Insurance claim settlement processing method and device, computer equipment and storage medium

Abstract

The embodiment of the invention belongs to the field of artificial intelligence, and relates to an insurance claim settlement processing method, which comprises the following steps: obtaining case information of an insurance claim settlement case, the case information comprising case basic information and case processing information; inputting the case information into a decision model to obtain a plurality of candidate decisions and reward values corresponding to the candidate decisions; selecting the candidate decision corresponding to the maximum reward value as a target decision; and calling at least one agent according to the target decision, so as to process the case information through the at least one agent. The invention further provides an insurance claim settlement processing device, computer equipment and a storage medium. The accuracy of the insurance claim settlement system is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an insurance claims processing method, apparatus, computer equipment and storage medium. Background Art

[0002] With the development of society and the change of concepts, the demand for insurance claims is also growing. The existing insurance claims process is usually cumbersome and complicated, involving a large number of manual operations, which is not only inefficient but also prone to errors. In recent years, the development of artificial intelligence technology has provided new possibilities for the intelligence of insurance claims. However, the current insurance claims systems based on artificial intelligence all use fixed linear task flows, calling one or more nodes through manually defined criteria and thinking chains to perform insurance claims tasks. This fixed task flow is difficult to meet the requirements of changing actual scenarios, because it is difficult for staff to fully and correctly describe an application scenario, resulting in insufficient system flexibility and inability to effectively handle unconventional insurance claims cases. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide an insurance claim processing method, apparatus, computer equipment and storage medium to solve the problem of poor flexibility of the insurance claim system.

[0004] In order to solve the above technical problems, the present application embodiment provides an insurance claim processing method, which adopts the following technical solutions:

[0005] Obtaining case information of insurance claim cases, wherein the case information includes case basic information and case processing information;

[0006] Inputting the case information into a decision model to obtain a plurality of candidate decisions and their corresponding reward values;

[0007] Select the candidate decision corresponding to the maximum reward value as the target decision;

[0008] At least one intelligent agent is called according to the target decision to process the case information through the at least one intelligent agent.

[0009] In order to solve the above technical problems, the embodiment of the present application further provides an insurance claim processing device, which adopts the following technical solution:

[0010] An information acquisition module is used to acquire case information of insurance claim cases, wherein the case information includes case basic information and case processing information;

[0011] An information input module, used to input the case information into a decision model to obtain a plurality of candidate decisions and their corresponding reward values;

[0012] A decision selection module is used to select the candidate decision corresponding to the maximum reward value as the target decision;

[0013] The case processing module is used to call at least one intelligent agent according to the target decision to process the case information through the at least one intelligent agent.

[0014] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0015] Obtaining case information of insurance claim cases, wherein the case information includes case basic information and case processing information;

[0016] Inputting the case information into a decision model to obtain a plurality of candidate decisions and their corresponding reward values;

[0017] Select the candidate decision corresponding to the maximum reward value as the target decision;

[0018] At least one intelligent agent is called according to the target decision to process the case information through the at least one intelligent agent.

[0019] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0020] Obtaining case information of insurance claim cases, wherein the case information includes case basic information and case processing information;

[0021] Inputting the case information into a decision model to obtain a plurality of candidate decisions and their corresponding reward values;

[0022] Select the candidate decision corresponding to the maximum reward value as the target decision;

[0023] At least one intelligent agent is called according to the target decision to process the case information through the at least one intelligent agent.

[0024] Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects: obtaining case information of insurance claim cases, including basic case information and case processing information, where the case processing information is the processing information of the case by the previous intelligent agent, thereby utilizing the capabilities or conclusions of other intelligent agents; the decision model is obtained based on reinforcement learning training, and can generate multiple candidate decisions and their corresponding reward values ​​according to the case information, and select the candidate decision corresponding to the maximum reward value as the target decision, thereby ensuring that the selected decision is based on the optimal return, thereby improving the accuracy and efficiency of claims processing; the target decision is a processing flow dynamically generated according to the actual situation of the case, and according to the target decision, at least one intelligent agent can be flexibly called to process the case information, which can effectively handle various insurance claim cases and improve the flexibility of insurance claims processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0026] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0027] Figure 2 is a flow chart of an embodiment of an insurance claim processing method according to the present application;

[0028] Figure 3 is a structural schematic diagram of an embodiment of an insurance claim processing device according to the present application;

[0029] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0031] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0032] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0033] like Figure 1As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0034] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0035] The terminal devices 101, 102, and 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0036] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0037] It should be noted that the insurance claim processing method provided in the embodiment of the present application is generally executed by a server, and accordingly, the insurance claim processing device is generally set in the server.

[0038] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0039] Continue to refer Figure 2 , shows a flow chart of an embodiment of an insurance claim processing method according to the present application. The insurance claim processing method comprises the following steps:

[0040] Step S201, obtaining case information of an insurance claim case, the case information includes basic case information and case processing information.

[0041] In this embodiment, the insurance claim processing method is executed on an electronic device (eg Figure 1 The server shown in the figure) can communicate with the terminal device through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 3G / 4G / 5G connection, Wi-Fi connection, Bluetooth connection, Wi MAX connection, Zigbee connection, UWB (Ultra Wi-Fi Band) connection, and other wireless connection methods currently known or to be developed in the future.

[0042] Specifically, the insurance claims processing system includes a decision model and multiple agents, where the decision model can also be regarded as an agent, which is the core control mechanism of each agent. The agent is based on a pre-trained model, or is built based on the large language model (LLM) and the visual language multimodal large model (LVLM), which is a customized expert model.

[0043] In insurance claims processing, case information includes two main parts:

[0044] Basic case information: This refers to the basic information of the case, such as the claimant’s information, the time and location of the accident, the accident description and other report information.

[0045] Case processing information: This refers to the information on the previous agent's processing of insurance claims, including the agent's processing results, relevant data during the processing, etc. In insurance claims processing, the decision model may have to make multiple decisions. When it is not the first decision, the decision model needs to consider the case processing information generated by the previous agent, so as to utilize the capabilities or conclusions of other agents.

[0046] Step S202: input case information into the decision model to obtain multiple candidate decisions and their corresponding reward values.

[0047] Specifically, case information is input into a decision model trained by reinforcement learning to obtain multiple candidate decisions. Each candidate decision has a corresponding reward value, and the reward value reflects the expected return or effect of the decision in a numerical value.

[0048] The reinforcement model in this application can be a DQN (Deep Q-Learning) model. The DQN model is a deep reinforcement learning model that can use a deep Q network DQN to optimize the long-term cumulative goal of sequential decision-making problems. Reinforcement learning is an artificial intelligence method in which an agent (the decision model in this application can be regarded as an agent) takes an action based on a certain strategy and obtains a reward for the state, and then optimizes the strategy through the reward obtained. The strategy refers to an action that should be taken under a specific state to maximize the expected cumulative reward. The DQN model uses a neural network to fit the strategy, inputs the state (such as accident type, damage level, reporting time, etc.) into the neural network, and the neural network outputs the Q value (expected cumulative reward) corresponding to each action. The action corresponding to the maximum Q value (such as querying historical cases, risk assessment, etc.) is the action that the DQN model believes should be selected. The larger the reward value, the higher the reward, and the more likely it is to take the corresponding candidate decision.

[0049] Reinforcement learning enables the model to learn from a large amount of historical data, optimize the decision-making process, and provide better claims settlement options. This application uses reinforcement learning to achieve competition and selection among agents and realize nonlinear scheduling of agents.

[0050] Step S203: Select the candidate decision corresponding to the maximum reward value as the target decision.

[0051] Specifically, among all the generated candidate decisions, the one with the highest reward value is selected as the target decision. The highest reward value means that the decision has the best performance in historical data, has the highest probability of success and the best return, ensuring that the system selects the optimal decision and maximizes the accuracy and efficiency of claims processing.

[0052] Step S204, calling at least one intelligent agent according to the target decision to process the case information through at least one intelligent agent.

[0053] Specifically, the decision model also has an agent scheduling function. The target decision records which agents need to be called, and the decision model will call at least one agent to process case information based on the target decision.

[0054] An intelligent agent is an expert model that can include a pre-trained model, an action instruction knowledge base, and a professional knowledge base, which can automatically process case information and generate intelligent agent processing results.

[0055] Taking the auto insurance claims scenario as an example, the auto insurance claims system contains the following types of agents:

[0056] Agent A: Vehicle damage assessment expert; responsible for conducting damage assessment based on the vehicle damage condition, and using image recognition technology and semantic understanding technology to quickly and accurately determine the extent of damage.

[0057] Agent B: Insurance clause expert; analyzes the contents of the insurance contract, generates and reviews the claims plan based on the insurance clauses and claims regulations.

[0058] Agent C: Online communication robot; communicates with car owners online, answers claims-related questions, and provides professional advice.

[0059] Agent D: An agent that searches for similar historical cases. It analyzes the historical claims case database to find cases similar to the current case and provides reference information, such as treatment plans, claim amounts, etc.

[0060] Agent E: Risk assessment expert; assesses the risk level of the accident based on accident scene photos and report forms, and provides risk warnings and suggestions for subsequent claims processing.

[0061] The decisions of the decision-making model are generated based on the actual situation of insurance claims cases, and the decision-making process will be flexibly adjusted according to the actual situation. This enables the system to effectively deal with various complex and unconventional insurance claims cases and improves the flexibility of processing.

[0062] In this embodiment, case information of insurance claim cases is obtained, including basic case information and case processing information. The case processing information is the processing information of the case by the previous intelligent agent, thereby utilizing the capabilities or conclusions of other intelligent agents; the decision model is obtained based on reinforcement learning training, and can generate multiple candidate decisions and their corresponding reward values ​​according to the case information, and select the candidate decision corresponding to the maximum reward value as the target decision, thereby ensuring that the selected decision is based on the optimal return, thereby improving the accuracy and efficiency of claims processing; the target decision is a processing flow dynamically generated according to the actual situation of the case, and according to the target decision, at least one intelligent agent can be flexibly called to process the case information, which can effectively handle various insurance claims cases and improve the flexibility of insurance claims processing.

[0063] Furthermore, after the above-mentioned step of obtaining case information of insurance claim cases, it can also include: obtaining the intelligent agent processing result in the case processing information, the intelligent agent processing result is generated by a pre-initiated intelligent agent; querying whether there is a preset decision corresponding to the intelligent agent processing result; when a preset decision exists, taking the preset decision as the target decision; when no preset decision exists, inputting the case information into the decision model to obtain multiple candidate decisions and their corresponding reward values.

[0064] Specifically, the decision model can obtain the previous agent's agent processing results on the case from the case processing information, so as to use the existing intermediate processing results to provide a reference for subsequent decisions.

[0065] The decision model queries whether there is a preset decision corresponding to the obtained agent processing result. For some agent processing results, the system pre-stores some standard decisions. If a corresponding preset decision is found, it is directly used as the target decision to reduce repeated calculations and improve processing efficiency.

[0066] For example, in a car insurance claim case, the above-mentioned agent A has determined that the accident type is a minor collision or scratch. For such agent processing results, a standard decision is pre-set as "select agent D to query historical similar cases and provide reference processing solutions", and the decision model can directly use this preset decision as the current target decision. Alternatively, if agent A determines that there is more serious damage or personal injury based on the photos of the accident scene, the risk assessment results of agent E will be given priority, and it will provide risk warnings and suggestions to ensure that the subsequent claims process is safe and smooth.

[0067] If there is no corresponding preset decision, the case information is input into the decision model to generate multiple candidate decisions and their corresponding reward values.

[0068] In this embodiment, the agent processing results generated by the previous agent are obtained. When there is a corresponding preset decision, the standardized decision is directly adopted to avoid repeated calculations and improve the processing speed; otherwise, a new candidate decision is generated by the decision model to ensure the flexibility and adaptability of the processing process.

[0069] Furthermore, the above step S202 may include: inputting the case information into a pre-trained model in the decision model to extract features of the case information through the pre-trained model to obtain a feature vector; inputting the feature vector into a reinforcement learning model in the decision model to obtain multiple candidate decisions and their corresponding reward values.

[0070] Specifically, the decision model includes a pre-trained model and a reinforcement learning model. The decision model first inputs the case information into the pre-trained model. The pre-trained model can be a large language model (LLM) and a large visual language multimodal model (LVLM), etc. It has the ability to handle complex tasks, understand and generate natural language, and analyze the combination of images and text. It is suitable for scenarios such as insurance claims that require the integration of multiple information. The pre-trained model can parse the case information and extract features to obtain a feature vector. The feature vector is a numerical representation of the case information, which contains the main features of the case information. The powerful representation ability of the pre-trained model is conducive to improving the accuracy of decision-making.

[0071] The feature vector is input into the reinforcement learning model in the decision model. The reinforcement learning model understands the current state based on the input feature vector and generates multiple candidate decisions and their corresponding reward values; each candidate decision has an expected reward value, which represents its expected effect or return.

[0072] In this embodiment, the decision model includes a pre-trained model and a reinforcement learning model; the case information is input into the pre-trained model for feature extraction to generate a feature vector, which can better capture the key points of the case information and ensure that the decision-making process is based on high-quality feature representation; then the reinforcement learning model is input to generate multiple candidate decisions and their corresponding reward values, and the optimal decision is generated based on high-quality features, thereby improving the efficiency and accuracy of handling insurance claims cases.

[0073] Furthermore, the above-mentioned step of processing case information through at least one intelligent agent may include: for each intelligent agent called, inputting the case information into a pre-trained model in the intelligent agent to obtain the task requirements corresponding to the case information; generating a task processing framework corresponding to the task requirements through the pre-trained model; for each task in the task processing framework, retrieving action instructions from the action instruction knowledge base of the intelligent agent to perform the task and obtain instruction execution results; based on the obtained instruction execution results and the professional knowledge base in the intelligent agent, generating the intelligent agent processing results through the pre-trained model.

[0074] Specifically, the agent contains pre-trained models, action instruction knowledge bases, and professional knowledge bases. The pre-trained models in each agent (including decision models) are customized and fine-tuned for the role and task of the agent to enhance its performance in a specific field. The action instruction knowledge base contains instructions for all actions and operations that the agent can perform. For example, the action instructions of agent A mentioned above may include "identify damaged parts", "estimate repair costs", "generate damage assessment reports", etc. The action instruction knowledge base also stores the rules and conditions for executing each action, such as under what circumstances to choose a specific action. The action instruction knowledge base can also provide strategies for how to efficiently execute these actions, including some optimization algorithms or operation guidelines to ensure the efficiency and accuracy of action execution. The professional knowledge base is an expert knowledge base customized for specific application scenarios and tasks, which contains in-depth knowledge and data in a specific field; for example, in the scenario of auto insurance claims, the professional knowledge base includes various professional knowledge in auto insurance claims, such as insurance terms, vehicle damage assessment standards, claims process, etc. The professional knowledge base can also include historical claims cases, common problems and solutions, etc., to provide reference for current tasks. The professional knowledge base can also include rules, laws and regulations, and industry norms in specific fields to ensure that tasks are performed in accordance with relevant standards.

[0075] The decision model will call at least one intelligent agent. For each called intelligent agent, the case information is first input into the pre-trained model. The pre-trained model will parse the case information and determine the corresponding task requirements. It will also generate a task processing framework based on the task requirements. It is a structured processing plan that includes the various steps required to perform the task. Each task in the task processing framework has corresponding action instructions, which are stored in the action instruction knowledge base. The intelligent agent executes these action instructions to perform specific tasks and obtains the instruction execution results. During the execution process, the intelligent agent will also refer to the professional knowledge and historical data in the professional knowledge base to ensure the accuracy and standardization of the task.

[0076] The pre-trained model will combine the command execution results and the information in the professional knowledge base to generate the final agent processing results. For example, when generating a damage report, the agent will refer to the vehicle damage assessment standards and historical case data in the professional knowledge base.

[0077] When the agent C mentioned above is called, it communicates with the customer through the online communication robot to answer questions, provide suggestions, and generate communication records or summaries based on the communication.

[0078] In this embodiment, the intelligent agent includes a pre-trained model, an action instruction knowledge base and a professional knowledge base; case information is input into the pre-trained model, task requirements are generated and a task processing framework is constructed, and then action instructions are retrieved from the action instruction knowledge base to execute the task, and the intelligent agent processing results are generated in combination with the professional knowledge base, ensuring that the intelligent agent can process case information systematically and effectively, and improve the professionalism, accuracy and processing efficiency of task execution.

[0079] Furthermore, the above step S204 may include: when the intelligent agent called according to the target decision is a decision model, generating a claim result of the insurance claim case according to the basic case information and case processing information; and sending the claim result to the terminal logged in by the target account.

[0080] Specifically, the decision model can make continuous decisions and executions. After executing a target decision, the decision model can select the next agent to perform the corresponding task based on the updated state. This process can be cyclical, and the execution results of the agent are continuously fed back to the decision model, and the decision model makes new decisions.

[0081] The agent called by the decision model according to the target decision can be the decision model itself. At this time, the final claim result is usually generated by integrating the basic case information and case processing information to ensure a comprehensive understanding of the insurance claim case. The claim result can be generated by the pre-trained model in the decision model. Then, the decision model sends the result to the terminal logged in by the target account to ensure that the customer or relevant personnel of the insurance company can receive the claim result in time and confirm or modify the claim result.

[0082] In this embodiment, when the target decision calls the decision model, the decision model generates the final claim result based on the basic case information and case processing information, and sends the claim result to the terminal logged in by the target account, thereby achieving accuracy and timeliness in the generation and delivery of the claim result.

[0083] Furthermore, when multiple agents are called according to the target decision and the agent processing results of each agent conflict, the insurance claim processing method may also include: obtaining the priority of each agent from the target decision, the priority being calculated based on the task importance score, historical accuracy score, resource consumption score and historical performance score of each agent; selecting the agent processing result of the highest priority agent as the execution result of the target decision.

[0084] Specifically, multiple agents may be called according to the target decision, and the agent processing results output by each agent may conflict or contradict each other. In this case, the priority of each agent is obtained from the target decision.

[0085] The priority of the agent is generated when the decision model generates candidate decisions. The priority is calculated based on the task importance score, historical accuracy score, resource consumption score and historical performance score of each agent. Among them, the task importance score includes time sensitivity (assessing the timeliness requirement for task completion, the more urgent the task, the higher the score), impact scope (assessing the impact of task results on the overall process, the greater the impact, the higher the score), risk level (assessing the impact of task results on system risk, the greater the risk, the higher the score), and user satisfaction (assessing the impact of task results on customer experience, the greater the impact, the higher the score). The historical accuracy score is a score determined based on the historical task success rate and accuracy of the agent. The resource consumption score assesses the computing resources and time required for the agent to perform the task, and the lower the consumption, the higher the score. The historical performance score assesses the performance of the agent on historical tasks, such as accuracy, efficiency, and the score given to the agent by insurance company staff. The scoring factors can be weighted and combined to calculate the total score of each agent, and the agents can be sorted according to the total score to determine the priority order.

[0086] When the results of various agents conflict, the agent processing result of the highest priority agent is selected as the execution result of the target decision.

[0087] In this embodiment, when the results of various intelligent agents conflict, the priority of each intelligent agent is obtained from the target decision. The priority is calculated based on the task importance score, historical accuracy score, resource consumption score and historical performance score, and the intelligent agent can be comprehensively evaluated. The result with the highest priority is selected to ensure the objectivity and reliability of the final decision and improve the quality of the decision.

[0088] Furthermore, when multiple agents are called according to the target decision and the agent processing results of each agent conflict, the insurance claim processing method may also include: obtaining the historical accuracy score and self-score of each agent; calculating the confidence score of each agent based on the historical accuracy score and self-score of each agent; and selecting the agent processing result of the agent with the highest confidence score as the execution result of the target decision.

[0089] Specifically, when multiple agents are called according to the target decision and the agent processing results of each agent conflict, the historical accuracy score and self-score of each agent can also be obtained. The historical accuracy score here is the same as the historical accuracy score that appears above. The self-score is the agent's self-evaluation of the agent processing results it outputs.

[0090] The confidence score of each agent can be obtained by weighted synthesis based on the historical accuracy score and self-score of each agent, which can accurately reflect the reliability of the agent. The agent processing result output by the agent with the highest confidence score is selected to ensure the reliability and accuracy of the decision.

[0091] For example, when the agents D and E mentioned above are called at the same time and their agent processing results conflict, if the confidence score of agent D is higher than that of agent E, even if the risk assessment result of agent E is more serious, it will focus on the reference information provided by agent D based on historical cases and give priority to the solution provided by agent D to reduce the probability of false risk alarms.

[0092] In this embodiment, when the results of each intelligent agent conflict, the historical accuracy score and self-score of each intelligent agent can also be obtained, and the confidence score can be calculated; the processing result of the intelligent agent with the highest confidence score is selected as the execution result, ensuring that the most reliable intelligent agent processing result is selected when there is a decision conflict, thereby improving the decision quality.

[0093] It should be emphasized that in order to further ensure the privacy and security of the above case information, the above case information can also be stored in a blockchain node.

[0094] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0095] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0096] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0097] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0098] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0099] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of an insurance claim processing device, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0100] like Figure 3 As shown, the insurance claim processing device 300 described in this embodiment includes: an information acquisition module 301, an information input module 302, a decision selection module 303 and a case processing module 304, wherein:

[0101] The information acquisition module 301 is used to acquire case information of insurance claim cases, and the case information includes basic case information and case processing information.

[0102] The information input module 302 is used to input case information into the decision model to obtain multiple candidate decisions and their corresponding reward values.

[0103] The decision selection module 303 is used to select the candidate decision corresponding to the maximum reward value as the target decision.

[0104] The case processing module 304 is used to call at least one intelligent agent according to the target decision to process the case information through the at least one intelligent agent.

[0105] In this embodiment, case information of insurance claim cases is obtained, including basic case information and case processing information. The case processing information is the processing information of the case by the previous intelligent agent, thereby utilizing the capabilities or conclusions of other intelligent agents; the decision model is obtained based on reinforcement learning training, and can generate multiple candidate decisions and their corresponding reward values ​​according to the case information, and select the candidate decision corresponding to the maximum reward value as the target decision, thereby ensuring that the selected decision is based on the optimal return, thereby improving the accuracy and efficiency of claims processing; the target decision is a processing flow dynamically generated according to the actual situation of the case, and according to the target decision, at least one intelligent agent can be flexibly called to process the case information, which can effectively handle various insurance claims cases and improve the flexibility of insurance claims processing.

[0106] In some optional implementations of this embodiment, the insurance claim processing device 300 may further include: a result acquisition module, a preset query module, and a preset determination module, wherein:

[0107] The result acquisition module is used to obtain the intelligent agent processing results in the case processing information, and the intelligent agent processing results are generated by the intelligent agent in advance.

[0108] The preset query module is used to query whether there is a preset decision corresponding to the intelligent agent processing result.

[0109] The preset determination module is used to take the preset decision as the target decision when there is a preset decision.

[0110] The information input module is also used to input case information into the decision model when there is no preset decision, so as to obtain multiple candidate decisions and their corresponding reward values.

[0111] In this embodiment, the agent processing results generated by the previous agent are obtained. When there is a corresponding preset decision, the standardized decision is directly adopted to avoid repeated calculations and improve the processing speed; otherwise, a new candidate decision is generated by the decision model to ensure the flexibility and adaptability of the processing process.

[0112] In some optional implementations of this embodiment, the information input module 302 may include: a feature extraction submodule and a vector input submodule, wherein:

[0113] The feature extraction submodule is used to input case information into the pre-trained model in the decision model so as to extract features of the case information through the pre-trained model to obtain a feature vector.

[0114] The vector input submodule is used to input the feature vector into the reinforcement learning model in the decision model to obtain multiple candidate decisions and their corresponding reward values.

[0115] In this embodiment, the decision model includes a pre-trained model and a reinforcement learning model; the case information is input into the pre-trained model for feature extraction to generate a feature vector, which can better capture the key points of the case information and ensure that the decision-making process is based on high-quality feature representation; then the reinforcement learning model is input to generate multiple candidate decisions and their corresponding reward values, and the optimal decision is generated based on high-quality features, thereby improving the efficiency and accuracy of handling insurance claims cases.

[0116] In some optional implementations of this embodiment, the case processing module 304 may include: an information input submodule, a framework generation submodule, a task execution submodule, and a result generation submodule, wherein:

[0117] The information input submodule is used to input the case information into the pre-trained model in each intelligent agent called to obtain the task requirements corresponding to the case information.

[0118] The framework generation submodule is used to generate the task processing framework corresponding to the task requirements through the pre-trained model.

[0119] The task execution submodule is used to retrieve action instructions from the action instruction knowledge base of the intelligent agent for each task in the task processing framework to execute the task and obtain the instruction execution result.

[0120] The result generation submodule is used to generate the agent processing results through the pre-training model based on the obtained instruction execution results and the professional knowledge base in the agent.

[0121] In this embodiment, the intelligent agent includes a pre-trained model, an action instruction knowledge base and a professional knowledge base; case information is input into the pre-trained model, task requirements are generated and a task processing framework is constructed, and then action instructions are retrieved from the action instruction knowledge base to execute the task, and the intelligent agent processing results are generated in combination with the professional knowledge base, ensuring that the intelligent agent can process case information systematically and effectively, and improve the professionalism, accuracy and processing efficiency of task execution.

[0122] In some optional implementations of this embodiment, the case processing module 304 may include: a claim generation submodule and a result sending submodule, wherein:

[0123] The claim generation submodule is used to generate the claim result of the insurance claim case based on the basic case information and case processing information when the intelligent agent called according to the target decision is a decision model.

[0124] The result sending submodule is used to send the claim result to the terminal logged in by the target account.

[0125] In this embodiment, when the target decision calls the decision model, the decision model generates the final claim result based on the basic case information and case processing information, and sends the claim result to the terminal logged in by the target account, thereby achieving accuracy and timeliness in the generation and delivery of the claim result.

[0126] In some optional implementations of this embodiment, the insurance claim processing device 300 may further include: a priority acquisition module and a first selection module, wherein:

[0127] The priority acquisition module is used to obtain the priority of each agent from the target decision. The priority is calculated based on the task importance score, historical accuracy score, resource consumption score and historical performance score of each agent.

[0128] The first selection module is used to select the agent processing result of the highest priority agent as the execution result of the target decision.

[0129] In this embodiment, when the results of various intelligent agents conflict, the priority of each intelligent agent is obtained from the target decision. The priority is calculated based on the task importance score, historical accuracy score, resource consumption score and historical performance score, and the intelligent agent can be comprehensively evaluated. The result with the highest priority is selected to ensure the objectivity and reliability of the final decision and improve the quality of the decision.

[0130] In some optional implementations of this embodiment, the insurance claim processing device 300 may further include: a score acquisition module, a confidence calculation module, and a second selection module, wherein:

[0131] The score acquisition module is used to obtain the historical accuracy score and self-score of each agent.

[0132] The confidence calculation module is used to calculate the confidence score of each agent based on its historical accuracy score and self-score.

[0133] The second selection module is used to select the agent processing result of the agent with the highest confidence score as the execution result of the target decision.

[0134] In this embodiment, when the results of each intelligent agent conflict, the historical accuracy score and self-score of each intelligent agent can also be obtained, and the confidence score can be calculated; the processing result of the intelligent agent with the highest confidence score is selected as the execution result, ensuring that the most reliable intelligent agent processing result is selected when there is a decision conflict, thereby improving the decision quality.

[0135] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0136] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.

[0137] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0138] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the insurance claim processing method, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0139] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the insurance claim processing method.

[0140] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0141] The computer device provided in this embodiment can execute the above-mentioned insurance claim processing method. Here, the insurance claim processing method can be the insurance claim processing method of each of the above-mentioned embodiments.

[0142] In this embodiment, case information of insurance claim cases is obtained, including basic case information and case processing information. The case processing information is the processing information of the case by the previous intelligent agent, thereby utilizing the capabilities or conclusions of other intelligent agents; the decision model is obtained based on reinforcement learning training, and can generate multiple candidate decisions and their corresponding reward values ​​according to the case information, and select the candidate decision corresponding to the maximum reward value as the target decision, thereby ensuring that the selected decision is based on the optimal return, thereby improving the accuracy and efficiency of claims processing; the target decision is a processing flow dynamically generated according to the actual situation of the case, and according to the target decision, at least one intelligent agent can be flexibly called to process the case information, which can effectively handle various insurance claims cases and improve the flexibility of insurance claims processing.

[0143] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the insurance claim processing method as described above.

[0144] In this embodiment, case information of insurance claim cases is obtained, including basic case information and case processing information. The case processing information is the processing information of the case by the previous intelligent agent, thereby utilizing the capabilities or conclusions of other intelligent agents; the decision model is obtained based on reinforcement learning training, and can generate multiple candidate decisions and their corresponding reward values ​​according to the case information, and select the candidate decision corresponding to the maximum reward value as the target decision, thereby ensuring that the selected decision is based on the optimal return, thereby improving the accuracy and efficiency of claims processing; the target decision is a processing flow dynamically generated according to the actual situation of the case, and according to the target decision, at least one intelligent agent can be flexibly called to process the case information, which can effectively handle various insurance claims cases and improve the flexibility of insurance claims processing.

[0145] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0146] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A method for handling insurance claims, It is characterized in that The steps include: Obtaining case information of insurance claim cases, wherein the case information includes case basic information and case processing information; Inputting the case information into a decision model to obtain a plurality of candidate decisions and their corresponding reward values; Select the candidate decision corresponding to the maximum reward value as the target decision; At least one intelligent agent is called according to the target decision to process the case information through the at least one intelligent agent.

2. The insurance claim processing method according to claim 1, It is characterized in that After the step of obtaining the case information of the insurance claim case, the method further includes: Obtaining an agent processing result in the case processing information, wherein the agent processing result is generated by a pre-made agent; Query whether there is a preset decision corresponding to the processing result of the intelligent agent; When the preset decision exists, taking the preset decision as the target decision; When the preset decision does not exist, the step of inputting the case information into the decision model to obtain multiple candidate decisions and their corresponding reward values ​​is performed.

3. The insurance claim processing method according to claim 1, It is characterized in that The step of inputting the case information into the decision model to obtain a plurality of candidate decisions and their corresponding reward values ​​comprises: Inputting the case information into a pre-trained model in a decision model, so as to extract features of the case information through the pre-trained model to obtain a feature vector; The feature vector is input into a reinforcement learning model in the decision model to obtain a plurality of candidate decisions and their corresponding reward values.

4. The insurance claim processing method according to claim 1, It is characterized in that The step of processing the case information by the at least one intelligent agent comprises: For each called agent, the case information is input into a pre-trained model in the agent to obtain the task requirements corresponding to the case information; Generate a task processing framework corresponding to the task requirement through the pre-trained model; For each task in the task processing framework, an action instruction is retrieved from the action instruction knowledge base of the agent to execute the task, and an instruction execution result is obtained; Based on the obtained instruction execution results and the professional knowledge base in the agent, the agent processing results are generated through the pre-trained model.

5. The insurance claim processing method according to claim 1, It is characterized in that The step of calling at least one intelligent agent according to the target decision to process the case information through the at least one intelligent agent comprises: When the intelligent agent called according to the target decision is the decision model, generating a claim result of the insurance claim case according to the case basic information and the case processing information; The claim settlement result is sent to the terminal logged in by the target account.

6. The insurance claim processing method according to claim 1, It is characterized in that When multiple agents are called according to the target decision and agent processing results of the agents conflict, the method further includes: Obtaining the priority of each agent from the target decision, wherein the priority is calculated based on the task importance score, historical accuracy score, resource consumption score and historical performance score of each agent; The agent processing result of the highest priority agent is selected as the execution result of the target decision.

7. The insurance claim processing method according to claim 1, It is characterized in that When multiple agents are called according to the target decision and agent processing results of the agents conflict, the method further includes: Obtaining the historical accuracy scores and self-scores of each of the agents; Calculate the confidence score of each agent based on the historical accuracy score and self-score of each agent; The agent processing result of the agent with the highest confidence score is selected as the execution result of the target decision.

8. An insurance claims processing device, It is characterized in that include: An information acquisition module is used to acquire case information of insurance claim cases, wherein the case information includes case basic information and case processing information; An information input module, used to input the case information into a decision model to obtain a plurality of candidate decisions and their corresponding reward values; A decision selection module is used to select the candidate decision corresponding to the maximum reward value as the target decision; The case processing module is used to call at least one intelligent agent according to the target decision to process the case information through the at least one intelligent agent.

9. A computer device, It is characterized in that It comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the insurance claim processing method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the insurance claim processing method according to any one of claims 1 to 7.

Images