Insurance product design method and device, electronic equipment, chip and storage medium

Abstract

The invention provides an insurance product design method and device, electronic equipment, a chip and a storage medium, and the method comprises the steps: obtaining the static information and behavior information of at least one user, and enabling the behavior information to be obtained through employing a first network to carry out the feature extraction of the operation of the at least one user; inputting the static information and the behavior information into a trained insurance business reasoning module to obtain structured data, the structured data being used for indicating a corresponding relationship among the static information, the behavior information, the insurance industry information and the risk information; and generating a target insurance product according to the structured data and the generation target. Insurance industry information, risk information and static information and behavior information of a user can be analyzed and sorted to determine a corresponding relationship between requirements and characteristics of the user in the insurance industry and information quality inspection related to an insurance product, and then a target insurance product can be generated according to structured data. The generated insurance product can better meet user requirements, and the risk is lower.

Description

Technical Field

[0001] This disclosure relates to the field of data processing, and more particularly to an insurance product design method, apparatus, electronic device, chip, and storage medium. Background Technology

[0002] With the continuous development of algorithms and artificial intelligence technologies, data-driven and algorithm-optimized insurance product design methods have received increasing attention. However, traditional machine learning-based methods, such as those using decision trees and random forests, suffer from inefficiency or insufficient accuracy when dealing with complex and ever-changing insurance business scenarios. Summary of the Invention

[0003] This disclosure provides an insurance product design method, apparatus, electronic device, chip, and storage medium to solve problems in the related art.

[0004] A first aspect of this disclosure proposes an insurance product design method, comprising: acquiring static information and behavioral information of at least one user, wherein the behavioral information is obtained by feature extraction of the user's operations using a first network; inputting the static information and behavioral information into a trained insurance business reasoning module to obtain structured data, wherein the structured data is used to indicate the correspondence between the static information, behavioral information, insurance industry information, and risk information; and generating a target insurance product based on the structured data and a generation target.

[0005] In some embodiments of this disclosure, the method further includes: acquiring multiple market research data, multiple user historical behavior data, and multiple user historical static data; performing data preprocessing on the multiple market research data, multiple user historical behavior data, and multiple user historical static data to obtain a first training dataset; performing cross-processing and partitioning on the first training dataset and the second training dataset to obtain a basic knowledge training dataset, a product design training dataset, and a market dynamic analysis training dataset, wherein the second training dataset includes multiple historical insurance industry information and multiple historical risk information; and training the initial model using the basic knowledge training dataset, the product design training dataset, and the market dynamic analysis training dataset to obtain a trained insurance business reasoning module.

[0006] In some embodiments of this disclosure, training an initial model using a basic knowledge training dataset, a product design training dataset, and a market dynamics analysis training dataset to obtain a trained insurance business reasoning module includes: determining the data type of each training data in the basic knowledge training dataset, product design training dataset, and market dynamics analysis training dataset; determining the target expert network corresponding to the data type of each training data in a multi-expert hybrid system, wherein the target expert network is any one of an actuarial prediction expert network, a compliance review expert network, and a market risk assessment expert network; inputting each training data into the corresponding target expert network to obtain the prediction result corresponding to each training data; and adjusting the initial model based on the prediction results until the adjusted initial model meets the training termination condition to obtain the trained insurance business reasoning module.

[0007] In some embodiments of this disclosure, generating a target insurance product based on structured data and generation objectives includes: generating an initial insurance product using an insurance product generation algorithm based on the generation objectives and structured data; conducting compliance review and business rule verification on the initial insurance product to obtain the actuarial balance coefficient, market competitiveness index, and regulatory compliance corresponding to the initial insurance product; determining the fitness function of the initial insurance product based on the actuarial balance coefficient, market competitiveness index, and regulatory compliance; and optimizing the initial insurance product based on the fitness function to obtain the target insurance product.

[0008] In some embodiments of this disclosure, the method further includes: conducting a market simulation of the target insurance product based on current insurance industry information, current risk information, and current market research data, and obtaining simulation results, the simulation results including at least one of market share, customer feedback, and potential risks; and adjusting the target insurance product based on the simulation results.

[0009] In some embodiments of this disclosure, the method further includes: triggering a model fine-tuning instruction when the change in market research data is detected to be greater than or equal to a preset threshold; in response to the model fine-tuning instruction, acquiring updated multiple market research data, updated multiple user historical behavior data, and updated multiple user historical static data; and training a model for the trained insurance business reasoning module based on the updated multiple market research data, updated multiple user historical behavior data, and updated multiple user historical static data to adjust the trained insurance business reasoning module.

[0010] A second aspect of this disclosure provides an insurance product design apparatus, comprising: a first processing unit for acquiring static information and behavioral information of at least one user, wherein the behavioral information is obtained by feature extraction of the user's operations using a first network; a second processing unit for inputting the static information and behavioral information into a trained insurance business reasoning module to obtain structured data, wherein the structured data is used to indicate the correspondence between the static information, behavioral information, insurance industry information, and risk information; and a third processing unit for generating a target insurance product based on the structured data and a generation target.

[0011] A third aspect of this disclosure provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the methods described in the first aspect of this disclosure.

[0012] A fourth aspect of this disclosure provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the methods described in the first aspect of this disclosure.

[0013] A fifth aspect of this disclosure provides a chip including at least one processor and a communication interface; the communication interface is used to receive signals input to the chip or signals output from the chip, and the processor communicates with the communication interface and implements the method described in the first aspect of this disclosure through logic circuits or executing code instructions.

[0014] In summary, the insurance product design method proposed in this disclosure can acquire user behavioral and static information to understand user needs and preferences. Then, a trained large model can be used to perform data analysis on insurance industry information, risk information, user behavioral information, and static information, extracting key features, determining the relationships between these information, and generating structured data. This structured data can then be used to generate target insurance products that better meet market requirements, industry demands, and user needs.

[0015] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0017] Figure 1 A flowchart illustrating an insurance product design method provided in this disclosure embodiment. Figure 1 ;

[0018] Figure 2 A flowchart illustrating an insurance product design method provided in this disclosure embodiment. Figure 2 ;

[0019] Figure 3 A flowchart illustrating an insurance product design method provided in this disclosure embodiment. Figure 3 ;

[0020] Figure 4A An architecture diagram of an insurance product design optimization system based on a large model, provided for embodiments of this disclosure;

[0021] Figure 4B A flowchart illustrating an insurance product design optimization method based on a large model, provided for embodiments of this disclosure;

[0022] Figure 4C A flowchart illustrating another insurance product design optimization method based on a large model provided in this disclosure embodiment;

[0023] Figure 5 This is a schematic diagram of the structure of an insurance product design device provided in an embodiment of the present disclosure;

[0024] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure;

[0025] Figure 7 This is a schematic diagram of the chip structure provided in an embodiment of this disclosure. Detailed Implementation

[0026] Embodiments of this disclosure are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting this disclosure.

[0027] Traditional insurance product design methods often rely on human experience and intuition, lacking a systematic approach and struggling to quickly respond to market changes and meet customer needs. In recent years, with the continuous development of algorithms and artificial intelligence technologies, data-driven and algorithm-optimized insurance product design methods have received increasing attention. While some machine learning-based methods exist, such as those using decision trees and random forests, these methods suffer from inefficiency or insufficient accuracy when handling complex and ever-changing insurance business scenarios.

[0028] Traditional product design solutions have the following drawbacks in optimizing insurance sales strategies:

[0029] Currently, some insurance companies are trying to use data analysis tools to assist in the design of insurance products. However, these tools are usually based on fixed algorithms and models, and have limited data processing and analysis capabilities, making it difficult to fully capture market dynamics and changes in customer needs.

[0030] Some companies use machine learning technology to optimize insurance product design, but the models they use are small in scale, have limited processing power, and are difficult to adapt to complex and ever-changing insurance business scenarios.

[0031] Therefore, in order to solve the above problems, this disclosure proposes an insurance product design method that integrates large model technology to achieve intelligent and optimized insurance product design, thereby improving the accuracy and market adaptability of product design.

[0032] The specific details of this method are as follows.

[0033] Figure 1 A flowchart illustrating an insurance product design method provided in this disclosure embodiment. Figure 1 .like Figure 1 As shown, the method may include the following steps.

[0034] Step 101: Obtain static information and behavioral information of at least one user.

[0035] In some embodiments, static information and behavioral information of the user can be obtained. Static information may include user-related information, such as the user's age, occupation, medical examination results, user query information, user input information, and user needs. User-related information may include historical order information, such as the user's past insurance purchases, historical policies, and coverage amounts. Optionally, static information can also be determined based on the user's historical information stored in a database. Static information can be used to indicate the user's basic conditions, such as physical health status, economic situation, and user preferences.

[0036] In some embodiments, user behavior information can be obtained by using a first network to extract features from the actions of at least one user. The first network can be, for example, a Temporal Convolutional Network (TCN). User actions include click actions, browsing actions, etc. Optionally, when users select or browse insurance product introductions on the official website, they usually browse products of interest for a longer period of time or click more times. Therefore, features can be extracted from user actions, such as dwell time, click stream, etc. User behavior information can be used to indicate user needs and user preferences.

[0037] In some embodiments, the database can store customer cases, product information, industry knowledge (such as regulatory requirements for the insurance industry, insurance-related knowledge, etc.), user historical information, user profiles, etc. That is, the database contains life insurance product information (such as product name, coverage, premium rates, claims conditions, etc.), industry dynamics (such as new policy releases, market trends, etc.), user historical behavior information, and customer cases. The database can store the above data in a structured manner. For example, various types of life insurance-related knowledge collected can be structured, and knowledge graph technology can be used to associate and represent the knowledge, thereby determining the relationships between various pieces of knowledge and generating structured data. This structured data can be stored in the database. This structured data storage method improves the efficiency of knowledge retrieval and reasoning ability when querying data from the database.

[0038] In some embodiments, when designing insurance products, static and behavioral information of users can be obtained based on the type of insurance product. For example, when the insurance product is auto insurance, data related to auto insurance can be retrieved from a database. This includes information on existing auto insurance products, order information for historical auto insurance products, and information about the user who purchased the auto insurance product, i.e., the user's static information, such as the user's age and occupation. Behavioral information can also be obtained, such as which auto insurance products the user browsed and the duration of their browsing.

[0039] In some embodiments, the database may optionally include a cache area and a non-cache area. The cache area includes at least one of the following: information on at least one hot candidate product, at least one historical user profile, and at least one historical customer case. In other words, the database may include a cache area that can cache information related to hot products. Hot products may be, for example, products that are purchased in large quantities, products with high sales volume over a period of time, or products with high sales volume in the current geographic area, such as travel insurance.

[0040] Optionally, the cache can cache user image information, such as user historical order information, user static feature information, user preference information, etc. The cache can also cache customer case information and the results of the above semantic similarity matching, that is, it can cache the correspondence between user needs and products. Optionally, when obtaining user static information and behavioral information, it can first determine whether relevant user static information and behavioral information exist in the database cache. If relevant user static information and behavioral information exist in the database cache, the information cached in the cache can be directly obtained. If relevant user static information and behavioral information do not exist in the database cache, the relevant information can be obtained from the non-cache area.

[0041] Step 102: Input static information and behavioral information into the trained insurance business reasoning module to obtain structured data.

[0042] In some embodiments, the trained insurance business reasoning module can be a large model. Leveraging the deep learning capabilities of this large model, it can extract key features from the input data. The trained insurance business reasoning module can identify potential patterns and regularities in the extracted feature data. These patterns may reveal trends in customer behavior, market changes, etc., which can help optimize product design strategies. The trained insurance business reasoning module can integrate the extracted features and identified patterns to form structured data. This structured data is used to indicate the correspondence between static information, behavioral information, insurance industry information, and risk information. In other words, the structured data contains a three-level feature system: Basic layer: static customer profile (age / occupation, etc.); Dynamic layer: temporal behavioral features (extracted through a TCN network); Prediction layer: risk probability distribution (generated based on Monte Carlo simulation).

[0043] In some embodiments, user static information and behavioral information can be used to reflect user-related characteristics such as user needs and preferences. Insurance industry information refers to information related to the insurance industry, such as basic insurance knowledge, insurance regulatory rules, insurance compliance rules, and insurance product requirements. Risk information is obtained by analyzing at least one of the static information, behavioral information, and insurance industry-related information, including potential risks in designing insurance products, as well as the probability and probability distribution of various risks. Therefore, risk issues can be used to highlight issues that need attention during the design of insurance products.

[0044] In some embodiments, the method further includes: acquiring multiple market research data, multiple user historical behavior data, and multiple user historical static data; performing data preprocessing on the multiple market research data, multiple user historical behavior data, and multiple user historical static data to obtain a first training dataset; performing cross-processing and partitioning on the first training dataset and the second training dataset to obtain a basic knowledge training dataset, a product design training dataset, and a market dynamic analysis training dataset, wherein the second training dataset includes multiple historical insurance industry information and multiple historical risk information; and training the initial model using the basic knowledge training dataset, the product design training dataset, and the market dynamic analysis training dataset to obtain a trained insurance business reasoning module.

[0045] In some embodiments, data collection methods may include API calls, web crawlers, database queries, etc. That is, at least one of multiple market research data, multiple user historical behavior data, and multiple user historical static data may be obtained from a database, or through API calls, or through web crawlers.

[0046] In some embodiments, market research data may be data related to insurance products in the market, such as historical sales data, market demand types, competitor product information, social media data, insurance regulatory requirements, publicly available industry reports and statistics, etc. Market research data can be used to indicate current market dynamics and reflect the demand and requirements for insurance products in the current market environment.

[0047] In some embodiments, a user's historical behavior data may include the user's historical health behavior data, such as exercise plans and dietary records, as well as user browsing operations, click operations, and other behavioral data. In other words, the user's historical behavior data can be recorded, which can reflect the applicable groups of different types of insurance products, as well as the user's needs and preferences.

[0048] In some embodiments, multiple sets of historical static data for users can be the user's historical static information, such as information related to the user themselves, historical order information, etc. Information related to the user themselves includes, for example, the user's age, occupation, medical examination results, user query information, user input information, user needs, etc. Historical order information includes, for example, data on the user's historical insurance purchases, historical policies, and coverage amounts. Optionally, static information can also be determined based on the user's historical information stored in the database. Static information can be used to indicate the user's basic conditions, such as physical health status, economic situation, and user preferences.

[0049] In some embodiments, multiple market research data sets, multiple user historical behavior data sets, and multiple user historical static data sets can be preprocessed to obtain a first training dataset. Data preprocessing includes, for example, data cleaning, data deduplication, data formatting, and data quality verification. Data cleaning cleans the collected data, including removing duplicate data, handling missing values, and correcting erroneous data. Data deduplication can use hash algorithms or similarity calculation methods to ensure data uniqueness. Data formatting converts the data into a uniform format for easier subsequent processing and analysis. Data quality verification ensures data accuracy and consistency by setting data quality verification rules. Optionally, after obtaining the first training dataset, the training data in the first training dataset can be stored in a database, such as a distributed database or a cloud database.

[0050] In some embodiments, after obtaining the first training dataset, a second training dataset can be acquired. The second training dataset includes multiple historical insurance industry information sets and multiple historical risk information sets. Similarly, these sets can be obtained through API calls, web crawlers, database queries, etc. The historical insurance industry information and risk information can be information from before the model is put into use. That is, during model training, current insurance industry information and risk information can be used for training. After the model training is complete and the trained insurance business reasoning module is obtained, the insurance industry information and risk information may have been updated when the insurance business reasoning module is used. At this point, the insurance industry information and risk information used during training are historical insurance industry information and historical risk information, but the actual insurance industry information and historical insurance industry information are of the same type, and the historical risk information and risk information are of the same type. Therefore, for an explanation of historical insurance industry information and historical risk information, please refer to the explanation of insurance industry information and risk information.

[0051] In some embodiments, the first and second training datasets can be cross-processed and partitioned to obtain a basic knowledge training dataset, a product design training dataset, and a market dynamics analysis training dataset. Optionally, the first and second training datasets can be cross-fused first to obtain a new training dataset, which includes all data from both datasets. Then, the data can be partitioned according to data type. Optionally, for example, data related to basic insurance concepts can be partitioned into the basic knowledge training dataset, which includes basic insurance industry data such as insurance requirements and rules. Data related to product design can be partitioned into the product design training dataset, which includes product design-related data such as insurance product format requirements, design principles, and risk information. Data related to market dynamics analysis can be partitioned into the market dynamics analysis training dataset, which includes market dynamics-related information such as market trends and attention levels of insurance products. In other words, the solution of this application introduces a course learning mechanism, adjusting the training samples according to a difficulty gradient of "basic insurance concepts - product design rules - market dynamics analysis".

[0052] In some embodiments, the initial model can be trained using a basic knowledge training dataset, a product design training dataset, and a market dynamics analysis training dataset to obtain a trained insurance business reasoning module. Optionally, the initial model can be trained sequentially using the basic knowledge training dataset, the product design training dataset, and the market dynamics analysis training dataset, according to the aforementioned difficulty gradient. Optionally, when training the initial model, it can first be pre-trained on a general corpus, and then adversarial training methods can be used to adapt it to insurance domain data (actuarial reports, policy texts, etc.).

[0053] In some embodiments, training an initial model using a basic knowledge training dataset, a product design training dataset, and a market dynamics analysis training dataset to obtain a trained insurance business reasoning module includes: determining the data type of each training data in the basic knowledge training dataset, product design training dataset, and market dynamics analysis training dataset; determining the target expert network corresponding to the data type of each training data in a multi-expert hybrid system, wherein the target expert network is any one of an actuarial prediction expert network, a compliance review expert network, and a market risk assessment expert network; inputting each training data into the corresponding target expert network to obtain the prediction result corresponding to each training data; and adjusting the initial model based on the prediction results until the adjusted initial model meets the training termination condition to obtain the trained insurance business reasoning module.

[0054] In some embodiments, the initial model of this application includes a Mixture of Experts (MoE) system, comprising actuarial forecasting experts, compliance review experts, and market risk assessment experts. Optionally, the actuarial forecasting experts may also include multiple sub-experts, each of whom may be responsible for a type of insurance. The data type of the training data may be the type of insurance to which the training data belongs; for example, auto insurance data may be processed by auto insurance-related sub-experts. Alternatively, the data type of the training data may be the domain to which the training data belongs; for example, risk information may be input to market risk assessment experts for processing.

[0055] Optionally, the prediction results can be structured data, i.e., predicting the correspondence between static information, behavioral information, insurance industry information, and risk information. Then, based on the correctness of the predicted correspondence and the similarity between the predicted and actual correspondences, it can be determined whether the model needs adjustment. When model adjustment is necessary, the model parameters, model architecture, optimization strategies, training data, etc., can be adjusted to improve the model's prediction accuracy. Optionally, termination conditions can be, for example, meeting a preset number of training iterations, model convergence, or the model's prediction accuracy reaching a preset threshold; this disclosure does not limit these conditions.

[0056] Optionally, a reward model based on reinforcement learning can be designed to transform insurance regulatory rules into a quantifiable reward function. That is, after obtaining the predicted structured data, it can be determined whether the predicted structured data meets the insurance regulatory rules, and the corresponding reward function can be determined. Then, the model can be adjusted and optimized based on the value of the reward function.

[0057] Step 103: Generate the target insurance product based on the structured data and the generated target.

[0058] In some embodiments, key factors such as market demand, customer preferences, and competitive landscape can be analyzed in depth to clarify the goals and positioning of product design, i.e., to determine the generation target. Then, static information and behavioral information can be input into the trained insurance business reasoning module to obtain structured data. Finally, the target insurance product can be generated based on the structured data and the generation target.

[0059] Optionally, a suitable product design algorithm can be designed by combining insurance business rules and actuarial principles. The algorithm needs to be able to receive structured data output from a large model and generate a preliminary product design scheme based on this data. Optionally, a detailed product design script can be written based on the algorithm design. The script needs to include multiple modules such as data processing, scheme design, and rule validation to ensure that a preliminary product design scheme can be automatically generated, including aspects such as coverage, premium calculation method, claims process, and exclusion clauses.

[0060] In summary, the above embodiments of this disclosure can acquire user behavior information and static information to obtain user needs and preferences. Subsequently, a trained large model can be used to perform data analysis on insurance industry information, risk information, user behavior information, and static information, extract key features, determine the relationships between insurance industry information, risk information, user behavior information, and static information, and generate structured data. Then, target insurance products can be generated based on the structured data, making the generated target insurance products more in line with market requirements, industry requirements, and user needs.

[0061] Figure 2 A flowchart illustrating an insurance product design method provided in this disclosure embodiment. Figure 2 .like Figure 2 As shown, based on Figure 1 The illustrated embodiment shows that the method includes the following steps.

[0062] Step 201: Based on the generation target and structured data, an initial insurance product is generated using an insurance product generation algorithm.

[0063] Optionally, a suitable product design algorithm (i.e., product generation algorithm) can be designed by combining insurance business rules and actuarial principles. The product generation algorithm needs to be able to receive structured data output from a large model and generate a preliminary product design scheme, i.e., an initial insurance product, based on this data and the generation target.

[0064] Step 202 involves conducting a compliance review and business rule verification of the initial insurance product to obtain the actuarial balance coefficient, market competitiveness index, and regulatory compliance level corresponding to the initial insurance product.

[0065] In some embodiments, the actuarial balance coefficient can be used to measure the balance between premium income and expected claims / costs to ensure the long-term financial sustainability of insurance products. The market competitiveness index can be used to indicate the competitiveness of the initial insurance product in the current market. The regulatory compliance can be used to indicate whether the initial insurance product meets regulatory and business requirements and whether there are risks, etc.

[0066] In some embodiments, a compliance review can be conducted on the initial insurance product. This review can assess the preliminary product design to ensure it complies with insurance business rules and legal regulations. The review includes assessing the legality of the terms and conditions, and the reasonableness of the premium rates. Optionally, the initial insurance product can be validated against business rules. This means that the plan can be further validated according to the insurance company's business rules. The validation includes assessing the plan's feasibility and market competitiveness, resulting in actuarial balance coefficients, market competitiveness index, and regulatory compliance for the initial insurance product.

[0067] Optionally, the actuarial balance coefficient can be determined based on the claim probability, sum insured, and premium of the initial insurance product. For example, the actuarial balance coefficient can be "claim probability × sum insured - premium". The market competitiveness index is determined by the similarity with competitors. For example, the market competitiveness index can be the inverse ratio of the similarity between the initial insurance product and competitors. The regulatory compliance can be determined based on whether the terms of the initial insurance product meet business rules and compliance review. Optionally, the terms can be deemed to be in violation based on compliance review and business rule verification, and points can be deducted for the terms to obtain the final regulatory compliance.

[0068] Step 203: Determine the fitness function of the initial insurance product based on the actuarial balance coefficient, market competitiveness index, and regulatory compliance.

[0069] In some embodiments, an actuarial balance coefficient, a market competitiveness index, and regulatory compliance can be used to determine the fitness function of the initial insurance product, which is used to indicate the gap between the initial insurance product and the generation target.

[0070] Step 204: Optimize the initial insurance product according to the fitness function to obtain the target insurance product.

[0071] In some embodiments, the initial product design can be modified and optimized based on a fitness function. Modifications may involve adjustments to terms and conditions, changes in rates, etc., to ensure the design better meets market demands and insurance company requirements.

[0072] Optionally, when optimizing the initial insurance product, reinforcement learning-driven dynamic optimization can be used. For example, a Markov Decision Process (MDP) model can be constructed to optimize the initial insurance product. The state space of the MDP model is the market dynamics plus the company's capital adequacy ratio, the action space is the combination of premium adjustment range and coverage scope, and the reward function is expected profit plus customer satisfaction plus compliance score. Optionally, a dual-delay deep deterministic policy gradient algorithm can also be developed for policy optimization.

[0073] In some embodiments, the method further includes: conducting a market simulation of the target insurance product based on current insurance industry information, current risk information, and current market research data, and obtaining simulation results, the simulation results including at least one of market share, customer feedback, and potential risks; and adjusting the target insurance product based on the simulation results.

[0074] In other words, current insurance industry information, current risk information, and current market research data can be obtained to conduct market simulations of target insurance products. For example, Monte Carlo simulations can be used to simulate the market for target insurance products, predicting key indicators such as market share and customer feedback (e.g., customer satisfaction). The simulated market share and customer satisfaction indicators can be analyzed to assess the potential risks and challenges facing the product design, including market risk and credit risk. Based on the simulation and assessment results, further predictions and adjustments can be made to the product design, potentially involving expanding coverage, simplifying claims processes, etc., to enhance the product's market competitiveness.

[0075] In summary, the above embodiments of this application can generate an initial insurance product based on structured data, and optimize the initial insurance product to obtain a target insurance product. Optionally, market simulation can be performed on the target insurance product, and the target insurance product can be further optimized based on the simulation results. This can realize the integration of large model technology for insurance data processing, improve the efficiency and accuracy of data processing, provide strong data support for product design, optimize the product design process, reduce the influence of subjective factors, improve the accuracy and market adaptability of product design, improve the scientificity and rationality of product design, and reduce market risks.

[0076] Figure 3 A flowchart illustrating an insurance product design method provided in this disclosure embodiment. Figure 3 .like Figure 3 As shown, based on Figure 1 The illustrated embodiment shows that the method includes the following steps.

[0077] Step 301: If the change in market research data is greater than or equal to a preset threshold, a model fine-tuning instruction is triggered.

[0078] In some embodiments, changes in market research data can be monitored. If the market research data changes and the magnitude of the change is greater than or equal to a preset threshold, it is determined that a significant change has occurred in the market research data. In this case, market demand for insurance products, industry regulatory requirements for insurance products, etc., may have changed, and the model needs to be adjusted. Optionally, the preset threshold can be set according to actual needs, which can trigger a model fine-tuning instruction. This disclosure does not limit this.

[0079] Step 302: In response to the model fine-tuning instruction, obtain updated market research data, updated user historical behavior data, and updated user historical static data.

[0080] In some embodiments, in response to a model fine-tuning instruction, multiple updated market research data, multiple updated user historical behavior data, and multiple updated user historical static data can be obtained. In other words, the current market research data, user behavior data, and user static data can be re-obtained to determine the content that has been updated.

[0081] Step 303: Train the model of the trained insurance business reasoning module based on the updated market research data, updated user historical behavior data, and updated user historical static data, so as to adjust the trained insurance business reasoning module.

[0082] In some embodiments, the trained insurance business reasoning module can be trained based on updated market research data, updated user historical behavior data, and updated user historical static data, thereby updating the model to meet updated market demands and requirements.

[0083] In summary, the above embodiments of this disclosure can enable timely adjustment of the model when the change in market research data is greater than or equal to a preset threshold, thereby improving the model's generalization ability and the accuracy of the structured data output by the model.

[0084] The technical solutions of this disclosure will be further described in detail below with reference to specific application embodiments.

[0085] The following is an embodiment of an insurance product design method and optimization system based on a large model provided in this disclosure.

[0086] like Figure 4A The diagram shown is an architecture diagram of an insurance product design optimization system based on a large model. Figure 4B The diagram shows a flowchart of an insurance product design method based on a large model.

[0087] 1. System Architecture

[0088] In some embodiments, the insurance product design optimization system based on a large model mainly consists of four modules: a data collection module, a large model processing module, a product design module, and an optimization feedback module.

[0089] Data collection module: Responsible for collecting data from multiple sources, including market research reports, customer behavior data, competitor product information, social media data, etc.; data collection methods may include API calls, web crawling, database queries, etc.; the collected data needs to be initially cleaned and formatted to ensure data quality and consistency.

[0090] Large Model Processing Module: Select a large model suitable for insurance business, such as a natural language processing model; pre-train the large model to enable it to understand and process insurance-related text and data; input the cleaned data into the large model for in-depth analysis and mining to extract valuable information and patterns, such as customer demand distribution, market competition landscape, and risk points; the output of the large model processing module will serve as the input to the product design module.

[0091] Product Design Module: Based on the output of the large model processing module, combined with insurance business rules and actuarial principles, a preliminary product design scheme is generated. The product design scheme includes coverage, premium calculation method, claims process, exclusion clauses, etc. The product design module also needs to consider factors such as product market positioning, target customer group, and sales channels. The preliminary product design scheme will serve as input for the optimization feedback module.

[0092] The optimization feedback module simulates and forecasts the preliminary product design scheme, and evaluates its market performance and customer feedback. Simulation and forecasting methods may include Monte Carlo simulation, market research, customer interviews, etc. Based on the simulation and forecasting results, the preliminary product design scheme is optimized and adjusted to form the final product design result. The final product design result will be fed back to relevant departments for implementation, and the market performance of the product will be continuously tracked and evaluated.

[0093] 2. For example Figure 4C As shown, the technical implementation process of the insurance product design optimization method based on the large model is as follows.

[0094] 1) Data collection:

[0095] Diversified sources: Ensure the diversity of data collection sources, including but not limited to market research reports, customer behavior data, competitor product information, social media data, publicly available industry reports and statistics.

[0096] API Interface Calls: Establish partnerships with third-party data providers to obtain relevant data in real time through API interfaces.

[0097] Web crawling technology: For data that cannot be directly obtained through APIs, develop customized web crawler programs to scrape data from relevant websites and platforms.

[0098] Database query: Integrates internal database resources and retrieves the required data through SQL query statements.

[0099] Data preprocessing:

[0100] Data cleaning: Cleaning the collected data, including removing duplicate data, handling missing values, and correcting erroneous data.

[0101] Data deduplication: Using hash algorithms or similarity calculation methods to ensure the uniqueness of data.

[0102] Data formatting: Converting data into a uniform format to facilitate subsequent processing and analysis.

[0103] Data quality verification: By setting data quality verification rules, we ensure the accuracy and consistency of the data.

[0104] Data storage:

[0105] Database selection: Choose a database system suitable for storing big data, such as a distributed database or a cloud database.

[0106] Data indexing and optimization: Indexing and optimizing the database to improve the efficiency of data querying and processing.

[0107] 2) Large-scale model analysis and mining:

[0108] a. Large Model Selection and Pre-training

[0109] Model selection: Based on the characteristics and needs of the insurance business, the most suitable model for this application is selected from a large pool of models. Selection criteria include model accuracy, processing speed, and scalability.

[0110] Data preparation: Collect large-scale insurance-related datasets for pre-training large models. The datasets should cover different types of insurance business scenarios to ensure the model's generalization ability.

[0111] Pre-training process: A selected large model is pre-trained using either unsupervised or supervised learning methods. During pre-training, the model's parameters and structure are adjusted to better understand and process insurance-related text and data.

[0112] b. Data Input and Processing

[0113] Data encoding: Transforming the cleaned and formatted data into an input format acceptable to large models. This includes preprocessing steps such as word segmentation, stop word removal, and stemming of text data.

[0114] Data batching: To improve processing efficiency, the data is divided into multiple batches, each containing a certain number of data samples. The batch size should be reasonably set based on the memory and processing capabilities of the large model.

[0115] c. Large-scale model in-depth analysis and mining

[0116] Feature extraction: Leveraging the deep learning capabilities of large models, key features are extracted from the input data. These features may include customer demand distribution, market competition landscape, risk points, etc., which are of significant value for subsequent product design.

[0117] Pattern recognition: Through the training and learning of large models, it identifies potential patterns and regularities in data. These patterns may reveal trends in customer behavior, market changes, etc., which can help optimize product design strategies.

[0118] Information Integration: The extracted features and identified patterns are integrated to form structured data analysis results. These results will serve as input for subsequent product design modules, guiding the design and optimization of insurance products.

[0119] d. Results Output and Visualization

[0120] Structured data output: The analysis and mining results of the large model are output in the form of structured data. This includes feature vectors, pattern recognition results, etc., which facilitates further processing and analysis by subsequent modules.

[0121] Data visualization: Using visualization techniques such as charts and heatmaps to display the results of analysis and data mining. This helps business personnel to understand the information and patterns in the data more intuitively, providing a clear basis for product design.

[0122] 3) Product design generation:

[0123] a. Product design algorithm and script development

[0124] Needs Analysis: First, we conduct an in-depth analysis of key factors such as market demand, customer preferences, and competitive landscape to clarify the goals and positioning of the product design.

[0125] Algorithm Design: Combining insurance business rules and actuarial principles, design a suitable product design algorithm. The algorithm needs to be able to receive structured data output from large models and generate preliminary product design schemes based on this data.

[0126] Script Writing: Based on algorithm design, write detailed product design scripts. The scripts should include multiple modules such as data processing, solution design, and rule validation to ensure the automated generation of preliminary product design solutions.

[0127] b. Generation of preliminary product design scheme

[0128] Data Input: The structured data after processing the large model is input into the product design script as input parameters for solution generation.

[0129] Solution Design: Based on the input data and preset algorithms, the script automatically generates a preliminary product design plan. The plan includes multiple aspects such as coverage, premium calculation method, claims process, and exclusion clauses.

[0130] Solution output: Output the generated preliminary product design solution in the form of a document or spreadsheet for subsequent review and optimization.

[0131] c. Plan review and revision

[0132] Compliance Review: Conduct a compliance review of the preliminary product design to ensure that the plan complies with insurance business rules and legal regulations. The review includes assessing the legality of the terms and conditions, and the reasonableness of the premium rates.

[0133] Business rule verification: The plan is further verified according to the insurance company's business rules. Verification includes the plan's feasibility and market competitiveness.

[0134] Modification and Optimization: Based on the review results, the preliminary product design will be modified and optimized. Modifications may involve adjustments to terms and conditions, changes to rates, etc., to ensure the plan better meets market demands and the requirements of the insurance company.

[0135] d. Scheme simulation and prediction

[0136] Market simulation: Using methods such as Monte Carlo simulation, market simulations are conducted on the optimized product design. The simulation includes key indicators such as market share and customer satisfaction.

[0137] Risk assessment: Analyze the simulation results to assess the potential risks and challenges facing the product design. The assessment includes market risk, credit risk, etc.

[0138] Forecast Adjustment: Based on simulation and evaluation results, further forecasts and adjustments are made to the product design. These adjustments may involve expanding coverage, simplifying claims processes, etc., to enhance the product's market competitiveness.

[0139] e. Final product design scheme determination

[0140] Solution integration: Integrate the product design solutions that have been reviewed, modified, simulated, and predicted to form the final product design solution.

[0141] Solution Review: Organize experts to review the final product design solution to ensure its feasibility and innovativeness. The review includes the solution's technical implementation and market prospects.

[0142] Finalization of the design: Based on the review results, the final product design will be finalized. The finalized design will serve as the basis for subsequent product development and marketing.

[0143] 4) Solution optimization and evaluation:

[0144] Write simulation and forecasting scripts or programs to simulate and forecast the market performance and customer feedback of preliminary product design schemes.

[0145] Based on the simulation and prediction results, the preliminary product design scheme was optimized and adjusted.

[0146] The optimized product design scheme, after further review and modification, forms the final product design result.

[0147] The final product design results are fed back to relevant departments for implementation, and the market performance of the product is continuously tracked and evaluated.

[0148] 3. Implementation of key technologies

[0149] Selection and training of large models:

[0150] Based on the needs and characteristics of insurance business, a large model with strong natural language processing and understanding capabilities is selected.

[0151] Large models are pre-trained to understand and process insurance-related text and data. The pre-training process can include language model training, domain-adaptive training, and more.

[0152] Regularly update and optimize the large model to improve its accuracy and adaptability.

[0153] Data fusion and mining:

[0154] Write data fusion scripts or programs to merge and integrate multi-dimensional data to form a unified data view.

[0155] By leveraging the deep learning capabilities of large models, we can mine and analyze the fused data to extract valuable information and patterns.

[0156] Visualize and display the mining results to facilitate understanding and use by business personnel.

[0157] Solution optimization and evaluation:

[0158] Write simulation and forecasting scripts or programs to simulate and predict the market performance and customer feedback of preliminary product design schemes. Simulation and forecasting methods may include Monte Carlo simulation, market research questionnaires, and analysis of customer interview records.

[0159] Based on the simulation and prediction results, write optimization algorithms or scripts to optimize and adjust the preliminary product design.

[0160] The optimized product design needs to be reviewed and revised again to ensure that it complies with insurance business rules and legal regulations.

[0161] In summary, the above-described examples of this disclosure, through the integration of large model technology, have achieved intelligent and optimized insurance product design, improved the accuracy and market adaptability of product design, optimized the product design process, reduced the influence of subjective factors, and improved the efficiency and scientific nature of product design.

[0162] The solution in this application is improved in the following aspects:

[0163] 1. Enhanced structured data generation mechanism.

[0164] 1) Multimodal fusion processing mechanism.

[0165] In the data preprocessing stage, an unstructured data vectorization method based on multi-head attention mechanism is added, and a vector space mapping model specific to the insurance domain is established (such as using the BERT-Insurance embedding layer); a dynamic fusion algorithm for time-series data and spatial data is used, and an LSTM-GCN hybrid network is used to process the spatiotemporal correlation features in insurance business; an automated construction module for domain knowledge graphs is established, which generates dynamically updated insurance knowledge graphs through the entity relationship extraction capability of large models.

[0166] 2) Dynamic update mechanism for structured data.

[0167] Design a feedback-driven incremental learning system that automatically triggers model fine-tuning when market data fluctuations exceed a threshold (e.g., ±15%); develop a distributed data update architecture based on federated learning to achieve cross-institutional data collaboration while protecting privacy.

[0168] Structured data includes a three-level feature system:

[0169] Basic layer: Static customer profile (age / occupation, etc.)

[0170] Dynamic layer: Temporal behavioral features (extracted via TCN network)

[0171] Prediction layer: Risk probability distribution (generated based on Monte Carlo simulation)

[0172] 2. Improved training methods for large models.

[0173] 1) Domain-adaptive training strategy.

[0174] Design a two-stage training framework:

[0175] Phase 1: Pre-training on general corpora (such as WikiText);

[0176] Phase Two: Adversarial training methods are used to adapt data in the insurance field (actual reports, policy texts, etc.);

[0177] A course learning mechanism is introduced, and the training samples are adjusted according to the difficulty gradient of "basic insurance concepts → product design rules → market dynamics analysis".

[0178] 2) Model architecture improvement.

[0179] Insert an insurance-specific adapter into the Transformer architecture. Develop a multi-expert hybrid system (MoE) comprising: actuarial prediction experts, compliance review experts, and market risk assessment experts. Design a reinforcement learning-based reward model to transform insurance regulatory rules into quantifiable reward functions.

[0180] 3. Optimization of product design generation algorithm.

[0181] 1) Dynamic optimization driven by reinforcement learning.

[0182] Construct a Markov Decision Process (MDP) model: State space: market dynamics + company capital adequacy ratio; Action space: premium adjustment range / coverage combination; Reward function: expected profit + customer satisfaction + compliance score.

[0183] We developed a dual-delay deep deterministic policy gradient (TD3) algorithm for policy optimization.

[0184] 2) Improved genetic algorithm.

[0185] The fitness function is designed to include: actuarial balance coefficient (claim probability × sum assured - premium); market competitiveness index (inversely proportional to similarity with competitors); and regulatory compliance (deductions for policy violations).

[0186] A quantum genetic algorithm is introduced to improve the mutation operation and enhance the efficiency of parameter search.

[0187] 4. Optimize the upgrade of the feedback module.

[0188] 1) Real-time feedback mechanism.

[0189] Establish an online A / B testing platform and dynamically allocate traffic using the bandit algorithm; develop a counterfactual evaluation model based on causal inference to quantify the market impact of each design element.

[0190] 2) Automated compliance checks.

[0191] Construct a knowledge graph of insurance terms (containing 3000+ regulatory rule nodes); develop a rule parsing engine to convert natural language terms into executable logical expressions; and achieve automatic detection of term conflicts and generation of correction suggestions.

[0192] The following are some scenario-based examples provided by this solution.

[0193] Scenario 1: Customized insurance products based on customer behavior data.

[0194] Background: An insurance company wants to analyze data such as customers' purchase history, browsing behavior, and social media interactions to customize personalized insurance products for different customer groups.

[0195] Application: Using this application system, customer behavior data is first acquired through the data collection module. Then, the big data processing module analyzes and mines this data to extract customers' potential needs and risk preferences. Finally, the product design module combines these analysis results to generate personalized insurance product design solutions that meet the customer's needs.

[0196] Effect: This example demonstrates the application of this application in enhancing the personalization capabilities of insurance products, which helps insurance companies better meet customer needs and improve market competitiveness.

[0197] Scenario 2: Insurance product innovation based on market competition.

[0198] Background: With the continuous development of the insurance market, competitors' product strategies and market dynamics have a significant impact on insurance companies' product design strategies.

[0199] Application: Using this application's system, the large-scale model processing module can conduct in-depth analysis and mining of competitors' product information, market share, marketing strategies, and other data to extract key characteristics of the market competition landscape. Then, the product design module combines these analysis results to innovatively design insurance products to stand out from the competition.

[0200] Effect: This example demonstrates the application of this application in assisting insurance companies in product innovation, helping them to respond quickly to market changes and improve the market competitiveness and profitability of their products.

[0201] Scenario 3: Optimization of insurance product pricing based on risk prediction.

[0202] Background: The pricing of insurance products needs to take into account a variety of risk factors, such as natural disasters, traffic accidents, and the incidence of diseases.

[0203] Application: Using the system described in this application, the large-scale model processing module can perform deep learning and analysis on historical risk data to predict the probability of future risks and the extent of losses. Then, the product design module combines actuarial principles with the analysis results of the large-scale model to price insurance products reasonably, balancing the insurance company's risks and returns.

[0204] Results: This example demonstrates the application of this application in optimizing insurance product pricing, which helps insurance companies to more accurately assess risks, formulate reasonable insurance product prices, and improve profitability. Specific implementation examples:

[0206] Example 1: Design and optimization of health insurance products.

[0207] Background and Objectives: With increasing health awareness, the demand for health insurance is growing, but customer needs are diverse and traditional design methods are insufficient to meet them. This embodiment aims to utilize the system of this application to design a health insurance product that better meets market demands and customer preferences.

[0208] Implementation steps:

[0209] Data collection:

[0210] Data sources include market research reports, medical institution data, customer health behavior data (such as step count, diet records, etc.), and social media discussions on health topics.

[0211] Collection methods: Data is collected by calling third-party data providers through API interfaces, web crawlers are used to crawl publicly available data from social media and medical institutions, and internal customer databases are integrated.

[0212] Data preprocessing:

[0213] Data cleaning: removing duplicate data, handling missing values ​​(e.g., filling with the mean or median), and correcting erroneous data.

[0214] Data formatting: Converting data from different sources into a uniform format to facilitate subsequent processing.

[0215] Large-scale model analysis and mining:

[0216] Model selection: The DeepSeek series of large models was selected, which performs excellently in natural language processing and understanding.

[0217] Pre-training: The model is pre-trained using a large-scale insurance and health-related dataset to enable it to better understand and process health insurance-related data.

[0218] Data input: Input the cleaned and formatted data into the large model, including customer health behavior, medical expenditure trends, disease incidence, etc.

[0219] Analysis and Mining: The large model extracts information such as customer health risk characteristics, disease association patterns, and market demand trends.

[0220] Product Design:

[0221] In accordance with business rules: Design the product's coverage (such as hospitalization medical care, critical illness protection, etc.) and premium calculation method (such as tiered pricing based on age, gender, and health status) based on insurance business rules and actuarial principles.

[0222] Consider market positioning: Design differentiated product solutions for different customer groups, such as young and healthy people and middle-aged and elderly people with chronic diseases.

[0223] Generate a preliminary plan: Based on the analysis results of the large model, combined with business rules and market positioning, generate a preliminary health insurance product design plan.

[0224] Solution optimization and evaluation:

[0225] Simulation and forecasting: Using methods such as Monte Carlo simulation, simulate and forecast the market performance and customer feedback of the preliminary plan.

[0226] Optimization and Adjustment: Based on the simulation and prediction results, optimization and adjustment will be made to the scope of coverage, premium prices, claims process, etc.

[0227] Final review and approval: Organize experts to review the final plan to ensure its feasibility and innovation, and finalize the plan.

[0228] Implementation Results: The health insurance products designed using this application system are more in line with market demands and customer preferences, enhancing their market competitiveness. Simultaneously, the application of large-scale modeling technology improves the efficiency and accuracy of product design, reducing the impact of subjective factors.

[0229] Example 2: Auto Insurance Product Design and Optimization:

[0230] Background and Objectives: With the increasing number of cars on the road, competition in the auto insurance market is becoming increasingly fierce. This embodiment aims to utilize the system of this application to design a more competitive auto insurance product, thereby increasing the insurance company's market share and customer satisfaction.

[0231] Implementation steps (similar to health insurance, but adjusted for the characteristics of car insurance):

[0232] Data collection:

[0233] Data sources include vehicle accident data, driving behavior data (such as sudden braking, speeding, etc.), vehicle maintenance records, and social media discussions about car insurance.

[0234] Data collection methods: Data is collected by calling traffic management departments and vehicle repair shops through API interfaces, crawling social media data using web crawlers, and integrating internal customer databases.

[0235] Data preprocessing: (same as health insurance).

[0236] Large-scale model analysis and mining:

[0237] Model selection and pre-training: (same as health insurance).

[0238] Data inputs include vehicle accident rates, driving behavior risk characteristics, and vehicle maintenance costs.

[0239] Analysis and mining: Extract information such as driving behavior risk patterns, vehicle accident correlation factors, and customer preferences for car insurance.

[0240] Product Design:

[0241] In accordance with business rules: Design the product coverage (such as vehicle damage, third-party liability, etc.) and premium calculation method (such as pricing based on vehicle type and driving behavior score) based on auto insurance business rules and actuarial principles.

[0242] Consider market positioning: Design differentiated product solutions for different customer groups such as new car owners, experienced drivers, and high-risk drivers.

[0243] Generate a preliminary plan: (same as health insurance).

[0244] Plan optimization and evaluation: (same as health insurance).

[0245] Figure 5 This is a schematic diagram of the structure of an insurance product design device 500 provided in an embodiment of this disclosure. Figure 5As shown, the device includes: a first processing unit 510, used to acquire static information and behavioral information of at least one user, wherein the behavioral information is obtained by feature extraction of the operations of at least one user using a first network; a second processing unit 520, used to input the static information and behavioral information into a trained insurance business reasoning module to obtain structured data, wherein the structured data is used to indicate the correspondence between static information, behavioral information, insurance industry information and risk information; and a third processing unit 530, used to generate a target insurance product based on the structured data and the generation target.

[0246] In some embodiments, the insurance product design apparatus further includes a fourth processing unit, configured to acquire multiple market research data, multiple user historical behavior data, and multiple user historical static data; perform data preprocessing on the multiple market research data, multiple user historical behavior data, and multiple user historical static data to obtain a first training dataset; perform cross-processing and partitioning on the first training dataset and the second training dataset to obtain a basic knowledge training dataset, a product design training dataset, and a market dynamic analysis training dataset, wherein the second training dataset includes multiple historical insurance industry information and multiple historical risk information; and train the initial model using the basic knowledge training dataset, the product design training dataset, and the market dynamic analysis training dataset to obtain a trained insurance business reasoning module.

[0247] In some embodiments, the fourth processing unit is further configured to: determine the data type of each training data in the basic knowledge training dataset, the product design training dataset, and the market dynamics analysis training dataset; determine the target expert network corresponding to the data type of each training data in the multi-expert hybrid system, wherein the target expert network is any one of the actuarial prediction expert network, the compliance review expert network, and the market risk assessment expert network; input each training data into the corresponding target expert network to obtain the prediction result corresponding to each training data; and adjust the initial model based on the prediction result until the adjusted initial model meets the training termination condition to obtain the trained insurance business inference module.

[0248] In some embodiments, the third processing unit is further configured to generate an initial insurance product using an insurance product generation algorithm based on the generation target and structured data; conduct compliance review and business rule verification on the initial insurance product to obtain the actuarial balance coefficient, market competitiveness index, and regulatory compliance corresponding to the initial insurance product; determine the fitness function of the initial insurance product based on the actuarial balance coefficient, market competitiveness index, and regulatory compliance; and optimize the initial insurance product based on the fitness function to obtain the target insurance product.

[0249] In some embodiments, the insurance product design apparatus further includes a fifth processing unit, configured to perform market simulation on the target insurance product based on current insurance industry information, current risk information, and current market research data, and obtain simulation results, the simulation results including at least one of market share, customer feedback, and potential risks; and adjust the target insurance product based on the simulation results.

[0250] In some embodiments, the fourth processing unit is further configured to trigger a model fine-tuning instruction when the change in market research data is detected to be greater than or equal to a preset threshold; in response to the model fine-tuning instruction, acquire updated multiple market research data, updated multiple user historical behavior data, and updated multiple user historical static data; and perform model training on the trained insurance business reasoning module based on the updated multiple market research data, updated multiple user historical behavior data, and updated multiple user historical static data to adjust the trained insurance business reasoning module.

[0251] In summary, the insurance product design device 500 can acquire user behavioral and static information to understand user needs and preferences. Then, a trained large model can be used to perform data analysis on insurance industry information, risk information, user behavioral information, and static information, extracting key features, determining the relationships between these information, and generating structured data. Based on this structured data, a target insurance product can be generated, making the generated product more aligned with market requirements, industry standards, and user needs.

[0252] The methods and apparatus provided in the embodiments of this application have been described above. To implement the functions of the methods provided in the embodiments of this application, the electronic device may include a hardware structure and software modules, and may implement the above functions in the form of a hardware structure, software modules, or a hardware structure plus software modules. One of the above functions may be executed in the form of a hardware structure, software modules, or a hardware structure plus software modules.

[0253] Figure 6 This is a block diagram illustrating an electronic device 600 for implementing the above-described method according to an exemplary embodiment. For example, the electronic device 600 may be a mobile phone, computer, messaging device, game console, tablet device, medical device, fitness equipment, personal digital assistant, etc.

[0254] Reference Figure 6The electronic device 600 may include one or more of the following components: a processing component 602, a memory 604, a power supply component 606, a multimedia component 608, an audio component 610, an input / output (I / O) interface 612, a sensor component 614, and a communication component 616.

[0255] Processing component 602 typically controls the overall operation of electronic device 600, such as operations associated with display, telephone calls, data communication, camera operation, and recording operations. Processing component 602 may include one or more processors 620 to execute instructions to perform all or part of the steps of the methods described above. Furthermore, processing component 602 may include one or more modules to facilitate interaction between processing component 602 and other components. For example, processing component 602 may include a multimedia module to facilitate interaction between multimedia component 608 and processing component 602.

[0256] Memory 604 is configured to store various types of data to support the operation of electronic device 600. Examples of this data include instructions for any application or method operating on electronic device 600, contact data, phonebook data, messages, pictures, videos, etc. Memory 604 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0257] Power supply component 606 provides power to various components of electronic device 600. Power supply component 606 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 600.

[0258] Multimedia component 608 includes a screen that provides an output interface between electronic device 600 and user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of touch or swipe actions but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 608 includes a front-facing camera and / or a rear-facing camera. When electronic device 600 is in an operating mode, such as a shooting mode or video mode, the front-facing camera and / or rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0259] Audio component 610 is configured to output and / or input audio signals. For example, audio component 610 includes a microphone (MIC) configured to receive external audio signals when electronic device 600 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 604 or transmitted via communication component 616. In some embodiments, audio component 610 also includes a speaker for outputting audio signals.

[0260] I / O interface 612 provides an interface between processing component 602 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0261] Sensor assembly 614 includes one or more sensors for providing state assessments of various aspects of electronic device 600. For example, sensor assembly 614 may detect the on / off state of electronic device 600, the relative positioning of components such as the display and keypad of electronic device 600, changes in position of electronic device 600 or a component of electronic device 600, the presence or absence of user contact with electronic device 600, orientation or acceleration / deceleration of electronic device 600, and temperature changes of electronic device 600. Sensor assembly 614 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 614 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 614 may also include an accelerometer, gyroscope, magnetometer, pressure sensor, or temperature sensor.

[0262] Communication component 616 is configured to facilitate wired or wireless communication between electronic device 600 and other devices. Electronic device 600 can access wireless networks based on communication standards, such as WiFi, 2G or 3G, 4G LTE, 5G NR (NewRadio), or combinations thereof. In one exemplary embodiment, communication component 616 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 616 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0263] In an exemplary embodiment, the electronic device 600 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0264] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 604 including instructions, which can be executed by a processor 620 of an electronic device 600 to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0265] Embodiments of this disclosure also provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the methods described in the above embodiments of this disclosure.

[0266] Figure 7 This is a schematic diagram illustrating the structure of a chip 700 for implementing the above method according to an exemplary embodiment. (Refer to...) Figure 7 The chip 700 includes a communication interface 701 and at least one processor 702. The communication interface 701 is used to receive signals input to the chip 700 or signals output from the chip 700. The processor 702 communicates with the communication interface 701 and implements the methods described in the above embodiments of this disclosure through logic circuits or executing code instructions.

[0267] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0268] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in at least one embodiment or example.

[0269] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0270] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processing module, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having at least one wiring (control method), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0271] It should be understood that various parts of the embodiments of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0272] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0273] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc.

[0274] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for designing an insurance product, characterized in that, The method includes: Acquire static information and behavioral information of at least one user, wherein the behavioral information is obtained by feature extraction of the operations of the at least one user using a first network; The static information and the behavioral information are input into a trained insurance business reasoning module to obtain structured data, which is used to indicate the correspondence between the static information, the behavioral information, the insurance industry information, and the risk information. Based on the structured data and the target, a target insurance product is generated.

2. The method according to claim 1, characterized in that, The method further includes: Acquire multiple market research data, multiple user historical behavior data, and multiple user historical static data; The multiple market research data, the multiple user historical behavior data, and the multiple user historical static data are preprocessed to obtain the first training dataset; The first training dataset and the second training dataset are cross-processed and divided to obtain a basic knowledge training dataset, a product design training dataset, and a market dynamics analysis training dataset. The second training dataset includes multiple historical insurance industry information and multiple historical risk information. The initial model is trained using the aforementioned basic knowledge training dataset, the aforementioned product design training dataset, and the aforementioned market dynamics analysis training dataset to obtain the trained insurance business reasoning module.

3. The method according to claim 2, characterized in that, The initial model is trained using the aforementioned basic knowledge training dataset, the aforementioned product design training dataset, and the aforementioned market dynamics analysis training dataset to obtain the trained insurance business reasoning module, which includes: The data types of each training data in the aforementioned basic knowledge training dataset, the aforementioned product design training dataset, and the aforementioned market dynamics analysis training dataset are based on the data type of each training data. In a multi-expert hybrid system, a target expert network is determined corresponding to the data type of each training data. The target expert network is any one of an actuarial prediction expert network, a compliance review expert network, and a market risk assessment expert network. Each training data point is input into the corresponding target expert network to obtain the prediction result for each training data point. The initial model is adjusted based on the prediction results until the adjusted initial model meets the training termination condition, thus obtaining the trained insurance business reasoning module.

4. The method according to claim 1, characterized in that, The step of generating a target insurance product based on the structured data and the target includes: Based on the generation target and the structured data, an initial insurance product is generated using an insurance product generation algorithm; The initial insurance product undergoes compliance review and business rule verification to obtain the actuarial balance coefficient, market competitiveness index, and regulatory compliance level corresponding to the initial insurance product. The fitness function of the initial insurance product is determined based on the actuarial balance coefficient, the market competitiveness index, and the regulatory compliance. The initial insurance product is optimized according to the fitness function to obtain the target insurance product.

5. The method according to claim 4, characterized in that, The method further includes: Based on current insurance industry information, current risk information, and current market research data, a market simulation is conducted on the target insurance product to obtain simulation results. The simulation results include at least one of the following: market share, customer feedback, and potential risks. The target insurance product is adjusted based on the simulation results.

6. The method according to claim 2, characterized in that, The method further includes: If the change in the market research data is detected to be greater than or equal to a preset threshold, a model fine-tuning instruction is triggered. In response to the model fine-tuning instruction, the system acquires updated market research data, updated user historical behavior data, and updated user historical static data. The trained insurance business reasoning module is trained using the updated market research data, the updated user historical behavior data, and the updated user historical static data to adjust the trained insurance business reasoning module.

7. An insurance product design device, characterized in that, The device includes: The first processing unit is configured to acquire static information and behavioral information of at least one user, wherein the behavioral information is obtained by feature extraction of the operations of the at least one user using a first network. The second processing unit is used to input the static information and the behavioral information into the trained insurance business reasoning module to obtain structured data, which is used to indicate the correspondence between the static information, the behavioral information, the insurance industry information and the risk information. The third processing unit is used to generate a target insurance product based on the structured data and the generated target.

8. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

10. A chip, characterized in that, It includes at least one processor and a communication interface; the communication interface is used to receive signals input to the chip or signals output from the chip, and the processor communicates with the communication interface and implements the method as described in any one of claims 1 to 6 through logic circuits or executing code instructions.

Images