Ai-based insurance claim information generation method and device

Abstract

A method by which an AI-based insurance claim information generation device generates AI-based insurance claim information is disclosed. According to one embodiment, the method may comprise the steps of: receiving estimate information from a repair shop; extracting text elements from the estimate information and embedding the extracted text elements; extracting standard-form estimate elements from a pre-established vector DB on the basis of vector values generated through embedding of the estimate elements from among the embedding results; generating, on the basis of the received estimate information, a standard-form estimate in which the standard-form estimate elements are reflected; predicting the probability of approval of the generated standard-form estimate by using a pre-established approval prediction model; and generating insurance claim information on the basis of the prediction result.

Description

AI-based insurance claim information generation method and device

[0001] The present invention relates to a method and apparatus for generating insurance claim information based on AI. More specifically, it relates to a method and apparatus for analyzing repair shop estimates through AI-based image and text analysis and generating insurance claim information based thereon.

[0002] When repairing a vehicle, while costs may vary slightly depending on the repair shop, the repair duration, parts prices, and labor costs are generally set similarly.

[0003] However, in the case of used cars, repair costs can vary depending on whether reused or new parts are used. Furthermore, repair estimates may include items that are denied insurance payouts based on whether the repair falls under the scope of the used car warranty. Consequently, verifying repair shop estimates requires significant effort and time when reviewing insurance claims. Additionally, from the insurance company's perspective, inconsistencies in the method of listing repair items and terminology on estimates among different workers at various shops lead to situations where the approval of payouts varies depending on the experience level of the insurance officer handling the claim.

[0004] As repair estimates vary from shop to shop, data-driven insurance design solutions are being developed in the life insurance sector, whereas automatic repair estimate generation and data-driven insurance design remain difficult in the automobile insurance sector.

[0005] The technical problem that the present invention aims to solve to resolve the above-mentioned issues is to provide an apparatus and method for automatically generating repair estimates that are easy to review and approve, so that they can be utilized by insurance companies.

[0006] Specifically, the technical problem that the present invention aims to solve is to provide a method and apparatus for automatically generating a repair estimate submitted by a repair shop, modified with items and costs within a pre-learned insurance approval scope.

[0007] In other words, another technical objective of the present invention is to provide a method and apparatus for determining an overall repair estimate by identifying repair items belonging to the insurance coverage matched for each insurance product / service at an insurance company and applying a pre-registered estimate range for each repair item.

[0008] In addition, another technical problem that the present invention aims to solve is to provide an apparatus and method for generating a repair estimate written in unified terms for verification by an insurance company representative.

[0009] Specifically, another technical problem that the present invention aims to solve is to provide a method and apparatus for automatically generating a technical opinion written in terms commonly used by insurance companies, based on vehicle information, vehicle type data, insurance information, etc.

[0010] Furthermore, another technical problem that the present invention aims to solve is to provide a method and apparatus for designing insurance products / services that can directly design variables affecting the insurance loss ratio.

[0011] The technical problems of the present invention are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by a person skilled in the art from the description below.

[0012] A method performed by an AI-based insurance claim information generation device according to an embodiment of the present invention for solving the above problem may include the steps of receiving estimate information from a repair shop, extracting character elements from the estimate information and embedding the extracted character elements, extracting standard form estimate elements from a pre-built vector DB based on vector values ​​generated through embedding of estimate elements among the embedding results, generating a standard form estimate reflecting the standard form estimate elements based on the received estimate information, predicting the possibility of approval of the generated standard form estimate using a pre-built approval prediction model, and generating insurance claim information based on the prediction result.

[0013] In one embodiment, the step of predicting the likelihood of approval of the generated standard form estimate may include a first step of determining whether the estimate element is within the coverage of the subscribed insurance product; a second step of determining whether the cost estimate corresponding to the estimate element among the extracted character elements is a value within a preset threshold range; and a step of predicting the likelihood of approval of the generated standard form estimate based on at least one of the first and second determination results.

[0014] In one embodiment, the step of generating the insurance claim information may include, based on the first judgment result, removing an estimate element that is not within the insurance scope, or based on the second judgment result, setting a cost estimate outside the threshold range to be changed to a value within the preset threshold range, and updating the standard form estimate by reflecting the set result.

[0015] In one embodiment, the estimation element includes at least one element among maintenance items, cost estimates, and technical opinion reports, and the step of extracting the standard form estimation element includes the step of extracting a standard form estimation element having a vector value within a similar range on the vector DB based on a vector value generated after embedding for at least one element, and the standard form estimation element is pre-matched with a vector within the similar range, and the step of generating the standard form estimation document may include the step of converting terms for the estimation element in the estimation document information to the extracted standard form estimation element.

[0016] In one embodiment, if, as a result of the first determination, a predetermined maintenance item among the estimate elements is not within the insurance coverage, the step of setting to change the predetermined maintenance item may include: a step of classifying the predetermined maintenance item into a first sub-maintenance item and a second sub-maintenance item; a step of determining whether any one of the classified first sub-maintenance item and second sub-maintenance item belongs to the insurance coverage; and a step of setting the predetermined maintenance item to change to the one sub-maintenance item if it is determined that any one of the sub-maintenance items belongs to the insurance coverage.

[0017] In one embodiment, the step of extracting character elements on the estimate information and embedding the extracted character elements includes the step of embedding maintenance items and technical opinion reports among the estimate elements, respectively, and the step of matching a first vector value embedded for the maintenance item with a second vector value embedded for the technical opinion report, and the step of classifying the predetermined maintenance item into a first sub-maintenance item and a second sub-maintenance item may include the step of classifying maintenance items having vector values ​​having a difference within a preset range based on the matching result, and the step of identifying the first sub-maintenance item and the second sub-maintenance item on the classified maintenance item.

[0018] In one embodiment, the step of determining whether either of the classified first sub-maintenance item and the second sub-maintenance item belongs to the insurance coverage may include the step of identifying information on maintenance items for which premium payment has been approved for an insurance product using an artificial intelligence model built by learning the insurer's decision-making made on the received estimate information, and the step of determining the maintenance item among the first sub-maintenance item and the second sub-maintenance item that belongs to the insurance coverage based on the identified information on the approved maintenance item.

[0019] An AI-based insurance claim information generation device according to another embodiment of the present invention for solving the above problem comprises one or more processors, a network interface for receiving estimate information, a memory for loading a computer program performed by said processors, and a storage for storing said computer program, wherein the computer program may include an operation for receiving estimate information of a repair shop, an operation for extracting character elements on said estimate information and embedding the extracted character elements, an operation for extracting standard form estimate elements from a pre-built vector DB based on a vector value generated through embedding of the estimate elements among the embedding results, an operation for generating a standard form estimate reflecting said standard form estimate elements based on said received estimate information, an operation for predicting the possibility of approval of said generated standard form estimate using a pre-built approval prediction model, and an operation for generating insurance claim information based on said prediction result.

[0020] According to one embodiment of the present invention, it becomes possible to secure high-quality training data that serves as the basis for automobile insurance design. Accordingly, by eliminating variations in the recording methods and terminology of repair estimates from individual repair shops and learning from a large volume of standardized estimates, an estimate generation service can be provided for use by automobile insurance companies in insurance design.

[0021] According to one embodiment of the present invention, the time required for insurance approval is significantly reduced, which has the advantage of facilitating the smooth operation of repair shops. Accordingly, there is an effect of eliminating overall insurance premium waste, such as the reduction of the rental car period.

[0022] According to one embodiment of the present invention, an insurance company representative is provided with maintenance items using unified terminology, thereby enabling an insurance premium payment review to be conducted regardless of the representative's experience and knowledge.

[0023] According to one embodiment of the present invention, specific values ​​of items in a repair estimate are filtered, thereby resolving the problem of overcharging insurance premiums.

[0024] In addition, according to another embodiment of the present invention, as data-based automobile insurance design is performed, there is an advantage in that an insurance product capable of lowering the insurance loss ratio can be designed by reflecting variables affecting the insurance loss ratio, such as the selection of repair parts and vehicle type characteristics.

[0025] The effects of the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by a person skilled in the art from the description below.

[0026] FIG. 1 is an exemplary diagram of an AI-based insurance claim information generation system according to one embodiment of the present invention.

[0027] FIG. 2 is a block diagram of an AI-based insurance claim information generation device according to another embodiment of the present invention.

[0028] FIG. 3 is an example of AI-based insurance claim information generation software according to another embodiment of the present invention.

[0029] FIG. 4 is a flowchart of an AI-based insurance claim information generation method according to another embodiment of the present invention.

[0030] FIG. 5 is a flowchart of a method for generating personal information-separated training data, which is referenced in some embodiments of the present invention.

[0031] FIG. 6 is a block diagram of an AI-based insurance design device according to another embodiment of the present invention.

[0032] FIG. 7 is a conceptual diagram of an insurance design method referenced in some embodiments of the present invention, and FIG. 8 is a flowchart of the insurance design method.

[0033] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings. The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the attached drawings. However, the present invention is not limited to the embodiments disclosed below but can be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims. Throughout the specification, the same reference numerals refer to the same components.

[0034] Unless otherwise defined, all terms used herein (including technical and scientific terms) may be used in a meaning commonly understood by those skilled in the art to which the present invention pertains. Furthermore, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise. The terms used herein are for describing embodiments and are not intended to limit the present invention. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text.

[0035] In this specification, descriptions such as "module," "unit," and "part" refer to units of software and / or hardware, for example, an embedding execution part may refer to a code package that performs the function of identifying text such as maintenance items on a maintenance estimate and numbers such as cost estimates, which are referenced in some embodiments of the present invention, and generating embedding vector values ​​based thereon.

[0036] The above hardware module / unit / part may be, for example, a hardware resource existing as a processor-specific unit for performing operations of a specific function. The above module / unit / part does not exist solely as a software module / unit / part or a hardware module / unit / part, but may also refer to a unit in which specific software and hardware are combined.

[0037] In this specification, the AI-based insurance claim information generation method and device may each be abbreviated as the insurance claim information generation method and device.

[0038] FIG. 1 is an exemplary diagram of an AI-based insurance claim information generation system according to one embodiment of the present invention.

[0039] Referring to FIG. 1, the AI-based insurance claim information generation system may include an insurance claim information generation device (100), a repair shop system (200), and an insurance company system (300).

[0040] Each component of FIG. 1 may be a computing device capable of communicating data with one another.

[0041] In particular, the insurance claim information generation device (100) may exist in a combined form with an AI model that automatically generates an estimate according to one embodiment of the present invention.

[0042] The repair shop system (200) may include a computing device of the repair shop and / or a terminal of a repair shop manager that generates and transmits insurance claim information including a repair estimate. Additionally, the repair shop system (200) may include software as an application applied to the computing device or terminal that generates a repair estimate in a computerized manner according to a predetermined form, or scans a handwritten repair estimate for a received vehicle to document the information.

[0043] In one embodiment, when an accident or breakdown is reported to a repair shop, the repair shop system (200) can generate vehicle information, insurance information, breakdown symptom information provided by the owner of the vehicle during the repair consultation, and an estimate of the estimated repair cost that are matched to identification information (insurance case number, vehicle number, owner information, etc.) for the reported accident (or breakdown).

[0044] The repair shop system (200) sends estimate information to the insurance claim information generation device (100). The estimate information generated by the repair shop system (200) may include information corresponding to various field values, such as a technical opinion report, a planned repair area, and a photo taken before repair.

[0045] The insurance claim information generating device (100) can learn each field value of the estimate information and the description of the maintenance worker's technical opinion report by performing learning on the provided estimate information. In one embodiment, the insurance claim information generating device (100) can identify at least some items on the estimate information and matching descriptions on the technical opinion report through learning, and to this end, the AI ​​model applied to the insurance claim information generating device (100) may include a language model widely known in the technical field to which the present invention belongs.

[0046] In addition, the insurance claim information generation device (100) performs learning on the insurance company's decision-making regarding the repair shop's estimate information.

[0047] The insurance claim information generation device (100) can learn the decision of the insurer regarding the repair shop's estimate information, including items to be repaired in the future and cost estimates, for example, the technical opinion and cost estimate provided by the repair shop and the insurer's on-site opinion regarding it.

[0048] Using an AI model generated based on such learning results, the insurance claim information generating device (100) can automatically generate insurance claim information including a technical opinion and an estimate based on the vehicle type, year of manufacture, and symptoms of failure, and transmit it to the insurance company system (300).

[0049] The insurance company system (300) may include a business interface for reviewing, providing opinions on, and making decisions regarding approval or rejection of insurance claims. After reviewing the insurance claim information displayed through the business interface of the insurance company system (300), the business officer of the insurance company may provide opinions and make decisions regarding approval or rejection.

[0050] In one embodiment, the insurance claim information generation device (100) can improve the accuracy of the AI ​​model that generates insurance claim information by continuously learning decision-making. Accordingly, the technical opinion and estimate on the generated insurance claim information can become increasingly accurate. That is, insurance claim information that conforms to or converges with the decision-making of the insurance company system (300) can be generated.

[0051] According to one embodiment, the insurance claim information displayed on the business interface may include information on whether each maintenance item is covered by insurance and information on the range of estimates eligible for insurance approval for each maintenance item in order to support the work of an insurance company representative. The information on whether each maintenance item is covered by insurance and the range of estimates may be visually identified and displayed on the insurance claim information according to a pre-set form.

[0052] The insurance claim information generation device (100) performs learning on the part name, part price, labor cost, photo after maintenance, and opinion attached when creating the estimate on the estimate, and learns together the actual inspection opinion and approved estimate on such estimate to set the estimate range for each individual or all maintenance items on each estimate.

[0053] In another embodiment, when the estimate range for maintenance items or the entire insurance claim is set, the insurance claim information generating device (100) may automatically approve or reject the estimate provided by the repair shop.

[0054] In addition, the insurance claim information generation device (100) can learn together the modified parts photos, modified estimates, and inspection opinions after actual maintenance is performed based on the final determined maintenance items and cost estimates after an inspection by an insurance company representative.

[0055] Through this, the difference between the estimate provided by the initial repair shop system (200) and the estimate approved by the final insurance company system (300) is learned, and in the future, when an estimate is submitted from the repair shop system (200), an inspection opinion may be automatically generated even if it is not provided by the insurance company system (300).

[0056] FIG. 2 is a block diagram of an AI-based insurance claim information generation device according to another embodiment of the present invention. FIG. 3 is an example of AI-based insurance claim information generation software according to another embodiment of the present invention.

[0057] Referring to FIG. 2, an AI-based insurance claim information generation device (100) may include one or more processors (101), a network interface (102) that receives a repair estimate from a repair shop system (200) and transmits the generated insurance claim information to an insurance company system (300), a memory (103) that loads a computer program (105) executed by the processor (101), and a storage (104) that stores the computer program (105).

[0058] The processor (101) controls the overall operation of each component of the AI-based insurance claim information generation device (100). The processor (101) may be configured to include a CPU (Central Processing Unit), MPU (Micro Processor Unit), MCU (Micro Controller Unit), AP (Application Processor), or any form of processor well known in the art of the present invention. Additionally, the processor (101) may perform operations for at least one application and / or program for executing the method according to the embodiments of the present invention.

[0059] The network interface (102) supports wired and wireless internet communication of the AI-based insurance claim information generation device (100). In addition, the network interface (102) may support various communication methods in addition to the internet, which is a public communication network. The network interface (102) may provide a connection with the repair shop system (200) and the insurance company system (300).

[0060] According to an embodiment of the present invention, the network interface (102) may form an interface with an artificial neural network widely known in the art to which the present invention belongs.

[0061] According to one embodiment, an artificial intelligence estimate analysis model can be constructed using a language model for character analysis on an artificial intelligence estimate among artificial neural networks connected via a network interface from an AI-based insurance claim information generation device (100).

[0062] The AI-based insurance claim information generation device (100) can extract a technical opinion and maintenance items reflecting maintenance opinions on a repair estimate obtained using an artificial intelligence estimate analysis model, and perform embedding therein to generate a vector value corresponding to the maintenance item.

[0063] According to another embodiment, the AI-based insurance claim information generation device (100) can learn maintenance items and cost estimates for each maintenance item by using an artificial neural network connected through a network interface. Through this learning, a cost estimate analysis model can be constructed to identify and extract maintenance items on an estimate and set the cost estimate range for each individual maintenance item.

[0064] The AI-based insurance claim information generation device (100) can determine the suitability of the cost estimate for maintenance items by using the constructed artificial intelligence cost estimate analysis model.

[0065] The memory (103) stores various data, commands and / or information. The memory (103) may load one or more programs (105) from storage (104) to execute embodiments of the present invention. In FIG. 2, the memory (103) may be, for example, RAM.

[0066] Storage (104) can store one or more of the programs (105) and reward data (106). In FIG. 2, insurance claim information generation software (105) is shown as an example of one or more of the programs (105).

[0067] In one embodiment, the storage (104) may store quotation information (106). The quotation information (106) may include quotations generated from an AI-based insurance claim information generation device (100) and / or quotations collected from a repair shop system (200).

[0068] The storage (104) may be configured to include non-volatile memory such as ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), flash memory, a hard disk, a removable disk, or any form of computer-readable recording medium well known in the art to which the present invention belongs.

[0069] Additionally, although the storage (104) in FIG. 2 is exemplified as being a component of the AI-based insurance claim information generation device (100), the embodiments of the present invention are not limited thereto and may exist as an external component of the AI-based insurance claim information generation device (100), such as a network-connected cloud.

[0070] According to an embodiment of the present invention, the insurance claim information generation software (105) can enable the insurance claim information generation method to be implemented by having the processor (101) of the AI-based insurance claim information generation device (100) execute each operation.

[0071] FIG. 3 is an example of insurance claim information generation software according to another embodiment of the present invention. Referring to FIG. 3, an AI-based insurance claim information generation device (100) can execute software (105).

[0072] FIG. 3 is an example of AI-based insurance claim information generation software according to another embodiment of the present invention, and

[0073] The software (105) may be configured to include a plurality of 'parts' of functional units. In FIG. 3, the software (105) may include an embedding execution unit (310), a standard form estimate generation unit (320), an insurer decision learning unit (330), an estimate update unit (340), and a maintenance estimate approval unit (350). Hereinafter, the operation of each component of the software (105) is an operation performed by the AI-based insurance claim information generation device (100) by the operation of each component by the processor (101), but for convenience of explanation, each component is described as operating.

[0074] The embedding execution unit (310) performs embedding by analyzing the text of the consultation record regarding the fault symptoms received by the repair shop system (200), the repair items on the estimate, and the technical opinion report regarding the repair items.

[0075] In one embodiment, the embedding execution unit (310) can convert maintenance items for a predetermined failure symptom and technical opinion reports thereon into vector values ​​and match them with each other.

[0076] To this end, the embedding execution unit (310) may include an artificial intelligence estimate analysis model. The embedding execution unit (310) analyzes multiple estimates received from the repair shop system (200), and a vector database may be constructed through the vectorization of failure symptoms, each repair item, and technical opinion reports regarding them.

[0077] In one embodiment, when a new estimate is received from the repair shop system (200), the embedding execution unit (310) can perform embedding for at least a portion of the new estimate, for example, repair items.

[0078] The AI-based insurance claim information generation device (100) can identify vector values ​​for estimate elements within a preset similar range on a constructed vector DB based on vector values ​​generated by the embedding process. Here, the estimate elements may include received failure symptoms, maintenance items, technical opinion reports, and cost estimates for maintenance items.

[0079] The AI-based insurance claim information generation device (100) can determine the suitability of other quotation elements in the new quotation based on quotation elements in a vector DB whose vector values ​​are within a preset similar range.

[0080] This explanation takes the example of a case where a new estimate is received due to failure symptom A, including maintenance item a1 for part X and maintenance item a2 for part Y.

[0081] The embedding execution unit (310) can generate an embedding vector value by extracting character elements including text and numbers on the new quotation. Here, the character elements are distinguished from images and include text, numbers, and special characters.

[0082] Specifically, vector values ​​for each quotation element on a new quotation can be generated and then matched with one another. Alternatively, each vector value can be combined to generate an integrated vector value for the new quotation.

[0083] The insurance claim information generation device (100) can search the vector DB and extract an estimate that includes a first similar vector value within a preset similarity range and a first vector value of a new estimate among the vector values ​​of existing registered estimates.

[0084] For example, the above-described extracted estimate may include maintenance item a11 for part X1 and maintenance item a3 for part Z as maintenance items due to failure symptom A. In this case, the first similar vector value may be the vector value for maintenance item a11 due to failure symptom A.

[0085] Accordingly, the insurance claim information generating device (100) can identify vector values ​​other than the first similar vector value among a plurality of vector values ​​for each estimation element on the extracted estimation sheet. In the above example, the vector value of maintenance item a3 for part Z can be identified.

[0086] The insurance claim information generating device (100) can perform a similarity determination for vector values ​​other than the first vector value on the new estimate based on the vector value of the identified a3. In the above example, the insurance claim information generating device (100) can determine the similarity of the vector value for a2 on the new estimate based on the vector value of a3.

[0087] Based on the above judgment result, if it is within a preset similarity range, the insurance claim information generating device (100) can determine that the failure symptoms, maintenance items, and parts list of the new estimate are suitable.

[0088] On the other hand, if the pre-set similarity range is exceeded, the insurance claim information generating device (100) may determine that the failure symptoms, maintenance items, and parts list on the new estimate are unsuitable and may generate a modified estimate with the maintenance items and the parts list changed.

[0089] When characters on an estimate received from the embedding execution unit (310) are embedded and similar vector values ​​are not identified even when the insurance claim information generation device (100) searches the vector DB, the insurance claim information generation device (100) can create a separate group on the vector DB and store and manage the estimate and the corresponding vector value within the created group.

[0090] The standard form estimate generation unit (320) can generate a standard form estimate that is commonly used by insurance companies based on the estimate received from the repair shop system (200).

[0091] The standard form quotation generation unit (320) can convert the description of each quotation element into a standard form description based on the embedding vector value for each quotation element, such as maintenance items, parts, technical opinion, and failure symptoms on the received quotation.

[0092] The insurance claim information generation device (100) can extract a vector value of a standard form entry corresponding to the embedding vector value of each estimate element, identify a standard form entry matched to the extracted vector value, and convert and generate the entry on the received estimate. For example, the term 'Ass', used for workers at some repair shops, is derived from the abbreviation Ass'y of Assembly, but there may be cases where the meaning of the term is not understood by the insurance company representative. In this case, by converting to a standard form entry, the insurance company can save time and cost in deciding whether to approve the insurance claim.

[0093] In one embodiment, the insurance claim information generating device (100) can send a converted standard form estimate to a repair shop system (200) to receive confirmation from a repair shop worker that the standard form estimate has been appropriately converted based on the previously provided estimate. That is, the repair shop system (200) can verify whether the estimate elements, such as failure symptoms, repair items, and parts list on the sent estimate match the estimate elements on the converted standard form estimate in terms of the insurance claim subject.

[0094] In another embodiment, the insurance claim information generation device (100) may provide an automatic estimate generation solution through an automatic estimate generation application on the repair shop system (200). In this case, the insurance claim information generation device (100) functions as a server, and the repair shop system (200) operates as a client. When a fault symptom is input into the repair shop system (200), the application identifies vector values ​​for maintenance items, parts lists, and technical opinions that match the vector values ​​for the input fault symptom, and can generate descriptions for each estimate element using a pre-trained generative language model. Thus, upon receiving the fault symptom, maintenance items, required parts lists, and technical opinions are automatically generated, making it easier for the repair shop worker to prepare the estimate.

[0095] In addition, by using a generative language model, an estimate written in a standard format can be generated in the repair shop system (200), thereby increasing the convenience of the insurance company system (300).

[0096] Next, the insurance company decision learning unit (330) can learn the insurance company's approval or rejection decision made regarding the estimate generated by the repair shop system (200).

[0097] For example, the insurance company decision learning unit (330) may learn by matching the technical opinion and cost estimate provided by the repair shop with the insurance company's actual inspection opinion thereon.

[0098] In addition, the insurance company decision learning unit (330) learns by matching information on the estimate elements, including the maintenance items and cost estimates on the estimate sheet, with the result of the decision (approval or rejection), so that the insurance claim information generating device (100) can generate the range of approved cost estimates for each maintenance item.

[0099] For example, the insurance company decision learning unit (330) can learn about the part names, part prices, labor costs, photos after maintenance, and opinions attached when creating the estimate, and learn together the actual opinion report and approved estimate for such estimate to set the estimate range for each individual or all maintenance items on each estimate.

[0100] The insurance claim information generation device (100) can learn the decision of the insurer regarding the repair shop's estimate information, including items to be repaired in the future and cost estimates, for example, the technical opinion and cost estimate provided by the repair shop and the insurer's on-site opinion regarding it.

[0101] In one embodiment, the insurance claim information generation device (100) can evaluate an estimate generated in a standard form based on the decision learning results of the insurance claim system (300). Specifically, the insurance claim information generation device (100) can determine whether the maintenance items and cost estimates on the generated estimate fall within the cost estimate range set by the insurance company decision learning unit (330).

[0102] The insurance claim information generation device (100) can update the estimate based on the above judgment result.

[0103] For example, if the above judgment result exceeds the set cost estimate range, the cost estimate for the maintenance item on the estimate sheet may be updated to a predetermined value within the said cost estimate range. For example, the predetermined value within the cost estimate range may be determined as the maximum value within the said range to reduce the time required for the repair shop and the insurance company to reach an agreement on the cost estimate.

[0104] On the other hand, if the above cost estimate range is not exceeded, the insurance claim information generating device (100) may determine that the cost estimate for the corresponding maintenance item is subject to approval.

[0105] The maintenance estimate approval unit (350) can generate approval information for the entire estimate (or insurance claim information) when individual / total cost estimates on the estimate are determined to be subject to approval.

[0106] In one embodiment, the maintenance estimate approval unit (350) provides the generated approval information to the insurance company system (300), so that the insurance claim information received by the insurance company system (300) can be processed for approval.

[0107] In another embodiment, the maintenance estimate approval unit (350) may transmit to the insurance company system (300) if approval information for some estimate elements of the entire estimate has not been generated. If the insurance company system (300) has approved some estimate elements but has not received approval information for at least some of the remaining estimate elements, the insurance claim information may ultimately be rejected.

[0108] In another embodiment, even if insurance claim information is not provided to the insurance company system (300), the insurance claim information generating device (100) may evaluate the generated standard form estimate and make a decision to approve or reject the insurance claim information. That is, the insurance claim information generating device (100) may generate the decision content that the insurance company system (300) is expected to decide and provide it to the repair shop system (200).

[0109] FIG. 4 is a flowchart of an AI-based insurance claim information generation method according to another embodiment of the present invention.

[0110] Each step of FIG. 4 is performed by an AI-based insurance claim information generation device (100), and specifically, each step is executed as the processor (101) of the AI-based insurance claim information generation device (100) performs operations on each component of the software (105).

[0111] Referring to FIG. 4, the AI ​​insurance claim information generation device (100) can receive repair estimate information from the repair shop system (200) (S10).

[0112] The AI ​​insurance claim information generation device (100) can extract and embed text on the estimate information. Additionally, the AI ​​insurance claim information generation device (100) can identify maintenance items and cost estimates on the estimate information based on the embedding results.

[0113] Specifically, the AI ​​insurance claim information generation device (100) can identify the text of the maintenance item and the technical opinion report regarding the above estimate information to generate a first vector value for the vehicle to be repaired and generate a second vector value for the numerical value of the cost estimate for the above maintenance item. The numerical value of the cost estimate may be the amount of maintenance service, so-called labor cost.

[0114] In one embodiment, the AI ​​insurance claim information generation device (100) may match the first vector value and the second vector value.

[0115] The insurance claim information generating device (100) can classify a predetermined maintenance item into a first sub-maintenance item and a second sub-maintenance item based on the matching result.

[0116] In other words, each maintenance item may reflect differentiated entries in the technical opinion report, and the vector values ​​of the technical opinion reports matched by each maintenance item may differ. Among these, similar maintenance items or comprehensive maintenance items may have their vector values ​​for the technical opinion reports grouped into a similar range.

[0117] The insurance claim information generating device (100) can classify maintenance items having vector values ​​that have differences within a preset range based on matching results. That is, a comprehensive maintenance item may include multiple sub-maintenance items, and the sub-maintenance items may have component linkages with one another. In this case, the technical opinion report for each component or sub-maintenance item will have different vector values, and the insurance claim information generating device (100) can classify sub-maintenance items by vector value. However, among the various maintenance items clustered by vector value, the vector value of the estimate element belonging to the comprehensive maintenance item may have similarity that is grouped within a predetermined range.

[0118] You can also predict the likelihood of approval for the cost estimate by comparing it with the insurance cost range for each maintenance item.

[0119] Based on the above prediction results, maintenance item X outside the insurance coverage and maintenance item Y that falls within the insurance coverage but has an excessive cost estimate can be extracted by the approval prediction model. The insurance claim information generation device (100) can probabilistically calculate the probability that maintenance item X, maintenance item Y, and the entire estimate will be rejected.

[0120] The insurance claim information generating device (100) can generate insurance claim information based on the above prediction results (S40). The insurance claim information generating device (100) can remove maintenance items that are not covered by insurance on the estimate, or change cost estimates that are judged to be overcharged to values ​​within a threshold range pre-set for maintenance items. The insurance claim information generating device (100) can update the standard form estimate by reflecting such removal or change, and generate insurance claim information based on this.

[0121] In one embodiment, the AI ​​insurance claim information generation device (100) classifies the maintenance item into a first sub-maintenance item and a second sub-maintenance item when the maintenance item does not fall within the insurance coverage, and can determine whether either of the classified first and second sub-maintenance items falls within the insurance coverage.

[0122] Additionally, if the AI ​​insurance claim information generation device (100) determines that any one of the above sub-maintenance items is within the insurance coverage, it may change the maintenance item to a sub-maintenance item within the insurance coverage.

[0123] Meanwhile, if the AI ​​insurance claim information generation device (100) compares the cost estimate with the pre-learned approved cost estimate range for each maintenance item and the result exceeds a preset threshold range, it may send a maintenance estimate request message to another maintenance shop within a preset distance range from the maintenance shop based on the location information of the maintenance shop.

[0124] The AI ​​insurance claim information generation device (100) can process the approval of the claim on the insurance claim information upon receiving a confirmation and maintenance progress message from the repair shop system (200) regarding the generated insurance claim information (S50).

[0125] In one embodiment, the insurance claim information generating device (100) can send insurance claim information including updated information including changed repair items and cost estimates to the repair shop system (200).

[0126] When a standard form estimate is updated by the insurance claim information generation device (100), the repair shop system (200) needs to determine whether to proceed with the repairs in accordance with the estimate, as this results in the exclusion of some repair items or the modification of cost estimates for repair items. Therefore, when insurance claim information including the updated estimate is sent, the repair shop system (200) makes a decision on whether to proceed with the repair work and sends a message regarding this to the insurance claim information generation device (100).

[0127] The insurance claim information generation device (100) can process the approval of the repair estimate upon receiving the estimate confirmation and repair progress message.

[0128] In another embodiment, the insurance claim information generation device (100) may transmit an estimate confirmation and maintenance progress message to the insurance company system (300), and upon approval of the insurance claim information from the insurance company system (300), generate an approval message and send it to the repair shop system (200).

[0129] FIG. 5 is a flowchart of a method for generating personal information-separated training data, which is referenced in some embodiments of the present invention.

[0130] As described above in the explanation of step (S30) of Fig. 4, for learning the decision-making of an insurance company, it is necessary to learn information regarding the coverage of each insurance policyholder's insurance product, the type of insured vehicle, symptoms of failure, repair items and cost estimates, technical opinion, the insurance company's inspection opinion, and whether final approval or rejection is given. In this process, the insurance policyholder (vehicle owner) and vehicle number may be learned together, but since the use of such information is restricted by the Personal Information Protection Act, it needs to be refined in the training data.

[0131] Referring to FIG. 5, the insurance claim information generating device (100) can identify personal information, such as information on a subscriber who has signed up for an insurance product or the owner of an insured vehicle, on insurance claim information claimed to the insurance company system (300) (S501).

[0132] For example, an insurance claim information generation device (100) can classify information corresponding to an estimate element and personal information by filtering vector values ​​generated by embedding all information on the insurance claim information.

[0133] The insurance claim information generating device (100) can encrypt identified personal information (S502). For example, the insurance claim information generating device (100) can generate an identifier in which personal information is encrypted by combining an identifier for at least one of the types of insurance products and insured vehicle information with an encrypted value generated by encrypting personal information. For example, if the identifier for vehicle type ABC is 'a' and the insured Park is encrypted as 1023xx, then a1023xx can be generated as the identifier in which personal information is encrypted.

[0134] The insurance claim information generation device (100) can delete personal information identified in the insurance claim information and update the insurance claim information by reflecting the identifier in which the generated personal information is encrypted (S503).

[0135] The insurance claim information generation device (100) can learn the decision-making of an insurer based on the updated insurance claim information (S504). Through this, when utilizing the learning results in the future, the type of insured vehicle or insurance product can be identified or extracted and utilized based on some characters of the identifier in which personal information is encrypted.

[0136] Although a method of learning excluding personal information has been described regarding the previously subscribed insurance mentioned above, the embodiments of the present invention are not limited thereto.

[0137] According to another embodiment of the present invention, the insurance claim information generation device (100) can encrypt personal information during the insurance certificate generation process when signing up for insurance, and then block the generated insurance certificate.

[0138] Accordingly, the insurance claim information generating device (100) may subsequently receive quotation information with personal information deleted before step (S10) of FIG. 4, generate insurance claim information with personal information excluded in step S40), and process the insurance claim approval in step (S50). Based on this, the insurance claim information generating device (100) or the insurance company system (300) may control the insurance company system (300) to reflect the approval processing result in the blocked insurance certificate and automatically pay the insurance premium.

[0139] Up to now, an AI-based insurance claim information generation device and method according to an embodiment of the present invention have been described. The insurance claim information generation device described above may also be referred to as an AI-based insurance design device in that, according to another embodiment of the present invention, it learns AI-based generated insurance claim information and performs a design function for an insurance product / service.

[0140] Hereinafter, an AI-based insurance design device and method will be described with reference to FIGS. 6 to 8.

[0141] FIG. 6 is a block diagram of an AI-based insurance design device according to another embodiment of the present invention. Among the components of the insurance design device (600) of FIG. 6 and the descriptions thereof, any content that overlaps with the insurance claim information generation method (100) is omitted.

[0142] Referring to FIG. 6, an insurance design method can be implemented by the processor (601) of the AI-based insurance design device (600) executing each operation. Specifically, the insurance design method can be performed by the processor (601) of the insurance design device (600) executing each operation on the insurance design software (605).

[0143] Additionally, the storage (604) may store insurance product information (606). The insurance product information (606) may include information regarding an insurance product generated by the insurance design device (600).

[0144] According to an embodiment of the present invention, the AI-based insurance design device (600) may exist as a single integrated device rather than a separate device separated from the AI-based insurance claim information generation device (100).

[0145] FIG. 7 is a conceptual diagram of an insurance design method referenced in some embodiments of the present invention, and FIG. 8 is a flowchart of the insurance design method.

[0146] Referring to FIG. 7, information on the vehicle type subject to the insurance claim (701), information on the type of parts applied to each repair item (702), such as reused parts, used parts, and new parts, and information on linked maintenance items (703) that are subject to repair at high frequency simultaneously or within a specified period in a single maintenance item, influence the price calculation of the insurance product. The insurance design device (600) can collect information (701, 702, 703) from insurance claim information for which the payment of insurance premiums has been finally approved, and can automatically design an insurance product based on this. The insurance design device (600) can also collect information on the loss ratio generated by the actual payment of insurance premiums and automatically design an insurance product based on this.

[0147] For example, the insurance design device (600) can learn the subscribed insurance product, the coverage of the insurance product and the actual approved maintenance items, the list of parts used for each maintenance item, the cost estimate and loss ratio for each maintenance item.

[0148] The insurance design device (600) can classify the types of maintenance items and parts applied within the common insurance scope of the insurance product and design the insurance product based on this. The types of common maintenance items and parts actually approved according to the common insurance scope are referred to as common elements. On the other hand, approved maintenance items outside the common insurance scope, types of parts applied to maintenance outside the common insurance scope, or specific parts with low frequency among the parts applied to maintenance items within the common insurance scope are referred to as specific elements.

[0149] Using the learned results above, when specific elements, that is, insurance coverage other than common elements, are added, the insurance premium can be predicted based on the addition of maintenance items other than common elements and the application of specific parts types. In particular, the insurance design device (600) can learn the correlation between specific elements and the loss ratio and generate an insurance premium prediction model based on this.

[0150] For example, when designing an insurance product, the insurance design device (600) can calculate the insurance premium by considering the predicted loss ratio using a prediction model.

[0151] The loss ratio is determined by the ratio of the premiums paid for insurance claims to the amount remaining after deducting out-of-pocket expenses from the premiums paid by the policyholder to the insurer until the insurance claim is processed. Here, the premiums paid refer to expenses set to be paid according to the scope of insurance, such as repair costs and medical expenses, and the premiums paid refer to expenses paid up to the time of the insurance claim among the expenses agreed upon by the customer to pay to the insurer at the time of enrollment.

[0152] According to one embodiment, the insurance design device (600) can calculate the insurance premium by considering at least one of the loss ratio by insurance coverage, the loss ratio by vehicle type, and the loss ratio by part.

[0153] Through this, when designing insurance products, risk analysis is performed in terms of the loss ratio, and based on the loss ratio, the co-payment ratio can be predicted and determined whether to add it to the premium at the time of subscription.

[0154] An insurance service (700) designed in Fig. 7 is illustrated as an example.

[0155] In one embodiment, various special conditions may be reflected in the insurance product to be designed for loss ratio management. In particular, the insurance design device (600) can predict the loss ratio according to the special conditions and determine the insurance premium based on the predicted loss ratio.

[0156] In one embodiment, the insurance design device (600) can calculate the insurance premium by considering the density of repair shops by region, the population density by region, and the distance between the repair shop and the parts warehouse.

[0157] For example, as population density increases, labor costs may decrease, and as repair shop density increases, labor costs may decrease. On the other hand, as the distance between the repair shop and the parts warehouse increases, the price of parts rises, which affects the increase in overall labor costs. These factors are reflected in the final loss ratio and affect the insurance premium charged to the consumer. Accordingly, the insurance design device (600) can predict the loss ratio and calculate the insurance premium by considering the subscriber's address, the main operating area of ​​the vehicle, and information on frequently visited repair shops.

[0158] In this case, the insurance design device (600) can reflect a premium discount as a special clause in the insurance product when repairs are received at a repair shop at a predetermined location in a designated area.

[0159] In one embodiment, the insurance design device (600) may calculate the insurance premium by taking into account the type of vehicle.

[0160] For example, the insurance design device (600) can design the insurance product by determining the insurance period and premium range for the vehicle type based on information regarding the number of excess insurance claims determined for a specific maintenance item of the vehicle type when the insurance product is finally designed.

[0161] For example, in the case of a certain vehicle model, similar failure symptoms may frequently appear in models of a specific period. Additionally, there may be failure symptoms that frequently occur in vehicles with a driving distance exceeding a certain range. The insurance design device (600) may predict the loss ratio and calculate the insurance premium based on such vehicle information and / or driving information.

[0162] In one embodiment, the insurance design device (600) can determine the maintenance item Y of the second failure symptom as a linked maintenance item if, after the repair of the maintenance item X matched to the first failure symptom for a specific vehicle model is completed and a predetermined period has elapsed, the second failure symptom occurs at a frequency greater than a preset standard.

[0163] The insurance design device (600) may probabilistically predict the likelihood of a linked maintenance item occurring and reflect this in the premium increase calculation.

[0164] The insurance design device (600) may design an insurance product such that linked maintenance items are included in the coverage of the insurance product when subscribing to the insurance product.

[0165] In another embodiment, the insurance design device (600) may reflect a special clause for a premium discount in the insurance product when using a reused part among the parts.

[0166] In addition, the insurance design device (600) may design an insurance product by reflecting various options such as the insurance coverage period and the proportion of self-payment.

[0167] Each step of FIG. 8 is performed by an insurance design device (600), and specifically, each step is executed as the processor (601) of the insurance design device (600) performs operations on the software (605).

[0168] Referring to FIG. 8, the insurance design device (600) can collect insurance claim information for which premium payment has been approved (S801).

[0169] The insurance design device (600) can learn by matching insurance claim information, the insurer's actual opinion regarding it, the paid premium, and the final loss ratio.

[0170] The insurance design device (600) can classify common elements and specific elements among the collected insurance claim information based on the learned results (S802).

[0171] Common elements refer to insurance coverage exceeding a preset standard for insurance products in the collected insurance claim information, maintenance items approved exceeding a preset standard, and parts applied exceeding a preset standard within each maintenance item.

[0172] In one embodiment, the common element can be extracted for insurance claim information with premiums paid below a preset loss ratio.

[0173] Specific elements refer to maintenance items recognized as covered below the aforementioned preset standards, and parts applied below the preset standards.

[0174] In one embodiment, when extracting specific elements, insurance claim information for which premiums were paid below a preset loss ratio may be excluded. That is, if the loss ratio is low, the insurance design device (600) may not classify maintenance items or parts that were recognized as being below a preset standard as specific elements.

[0175] The insurance design device (600) can learn the correlation between specific factors and the loss ratio (S803). Based on this, the insurance design device (600) can generate an insurance premium prediction model.

[0176] The insurance design device (600) can design an insurance product using an insurance premium prediction model (S804).

[0177] The insurance design device (600) can design a preliminary insurance product composed only of common elements, and can also design a preliminary insurance product by adding at least one specific element among insurance coverage, maintenance items, and applicable parts to the common elements.

[0178] The insurance design device (600) can design a final insurance product by calculating the premium for each designed preliminary insurance product. At this time, the insurance design device (600) can calculate the premium by predicting the loss ratio based on additionally reflected specific factors using an insurance premium prediction model.

[0179] In one embodiment, the insurance design device (600) designs an insurance product that reflects the calculated premium, but may also design a final insurance product by reflecting special conditions that are pre-matched for each specific element (S805).

[0180] For example, when the insurance design device (600) is in the final design of the insurance product, the insurance design device (600) can create a special clause that adjusts the insurance premium when the use of the reusable part is agreed upon in a situation where repair of a part that is a specific element is required, by reflecting the condition regarding whether the reusable part is allowed in advance.

[0181] As another example, the insurance design device (600) can generate a special clause that adjusts the insurance premium when using a repair shop at a pre-registered location during the final design of the insurance product.

[0182] As another example, the insurance design device (600) can generate a special clause that adds the linked maintenance item to the insurance scope of the designed insurance product based on information regarding the extracted linked maintenance item when the insurance product is finally designed, and can also design an insurance product that reflects the generated special clause.

[0183] Specifically, when adding a linked maintenance item, the insurance service design device (100) may add a special clause that adjusts the self-payment ratio of the insurance product when an insurance claim is made by receiving maintenance for the linked maintenance item and the maintenance item within the insurance scope together, and reflect this in the generated special clause.

[0184] In another embodiment, the insurance design device (600) may design an insurance product by reflecting the driver's driving habits, etc.

[0185] The methods for determining and / or calculating a processor according to the embodiments of the present invention described so far with reference to the attached drawings may be performed by executing a computer program implemented in computer-readable code. The computer program may be transmitted from a first computing device to a second computing device via a network such as the Internet and installed on the second computing device, thereby being used on the second computing device. The first computing device and the second computing device include all fixed computing devices such as server devices and desktop PCs, and mobile computing devices such as laptops, smartphones, and tablet PCs.

[0186] Although embodiments of the present invention have been described above with reference to the attached drawings, those skilled in the art will understand that the present invention may be implemented in other specific forms without changing its technical concept or essential features. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive.

Claims

1. A method performed by an AI-based insurance claim information generation device, Step of receiving estimate information from the repair shop; A step of extracting character elements from quotation information and embedding the extracted character elements; Among the above embedding results, a step of extracting standard form estimation elements from a pre-established vector DB based on vector values ​​generated through embedding for estimation elements; A step of generating a standard form quotation reflecting the standard form quotation elements based on the received quotation information; A step of predicting the likelihood of approval of the generated standard form quotation using a pre-established approval prediction model; and A method comprising the step of generating insurance claim information based on the above prediction results, AI-based insurance claim information generation method.

2. In claim 1, the step of predicting the possibility of approval of the generated standard form quotation is, A first step of determining whether the above-mentioned quotation element is within the coverage of the subscribed insurance product; A second step of determining whether the cost estimate corresponding to the estimate element among the extracted character elements is a value within a preset threshold range; and A method comprising the step of predicting the possibility of approval of the generated standard form quotation based on at least one of the results of the first judgment and the second judgment. AI-based insurance claim information generation method.

3. In Paragraph 2, the step of generating the insurance claim information is, A step of removing an estimate element that is not within the insurance scope based on the first judgment result, or setting a cost estimate outside the threshold range to be changed to a value within the preset threshold range based on the second judgment result; and A step including updating the standard form quotation to reflect the above-set result, AI-based insurance claim information generation method.

4. In claim 1, the estimation element includes at least one element among maintenance items, cost estimates, and technical opinions, and The step of extracting the above-mentioned standard form quotation elements is, Based on a vector value generated after embedding for at least one element, the method includes the step of extracting a standard form estimation element having a vector value within a similar range on the vector DB. The estimation elements of the above standard form are pre-matched with vectors within the above similar range, and The step of generating the above standard form quotation is, A step comprising converting terms for quotation elements in quotation information into quotation elements of an extracted standard form, AI-based insurance claim information generation method.

5. In Paragraph 3, if, as a result of the first judgment above, a specified maintenance item among the estimation elements is not within the scope of insurance, The step of setting the above change is, A step of classifying the above-mentioned predetermined maintenance items into a first sub-maintenance item and a second sub-maintenance item; A step of determining whether any one of the first sub-maintenance item and the second sub-maintenance item classified above falls under the insurance coverage; and If it is determined that any one of the above sub-maintenance items falls within the scope of insurance, the method includes the step of setting the predetermined maintenance item to be changed to any one of the above sub-maintenance items. AI-based insurance claim information generation method.

6. In claim 5, the step of extracting character elements from the quotation information and embedding the extracted character elements is: Among the above-mentioned estimation elements, a step of embedding maintenance items and technical opinion reports respectively; and It includes the step of matching a first vector value embedded for the above maintenance item with a second vector value embedded for the above technical opinion report, and The step of classifying the above-mentioned predetermined maintenance items into first sub-maintenance items and second sub-maintenance items is Based on the above matching results, a step of classifying maintenance items having vector values ​​with differences within a preset range; and A step comprising identifying a first sub-maintenance item and a second sub-maintenance item on the above-classified maintenance item, AI-based insurance claim information generation method.

7. In Clause 6, the step of determining whether either of the classified first sub-maintenance item and second sub-maintenance item falls within the insurance coverage is: A step of identifying information on maintenance items for which premium payment has been approved for an insurance product, using an artificial intelligence model built by learning the insurer's decision-making regarding the received estimate information; and A step comprising determining a maintenance item that falls under insurance coverage among a first sub-maintenance item and a second sub-maintenance item based on information regarding the above-mentioned identified approved maintenance item, AI-based insurance claim information generation method.

8. One or more processors; Network interface for receiving quotation information; Memory for loading a computer program executed by the above processor; and It includes storage for storing the above computer program, The above computer program is, Operation to receive repair shop estimate information; An operation to extract character elements from the above quotation information and embed the extracted character elements; An operation to extract standard form estimation elements from a pre-established vector DB based on vector values ​​generated through embedding for estimation elements among the above embedding results; An operation to generate a standard form quotation reflecting the standard form quotation elements based on the received quotation information; An operation to predict the likelihood of approval of the generated standard form quotation using a pre-established approval prediction model; and Based on the above prediction results, including an operation to generate insurance claim information, AI-based insurance claim information generation device.

Images