loss scale is greater than the preset threshold, mark the case to be audited as a failed case and send a prompt message to the claim personnel, wherein the prompt message includes the loss scale.
[0012] In a third aspect, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the claim case loss method when executing the computer program.
[0013] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program, and the computer program implements the steps of the claim case loss method when executed by a processor.
[0014] In the scheme implemented by the claim case loss method, the device, the computer device, and the storage medium, the claim data including the claim amount, the product information, and the customer information are input into the pre-trained loss scale prediction model to obtain the loss scale of the case to be audited, and if the loss scale is greater than the preset threshold, the case to be audited is marked as a failed case, and a prompt message including the loss scale is sent to the claim personnel. Without manually judging the loss scale, the human cost is saved, and the efficiency of judging the loss scale is improved. Meanwhile, non-professional claim personnel can also verify the claim case according to the loss scale in the reference prompt message, fully utilize all human resources, save the claim cost of the insurance company, and improve the claim efficiency. BRIEF DESCRIPTION OF DRAWINGS
[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the description of the embodiments of the present application. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.
[0016] Figure 1 is an application environment schematic diagram of the claim case loss method provided by an embodiment of the present application.
[0017] Figure 2 is an implementation flow schematic diagram of the claim case loss method provided by an embodiment of the present application.
[0018] Figure 3 is Figure 2 is a flow schematic diagram of another specific implementation of step S32 in the method.
[0019] Figure 4 is Figure 3 is a flow schematic diagram of another specific implementation of step S32 in the method.
[0020] Figure 5 is Figure 3 Another specific embodiment flowchart of step S33 in the method.
[0021] Figure 6 is Figure 2 Another specific embodiment flowchart after step S20 in the method.
[0022] Figure 7 is Figure 6 Another specific embodiment flowchart after step S40 in the method.
[0023] Figure 8 is a schematic diagram of the claim case reduction device in an embodiment of the present application
[0024] Figure 9 is a structural schematic diagram of the computer device in an embodiment of the present application. DETAILED DESCRIPTION
[0025] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.
[0026] The claim case reduction method provided by the embodiments of the present application can be applied in an application environment such as Figure 1 , and Figure 1 is a schematic diagram of an application environment of the claim case reduction method provided by an embodiment of the present application, as shown in Figure 1 , which includes a terminal 10 and a server end 20. The terminal 10 is a terminal device used by a claim personnel, and the server end 20 is a server end terminal device. The terminal 10 and the server end 20 jointly execute the claim case reduction method. It should be noted that the terminal 10 and the server end 20 can be a smart phone, a tablet computer, a notebook computer, a desktop computer, etc., but are not limited thereto. The terminal 10 and the server end 20 can be connected through Bluetooth, a universal serial bus (USB) or other communication connection modes, which are not limited in the present application.
[0027] Figure 2 is a specific embodiment flowchart of the claim case reduction method provided by an embodiment of the present application, as shown in Figure 2 , which can specifically include the following steps:
[0028] S10, obtaining claim data of a to-be-audited claim case, the claim data comprising a claim amount, product information and customer information.
[0029] In this step, the claim data of the to-be-audited claim case is obtained. The claim data comprises a claim amount, product information and customer information. The claim amount refers to the specific amount of money required by the applicant for compensation or the specific amount of money initially determined by the claim personnel. The product information includes the type of insurance product, the scope of protection, the term of protection, etc. The customer information includes the customer's age, occupation, previous claim record, financial status and other relevant information.
[0030] S20, inputting the claim data into a pre-trained loss scale prediction model to obtain the loss scale of the to-be-audited claim case.
[0031] In this step, the claim data is inputted into a pre-trained loss scale prediction model to obtain the loss scale of the to-be-audited claim case. The loss scale refers to the amount of claim that can be reduced after the audit, which is an important indicator for measuring the risk of claim.
[0032] S30, if the loss scale is greater than a preset threshold, marking the to-be-audited claim case as an audit failure case and sending a prompt message to the claim personnel, the prompt message comprising the loss scale.
[0033] In this step, it is determined whether the loss scale is greater than a preset threshold. The preset threshold can be determined according to the risk management strategy of the insurance company and historical data analysis, and is usually set to a value that can effectively screen out high-risk cases. If the loss scale is greater than the preset threshold, the to-be-audited claim case is marked as an audit failure case, and a prompt message is sent to the claim personnel, the prompt message comprising the loss scale.
[0034] In one specific embodiment, the to-be-audited claim case A is a car insurance claim case, and the claim amount is 86500 yuan. The product information is high-end comprehensive insurance, the coverage is 200000 yuan, the scope of protection includes vehicle damage + third party, the delay in reporting the case is 48 hours, and the vehicle type is an imported luxury SUV. The customer information is: male, 32 years old; the number of historical claims in the past two years is 5 times; the financial status is excellent. The above claim data is inputted into a pre-trained loss scale prediction model to obtain a loss scale of 30000 yuan. If the preset threshold is 20000 yuan, the loss scale of the to-be-audited claim case A is greater than the preset threshold, and a prompt message of "there is a loss of 30000 yuan in this case, please verify the customer's materials and fees, confirm the loss amount, and resubmit the audit" is sent to the claim personnel of the to-be-audited claim case A.
[0035] In another specific embodiment, the to-be-audited claim case B is a health insurance claim case, and the claim amount is 32000 yuan. The product information is: high-end medical insurance; the guarantee range is full coverage of special ward and imported drugs; the report delay is 7 days. The customer information is: female, 58 years old; 4 times of the same type of hospitalization claim within 3 years. The above claim data is input into the pre-trained loss scale prediction model, and the loss scale is 25000 yuan. If the preset threshold is 15000 yuan, the loss scale of the to-be-audited claim case B is greater than the preset threshold, and the prompt information "there is a loss space of 25000 yuan in this case. Please verify the customer materials and fees, confirm the loss amount, and resubmit the audit" is sent to the claim personnel of the to-be-audited claim case B.
[0036] The claim case loss method of the embodiment inputs the claim data including the claim amount, product information and customer information into the pre-trained loss scale prediction model to obtain the loss scale of the to-be-audited claim case. If the loss scale is greater than the preset threshold, the to-be-audited claim case is marked as an audit failure case, and prompt information including the loss scale is sent to the claim personnel. There is no need to judge the loss scale by artificial means, which saves labor costs and improves the efficiency of judging the loss scale. At the same time, non-professional claim personnel can also verify the claim case according to the loss scale in the reference prompt information, fully utilize all human resources, save the claim cost of the insurance company, and improve the claim efficiency.
[0037] Figure 3 is Figure 2 The flowchart of one specific embodiment in step S20 is shown in FIG. 2. As shown in FIG. 2, in some embodiments, the training steps of the violation detection model are as follows: Figure 3
[0038] S31, obtaining historical claim data of historical claim cases, the historical claim data including claim amount, product information, customer information and loss scale.
[0039] In this step, the historical claim data of the historical claim cases is obtained, and the historical claim data includes the claim amount, the product information, the customer information and the loss scale. The historical data is usually derived from the records of the claim cases handled by the insurance company in the past, which contains complete claim process information and final results.
[0040] S32, processing the historical claim data to obtain a target sample set, and dividing the target sample set into a target training set, a target validation set and a target test set according to a preset proportion.
[0041] In this step, the target sample set is divided into a target training set, a target validation set and a target test set according to a preset proportion. Usually, a 7:2:1 or 8:1:1 proportion is used for division to ensure the scientificity of model training and evaluation.
[0042] S33, training the target sample set by using the XGboost algorithm to obtain a trained loss scale prediction model.
[0043] In this step, the target sample set is trained by using the XGboost algorithm to obtain a trained loss scale prediction model. XGBoost (eXtreme Gradient Boosting) is a high-efficiency machine learning algorithm based on gradient boosting decision tree, which can output key influencing variables through feature importance analysis to assist business decision optimization. Through actual test, the XGboost algorithm performs outstandingly in loss scale prediction in the loss scale prediction model of the embodiment.
[0044] In the above embodiment, the loss scale prediction model is trained by taking the historical claim data as the training set, which realizes the knowledge precipitation of the artificial judgment loss rule, converts the loss labor tilt problem into an automatic technical solution, saves the labor cost, promotes the achievement of the individual value contribution index of the claim staff, and saves the claim cost of the insurance company.
[0045] In one specific embodiment, the historical claim data is obtained by setting a time window to move and filter regularly, such as setting the historical claim data set filtering condition as the claim data of the historical claim cases in the last three years. The purpose is to enable the latest loss success cases to participate in training as features in time, and when the claim settlement is completed, new claims also enter the database for model training, so as to ensure the accuracy and effectiveness of the model with version iteration prediction model, and facilitate the next claim loss judgment.
[0046] Figure 4 is Figure 3 Another specific embodiment flowchart of step S32 is shown in FIG. 3B, in which, in some embodiments, the historical claim data is processed in step S32 to obtain a target sample set, specifically including: Figure 4
[0047] S321, performing outlier processing on the historical claim data to obtain a first sample set, performing missing value processing on the first sample set to obtain a second sample set, performing feature selection and feature importance sorting on the second sample set to obtain the target sample set.
[0048] In this step, the outlier processing mainly adopts the box plot method and Z-score method to identify and process abnormal data points, including eliminating abnormal values such as claim amount exceeding the coverage. The missing value processing adopts methods such as mean / median filling, nearest neighbor filling or prediction model filling according to the characteristics of the data. In the feature selection process, the most valuable feature variables for damage scale prediction are selected through correlation analysis, variance analysis and other methods, such as claim frequency > claim amount > financial stability level.
[0049] Figure 5 is Figure 3 Another specific embodiment of step S33 is shown in the flowchart of Fig. 6. As shown in Fig. 6, in step S33, the XGboost algorithm is used to train the target sample set to obtain a trained damage scale prediction model, which includes: Figure 5
[0050] S331, using the XGboost algorithm to train the target training set to obtain a preliminarily trained damage scale prediction model;
[0051] S332, verifying and adjusting the preliminarily trained damage scale prediction model according to the target verification set to obtain a verification-completed damage scale prediction model;
[0052] S333, testing and evaluating the verification-completed damage scale prediction model according to the target test set to obtain a test evaluation result, and obtaining a trained damage scale prediction model when it is detected that the test evaluation result meets a preset requirement.
[0053] In this embodiment, the XGboost algorithm is used to train the target sample set to obtain a trained damage scale prediction model. The specific training process includes three stages: first, the XGboost algorithm is used to train the target training set to obtain a preliminarily trained damage scale prediction model; second, the preliminarily trained damage scale prediction model is verified and adjusted according to the target verification set to obtain a verification-completed damage scale prediction model, which mainly optimizes the model performance by adjusting the learning rate, the depth of the tree, the regularization parameter and other hyperparameters; finally, the verification-completed damage scale prediction model is tested and evaluated according to the target test set to obtain a test evaluation result, and a trained damage scale prediction model is obtained when it is detected that the test evaluation result meets a preset requirement. The preset requirement usually includes that the model accuracy, precision, recall, F1 score and other indicators reach a specific threshold.
[0054] In some embodiments, the prompt information in step S30 further includes:
[0055] Customer tags and prompt text and communication scripts generated based on the customer tags; the customer tags are generated based on the customer information using the K-means clustering algorithm, and the customer tags include high-risk management customers, low-risk management customers, financially stable customers, and financially stressed customers.
[0056] In this embodiment, customer tags are generated based on customer information using a K-means clustering algorithm. These tags represent the customer's personalized characteristics, including but not limited to high-risk management customers, low-risk management customers, financially stable customers, and financially strained customers. This allows claims personnel to select appropriate communication scripts based on the customer's individual characteristics when communicating with them. For different customer tags, corresponding prompts and communication scripts are automatically generated based on historical claims data, helping claims personnel communicate more effectively with customers, improving claims processing efficiency and customer satisfaction. In other embodiments, customer tags can also be used as samples for training and prediction of loss reduction prediction models.
[0057] In one specific implementation, when dealing with high-risk management clients, the following prompt text and communication script are generated:
[0058] Warning text: This client has multiple claims / high-value claims / complex cases / suspected fraud. Please carefully assess client risks and pay attention to your communication methods. The following script is for reference:
[0059] ① Regarding the XX yuan fee you mentioned, Clause 2.3 stipulates that three conditions must be met: Condition a, Condition b, and Condition c (broken down item by item). We have noticed [specific missing materials / non-compliance items] and suggest you provide [specific documents] to expedite the review. If these documents are indeed unavailable, we can also initiate a special case review process. ② We understand your dissatisfaction with the processing time; if I were in your shoes, you would likely feel the same way. The case is currently stuck at [specific stage], and we are coordinating with [specific departments / resources]. We expect to provide you with written progress feedback by the end of tomorrow at the latest. Is this solution acceptable to you? ③ During the review process, we noticed [specific anomalies, such as conflicting treatment times / loss amounts deviating from market prices]. According to regulatory requirements, these details need to be reviewed. Please provide [specific supporting documents]. The entire review process will not affect the normal claims process and will take approximately 3 business days. ④ You can rate this service; we highly value your feedback.
[0060] In another specific embodiment, when dealing with low-risk management clients, the following prompt text and communication script are generated:
[0061] This customer has a good claims history, a clear case, and complete documentation, classifying them as a low-risk customer. Please focus on rapid response, streamlined processes, and enhanced trust, employing the following communication techniques to improve the service experience:
[0062] ①Your claim application materials have been preliminarily reviewed, and only [specific file name] is missing. To save your time, you can submit the electronic version directly through the
APP upload portal
claim progress link
[0063] Figure 6 Yes Figure 2 A specific implementation mode flowchart after step S20 is shown in FIG. 4. After step S20, it further includes: Figure 6
[0064] S40, if the loss scale is less than the preset threshold, marking the to-be-reviewed claim case as a passed case and starting a claim payment process.
[0065] In this embodiment, if the loss scale is less than the preset threshold, the to-be-reviewed claim case is marked as a passed case and the claim payment process is started. In this way, low-risk cases can be quickly processed, and the claim efficiency can be improved.
[0066] Figure 7 Yes Figure 6 A specific implementation mode flowchart after step S40 is shown in FIG. 5. In some embodiments, after step S40, it further includes: Figure 7
[0067] S50, using the claim data in the passed case as training data to iteratively update the pre-trained loss scale prediction model.
[0068] In the embodiments provided by the invention, for the passed cases, the claim data thereof is used as training data to iteratively update the pre-trained loss scale prediction model. By continuously introducing new case data, the model can continuously learn and optimize, and the prediction accuracy can be improved.
[0069] The claim case reduction method provided by the embodiment of the present application can be constructed based on artificial intelligence, and the related data is acquired and processed based on artificial intelligence technology, so as to realize unattended claim case reduction. The artificial intelligence (AI) is the theory, method, technology and application system for simulating, extending and expanding human intelligence by using a digital computer or a machine controlled by a digital computer, perceiving environment, acquiring knowledge and using knowledge to obtain the best results. The artificial intelligence basic technology generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics and the like. The artificial intelligence software technology mainly includes computer vision technology, robot technology, biometric identification technology, speech processing technology, natural language processing technology and machine learning / deep learning and the like.
[0070] In an embodiment, a claim case reduction device is provided, which corresponds to the claim case reduction method in the above-mentioned embodiments. As shown in the figure, the claim case reduction device comprises: Figure 8
[0071] The claim data acquisition module 100 is configured to acquire claim data of a to-be-audited claim case, wherein the claim data comprises a claim amount, product information and customer information.
[0072] The reduction scale acquisition module 200 is configured to input the claim data into a pre-trained reduction scale prediction model to obtain a reduction scale of the to-be-audited claim case.
[0073] The audit determination module 300 is configured to, if the reduction scale is greater than a preset threshold, mark the to-be-audited claim case as an audit failure case and send a prompt information to a claim personnel, wherein the prompt information comprises the reduction scale.
[0074] Specifically, the reduction scale acquisition module 200 comprises a model training unit.
[0075] The model training unit is configured to acquire historical claim data of historical claim cases, wherein the historical claim data comprises a claim amount, product information, customer information and a reduction scale.
[0076] The model training unit is further configured to process the historical claim data to obtain a target sample set, and divide the target sample set into a target training set, a target verification set and a target test set according to a preset proportion.
[0077] The model training unit is further configured to train the target sample set by using an XGboost algorithm to obtain a trained reduction scale prediction model.
[0078] Specifically, the model training unit is further configured to perform outlier processing on the historical claim data to obtain a first sample set, perform missing value processing on the first sample set to obtain a second sample set, and perform feature selection and feature importance sorting on the second sample set to obtain the target sample set.
[0079] Specifically, the model training unit is further configured to train the target training set by using an XGboost algorithm to obtain a preliminarily trained loss scale prediction model, verify and adjust the preliminarily trained loss scale prediction model according to the target verification set to obtain a verification-completed loss scale prediction model, and test and evaluate the verification-completed loss scale prediction model according to the target test set to obtain a test evaluation result, and obtain a trained loss scale prediction model when the test evaluation result meets a preset requirement.
[0080] Specifically, the prompt information in the review and determination module 300 further includes a customer label and prompt text and communication rhetoric generated based on the customer label; the customer label is generated based on the customer information by using a K-means clustering algorithm, and the customer label includes a high-risk management customer, a low-risk management customer, a financially stable customer, and a financially tight customer.
[0081] Specifically, the review and determination module 300 is further configured to, if the loss scale is less than a preset threshold, mark the to-be-reviewed claim case as a review-passed case and start a claim payment process.
[0082] Specifically, the device further includes an iterative updating module.
[0083] The iterative updating module is configured to use claim data in the review-passed case as training data to iteratively update the pre-trained loss scale prediction model.
[0084] The specific limitations of the claim case loss reduction device can be referred to the limitations of the claim case loss reduction method described above, and will not be repeated here. Each module in the above claim case loss reduction device can be realized by software, hardware, and a combination thereof, in whole or in part. The above modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to each module.
[0085] Figure 9 For an embodiment of the structure of the computer device according to the present application, as shown in Figure 9 The present application further provides a computer device, which comprises:
[0086] A memory and a processor, the processor storing computer readable instructions, the computer readable instructions being executed by the processor to cause the processor to perform any of the steps of the claim case reduction method.
[0087] The application also provides a computer readable storage medium, the computer readable instructions being executed by one or more processors to cause the one or more processors to perform any of the steps of the claim case reduction method. It should be understood that the readable storage medium in the embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.
[0088] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.
Claims
1. A claim case reduction method, characterized by, The method comprises the following steps: obtaining claim data of a claim case to be audited, wherein the claim data comprises a claim amount, product information and customer information; inputting the claim data into a pre-trained loss scale prediction model to obtain a loss scale of the claim case to be audited; if the loss scale is greater than a preset threshold, marking the claim case to be audited as a failed audit case and sending a prompt message to a claim personnel, wherein the prompt message comprises the loss scale.
2. The claims case reduction method of claim 1, wherein, The training method of the loss scale prediction model comprises the following steps: obtaining historical claim data of historical claim cases, wherein the historical claim data comprises a claim amount, product information, customer information and a loss scale; processing the historical claim data to obtain a target sample set, and dividing the target sample set into a target training set, a target validation set and a target test set according to a preset proportion; training the target sample set by using an XGboost algorithm to obtain a trained loss scale prediction model.
3. The claims case reduction method of claim 2, wherein, The processing of the historical claim data to obtain the target sample set comprises the following steps: performing outlier processing on the historical claim data to obtain a first sample set, performing missing value processing on the first sample set to obtain a second sample set, performing feature selection and feature importance sorting on the second sample set to obtain the target sample set.
4. The claims case reduction method of claim 2, wherein, The training of the target sample set by using the XGboost algorithm to obtain the trained loss scale prediction model comprises the following steps: training the target training set by using the XGboost algorithm to obtain a preliminarily trained loss scale prediction model; verifying and adjusting the preliminarily trained loss scale prediction model according to the target validation set to obtain a verified loss scale prediction model; testing and evaluating the verified loss scale prediction model according to the target test set to obtain a test evaluation result, and obtaining the trained loss scale prediction model when it is detected that the test evaluation result meets a preset requirement.
5. The claims case reduction method of claim 1, wherein, The prompt message further comprises: a customer label and prompt text and communication tactics generated based on the customer label; the customer label is generated based on the customer information by using a K-means clustering algorithm, and the customer label comprises a high-risk management customer, a low-risk management customer, a financially stable customer and a financially tight customer.
6. The claims case reduction method of claim 1, wherein, After obtaining the loss scale of the claim case to be audited, the method further comprises the following steps: if the loss scale is less than the preset threshold, marking the claim case to be audited as a passed audit case and starting a claim payment process.
7. The claims case reduction method of claim 6, wherein, After marking the claim case to be audited as the passed audit case and starting the claim payment process, the method further comprises the following steps: iteratively updating the pre-trained loss scale prediction model by using the claim data in the passed audit case as training data.
8. A claim case reduction apparatus characterized by comprising: The method comprises the following steps: a claim data obtaining module is configured to obtain claim data of a claim case to be audited, wherein the claim data comprises a claim amount, product information and customer information; a loss scale obtaining module is configured to input the claim data into a pre-trained loss scale prediction model to obtain a loss scale of the claim case to be audited; and The auditing judgment module is configured to mark the to-be-audited claim case as a failed case and send a prompt message to a claim personnel if the loss scale is greater than a preset threshold, wherein the prompt message comprises the loss scale.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the steps of the claim case loss scale method according to any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. The computer program is executed by the processor to implement the steps of the claim case loss scale method according to any one of claims 1 to 7.