Intelligent robot car insurance order issuing system

Abstract

The invention discloses an intelligent robot car insurance issuing system, and relates to the technical field of intelligent financial services, and the system comprises the following components: a multi-channel intelligent interaction module, an information automatic processing module, a real-time underwriting and electronic insurance policy generation module, and a data synchronization module. According to the invention, through a multi-dimensional information verification algorithm unit in an information automatic processing module, an improved D-S evidence theory is utilized, and the weight is dynamically optimized by combining machine learning, so that the accuracy verification is carried out on certificate images and text information uploaded by clients; the defect that a traditional rule engine cannot process fuzzy evidences is overcome, the accuracy of information verification in a complex scene is remarkably improved, meanwhile, the algorithm can automatically trigger supplementary information guidance, manual intervention is reduced, and the overall processing efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent financial service technology, and specifically to an intelligent robot auto insurance policy issuance system. Background Art

[0002] With the continuous growth of car ownership and the rapid development of Internet technology, auto insurance business, as an important part of the financial insurance industry, is facing unprecedented challenges and opportunities.

[0003] The traditional auto insurance policy issuance process has several significant shortcomings. First, information verification relies heavily on manual review, which is not only time-consuming and labor-intensive, but also easily affected by human factors, resulting in inaccurate verification results. This is especially true when processing large amounts of customer information. The efficiency and accuracy of manual verification are difficult to guarantee. Second, insurance recommendations are often based on fixed rules and templates, lacking personalized analysis of individual customer characteristics, making it difficult to meet the diverse needs of customers.

[0004] In view of the inaccurate information verification and lack of personalization in insurance recommendations in the traditional auto insurance policy issuance process, it is particularly important to develop an intelligent robot auto insurance policy issuance system. Summary of the Invention

[0005] The purpose of this invention is to make up for the shortcomings of the existing technology and provide an intelligent robot car insurance issuance system. It can realize the automation and intelligence of the car insurance issuance process by integrating multi-channel intelligent interaction, automatic information processing, real-time underwriting and electronic policy generation, and data synchronization modules. The system adopts advanced multi-dimensional information verification algorithms, dynamic insurance combination optimization algorithms and hierarchical underwriting decision tree technical means, which significantly improves the accuracy and efficiency of information verification, and realizes personalized insurance recommendation and efficient underwriting.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: an intelligent robot automobile insurance policy issuance system, which includes the following components: a multi-channel intelligent interaction module, an automatic information processing module, a real-time underwriting and electronic policy generation module, and a data synchronization module;

[0007] The multi-channel intelligent interaction module is used to receive customer insurance requests through multiple channels and guide information entry;

[0008] The automatic information processing module is used to verify the insurance information and match the insurance type combination;

[0009] The real-time underwriting and electronic policy generation module is used to connect to the insurance company's backend to complete underwriting and generate insurance policies;

[0010] The data synchronization module is used to ensure that payment status and policy data are synchronized in real time between the insurance company system and the cloud platform.

[0011] Furthermore, the automatic information processing module includes a multi-dimensional information verification algorithm unit for verifying the accuracy of the document image and text information uploaded by the customer. The multi-dimensional information verification algorithm is constructed based on the improved DS evidence theory. The algorithm is:

[0012]

[0013] Where C is the comprehensive score of information credibility, ranging from 0 to 100, n represents the total number of information categories involved in the information credibility assessment, ω i is the weight coefficient of the i-th category information, which is determined by historical data training and iteratively optimized by the particle swarm optimization algorithm. The initial weight matrix is ​​generated based on the Gini index to calculate the feature importance, m i (A) is the quality of positive evidence for the i-th type of information, is the quality of negative evidence, δ i It is a correlation correction factor across information dimensions. When C ≥ 85, the information is automatically verified. If it is lower than the threshold, the robot is triggered to guide the customer to supplement the information. The algorithm integrates multi-source heterogeneous evidence through evidence theory and combines machine learning to dynamically optimize weights to solve the defect that traditional rule engines cannot handle fuzzy evidence, and improve the accuracy of information verification in complex scenarios.

[0014] Furthermore, the automatic information processing module includes a dynamic insurance combination optimization algorithm unit for generating personalized insurance plans based on the customer's vehicle parameters and risk preferences. The algorithm is based on an improved non-dominated sorting genetic algorithm, and the objective function is:

[0015]

[0016] The constraints are: Coverage ≥ RT;

[0017] where λ j is the insurance cost weight, μ k For risk coverage weight, based on the loss distribution data of the insurance industry, the prior probability is determined by learning the user's historical selection behavior through the Bayesian network. j is the premium cost of the j-th insurance type, Risk k is the uncovered probability of the kth type of risk, CI is the set of required types of compulsory traffic insurance, and RT is the minimum coverage required by regulators. The algorithm introduces an elite retention strategy and an adaptive crossover mutation operator to generate a set of non-inferior solutions that take both cost and risk into account on the Pareto front for customers to choose. This algorithm breaks through the limitations of traditional rule matching, realizes the dynamic generation of insurance combinations under multi-objective optimization, and improves the adaptability to customer needs.

[0018] Furthermore, the real-time underwriting and electronic policy generation module includes a hierarchical underwriting decision tree unit for automatically completing underwriting based on multi-dimensional risk factors. The decision tree is constructed using an improved C5.0 algorithm, and the splitting criterion is:

[0019]

[0020] Among them, GainRatio(S,A) is the information gain rate of feature A for data set S, InfoGain(S,A) is the reduction of information entropy of feature A for data set S, SplitInfo(S,A) is the complexity of feature A partitioning data set S, γ(A) is the risk factor correlation correction coefficient, and the information gain correlation between factors is calculated by the mutual information algorithm to suppress redundant attributes. The risk factor set F = {VehicleAge, DrivingExperience, ClaimsFrequency, CreditScore}, and the weight of each factor is obtained through gradient boosting tree training. The initial weight is set by the insurance actuary according to the risk pricing model. The tree leaf nodes of the decision tree correspond to the underwriting results. The pruning strategy adopts cost-complexity pruning based on the minimum description length. The underwriting model solves the rigidity of traditional underwriting rules by integrating expert knowledge and data-driven, and realizes dynamic adjustment and enhanced interpretability of risk assessment.

[0021] Furthermore, the multi-channel intelligent interaction module includes a dialogue strategy optimization algorithm unit to improve the efficiency of human-computer interaction. The algorithm is built based on the deep reinforcement learning (DRL) framework. The state space S includes features such as customer information completeness, question type, and historical interaction records. The action space A includes operations such as information guidance, solution recommendation, and FAQ. The reward function is designed as follows:

[0022] R=α·CR+β·RS-γ·ID

[0023] Among them, α, β, and γ are weighting coefficients, which were optimized through A / B testing and have initial values ​​of 0.5, 0.3, and 0.2, respectively. CR is the information collection completion rate. RS is based on the customer sentiment analysis score and uses the BERT model to perform sentiment classification on the conversation text. ID is the conversation turn. The algorithm improves training stability through the experience replay mechanism and target network update strategy, achieving an interactive upgrade from passive question-and-answer to active guidance, reducing customer operating costs.

[0024] Furthermore, the data synchronization module adopts a distributed transaction consistency protocol based on blockchain to ensure the reliability of payment status synchronization;

[0025] The protocol flow is as follows:

[0026] After the customer completes the payment, the payment gateway generates an event message containing the transaction hash;

[0027] The system broadcasts the message to the insurance company blockchain node and the cloud auto insurance platform node through smart contracts;

[0028] Each node verifies the validity of the transaction through the PBFT consensus algorithm and updates the status after reaching consensus;

[0029] The status update results are fed back to the front-end interface and underwriting module through the event bus. This mechanism uses the blockchain's tamper-proof characteristics and dynamic weight consensus to solve the payment status inconsistency problem that may be caused by traditional asynchronous callbacks and ensure data integrity.

[0030] Furthermore, the electronic policy management module adopts a hybrid encryption mechanism that is resistant to quantum attacks to ensure the security of policy data;

[0031] The mechanism consists of the following steps:

[0032] Based on the SM9 identification cryptographic algorithm, the system generates a unique identification key for each customer. This key is bound to the customer's identity information, enabling identity authentication without the need for traditional certificate management.

[0033] The AES-256 symmetric encryption algorithm is used to encrypt the policy text and generate a session key. The SM2 elliptic curve algorithm is also used to re-encrypt the session key. The encryption public key comes from the insurance company's digital certificate.

[0034] Embed a timestamp generated based on the SM9 algorithm during the encryption process to ensure the timeliness of policy data and prevent replay attacks;

[0035] The final encrypted policy data consists of three parts: the AES-encrypted policy content, the SM2-encrypted session key, and the authentication tag generated by SM9;

[0036] This encryption mechanism integrates national secret algorithms and quantum-resistant cryptographic technology to meet the requirements of the National Information Security Level Protection 2.0. While ensuring data security, it supports rapid decryption and verification to meet the needs of high-frequency issuance of auto insurance policies.

[0037] Furthermore, the system also includes an online learning and performance optimization module to continuously improve the accuracy and efficiency of the system. It collects multi-source data such as customer operation logs, underwriting results, and complaints and suggestions in real time, automatically identifies and annotates abnormal cases through natural language processing technology, and builds a dynamic training data set. Based on a micro-batch learning strategy, the system regularly conducts incremental training on information verification, insurance type matching, and underwriting rule models.

[0038] Update trigger conditions include:

[0039] The deviation between the customer's operation path and the predicted path exceeds the threshold for N consecutive times (N≥5);

[0040] The amount of newly annotated data reaches more than 5% of the historical data set;

[0041] The conversion rate of specific business scenarios continues to be lower than the benchmark value.

[0042] Compared with existing technologies, this intelligent robot auto insurance policy issuance system has the following beneficial effects:

[0043] 1. The system uses the multi-dimensional information verification algorithm unit in the automatic information processing module, utilizes the improved DS evidence theory, and combines machine learning to dynamically optimize weights to verify the accuracy of document images and text information uploaded by customers. This algorithm solves the defect of traditional rule engines that cannot handle fuzzy evidence by integrating multi-source heterogeneous evidence, significantly improving the accuracy of information verification in complex scenarios. At the same time, the algorithm can automatically trigger supplementary information guidance, reducing manual intervention and improving overall processing efficiency.

[0044] 2. The system generates personalized insurance plans based on the customer's vehicle parameters and risk preferences through the dynamic insurance combination optimization algorithm unit in the automatic information processing module. The algorithm is based on an improved non-dominated sorting genetic algorithm, which generates a set of non-inferior solutions that take into account both cost and risk on the Pareto front for customers to choose from, thus breaking through the limitations of traditional rule matching and realizing the dynamic generation of insurance combinations under multi-objective optimization. At the same time, the real-time underwriting and electronic policy generation module adopts a hierarchical underwriting decision tree unit to automatically complete underwriting based on multi-dimensional risk factors, and constructs a decision tree through the improved C5.0 algorithm, which solves the problem of rigid traditional underwriting rules, realizes dynamic adjustment and enhanced interpretability of risk assessment, and improves underwriting efficiency.

[0045] Other advantages, objects and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be learned from the practice of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0047] Figure 1 Develop a flow chart for the intelligent robot car insurance billing system;

[0048] Figure 2 This is the connection diagram of the intelligent robot car insurance policy issuance system. DETAILED DESCRIPTION

[0049] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0050] Example 1

[0051] Mr. is a busy office worker who plans to purchase auto insurance for his new car. Due to his busy work schedule, he hopes to complete the auto insurance purchase process quickly through a mobile app at night to avoid being distracted during daytime working hours.

[0052] Open the mobile APP and select the "Car Insurance Purchase" service. The system interface is simple and clear, guiding him through the steps.

[0053] The system first uses voice and text prompts to guide the entry of vehicle information, including license plate number, model, purchase date, etc. You can choose to enter it manually, or you can take a photo to identify the information on the driving license, and the system will automatically fill it in, greatly improving the entry efficiency.

[0054] Next, the system guides you to enter personal information, such as ID number, contact information, etc. Similarly, the system supports the function of taking photos to identify ID cards, reducing the error rate of manual input.

[0055] After the entry is completed, the automatic information processing module immediately starts the multi-dimensional information verification algorithm unit. The system first performs image recognition on the uploaded ID card photo, extracts key information, and compares it with the text input information to ensure consistency.

[0056] Next, the system uses the multi-dimensional information verification algorithm formula for verification:

[0057]

[0058] Among them, w i is the weight coefficient of the i-th type of information, which is set according to the type and importance of information. and are the quality of positive evidence and the quality of negative evidence, respectively, which are calculated by the risk assessment model built into the system, k i It is a correlation correction factor across information dimensions, used to adjust the correlation impact between different information. After the system verification is passed, it enters the next step.

[0059] Subsequently, the dynamic insurance combination optimization algorithm unit generates a personalized insurance plan based on vehicle parameters (such as vehicle model and year of purchase) and risk preferences (such as whether long-distance driving is frequent). The algorithm objective function is: The constraints are: Coverage ≥ RT, ensuring compulsory traffic insurance is required and the total coverage amount is not lower than the regulatory requirements. Taking into account the characteristics of the new car, a combination of full insurance + no deductible is recommended, which not only ensures comprehensive protection but also takes cost-effectiveness into consideration. The customer expressed satisfaction and selected this plan.

[0060] After the proposal is submitted, the real-time underwriting and electronic policy generation module immediately activates the hierarchical underwriting decision tree unit. The system automatically completes underwriting based on multi-dimensional risk factors (such as driving record, vehicle value, age, etc.). The decision tree splitting criteria are:

[0061]

[0062] Among them, r A It is the risk factor correlation correction coefficient, which is adjusted according to the specific situation. During the underwriting process, the system also takes into account the credit record and historical claims to ensure the accuracy of the underwriting results. After the underwriting is passed, the system will immediately generate an electronic insurance policy and send it to the mobile APP.

[0063] After the payment is completed, the data synchronization module adopts a distributed transaction consistency protocol based on blockchain to ensure that the payment status and policy data are synchronized in real time between the insurance company system and the cloud platform. The system broadcasts the transaction message to the insurance company blockchain node and the cloud auto insurance platform node through the smart contract. After each node updates its status, the PBFT consensus algorithm is used to verify the validity of the transaction. After reaching a consensus, the data synchronization is completed, and the policy status and details can be viewed at any time on the mobile APP.

[0064] Example 2

[0065] A woman came to an auto insurance store hoping to purchase auto insurance for her used car. The store staff used an intelligent robot auto insurance policy issuance system to assist her in completing the insurance process and provide more personalized and professional services.

[0066] After the lady arrived at the store, the staff warmly welcomed her and guided her to the self-service terminal. The terminal had a user-friendly interface and was easy to operate. The staff first helped the lady enter the vehicle information, including license plate number, model, purchase date, mileage, etc. Since it was a second-hand car, the staff also specifically asked about the vehicle's maintenance records and accident conditions in order to more accurately assess the risks.

[0067] Next, the staff guided the lady to enter her personal information, including her ID number, contact information, etc. The lady could choose to enter it manually or automatically fill in the information by scanning her ID card, which improved the entry efficiency.

[0068] After the entry is completed, the system verifies the driving license and ID card information uploaded by the woman through a multi-dimensional information verification algorithm unit. The system first performs image recognition, extracts key information, and compares it with the text input information to ensure consistency. At the same time, the system also checks the validity period of the driving license and the annual inspection status of the vehicle to ensure that the vehicle meets the insurance conditions.

[0069] Subsequently, the dynamic insurance combination optimization algorithm unit recommended a combination of basic insurance + third-party liability insurance based on the vehicle condition (such as age, mileage, maintenance record) and the woman's budget. The staff explained to the woman in detail the coverage and costs of each type of insurance to help her better understand and make a choice.

[0070] After the proposal was submitted, the system activated the hierarchical underwriting decision tree unit. Taking into account the possible historical risks of used cars, the underwriting process was more detailed. The system not only checked the woman's driving record and credit status, but also considered the vehicle's maintenance history and accident records to ensure the accuracy of the underwriting results. After the underwriting was passed, the system immediately generated an electronic insurance policy and printed out a paper policy for the woman to sign and confirm.

[0071] The data synchronization module ensures that the policy data is synchronized to the insurance company system in real time. The staff handed the paper policy to the lady and explained the policy content in detail, including the coverage, exemption clauses, etc. The lady was satisfied with the service and decided to purchase the car insurance product. At the same time, the system also provides the function of online viewing and management of the policy, so that the lady can understand the policy status at any time.

[0072] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as above in terms of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can, without departing from the scope of the technical solution of the present invention, make some changes or modifications to equivalent embodiments using the technical contents disclosed above. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. Intelligent robot car insurance billing system, characterized by: The system includes the following components: multi-channel intelligent interaction module, automatic information processing module, real-time underwriting and electronic policy generation module, and data synchronization module; The multi-channel intelligent interaction module is used to receive customer insurance requests through multiple channels and guide information entry; The automatic information processing module is used to verify the insurance information and match the insurance type combination; The real-time underwriting and electronic policy generation module is used to connect to the insurance company's backend to complete underwriting and generate insurance policies; The data synchronization module is used to ensure that payment status and policy data are synchronized in real time between the insurance company system and the cloud platform.

2. The intelligent robot car insurance billing system according to claim 1 is characterized in that: The automatic information processing module includes a multi-dimensional information verification algorithm unit, which is used to verify the accuracy of the document image and text information uploaded by the customer. The multi-dimensional information verification algorithm is constructed based on the improved DS evidence theory. The algorithm is: Where C is the comprehensive score of information credibility, n represents the total number of information categories involved in information credibility assessment, ω i is the weight coefficient of the i-th type of information, m i (A) is the quality of positive evidence for the i-th type of information, is the quality of negative evidence, δ i It is the correlation correction factor across information dimensions. When C≥85, the information automatically passes the verification. If it is lower than the threshold, the robot is triggered to guide the customer to supplement the information.

3. The intelligent robot car insurance billing system according to claim 1 is characterized in that: The automatic information processing module includes a dynamic insurance combination optimization algorithm unit, which is used to generate personalized insurance plans based on the customer's vehicle parameters and risk preferences. The algorithm is based on an improved non-dominated sorting genetic algorithm, and the objective function is: The constraints are: where λ j is the insurance cost weight, μ k is the risk coverage weight, Cost j is the premium cost of the j-th insurance type, Risk k is the uncovered probability of the kth type of risk, CI is the set of required types of compulsory traffic insurance, and RT is the minimum coverage required by regulators. The algorithm introduces an elite retention strategy and an adaptive crossover mutation operator to generate a set of non-inferior solutions that take both cost and risk into account in the Pareto front for customers to choose.

4. The intelligent robot car insurance billing system according to claim 1 is characterized in that: The real-time underwriting and electronic policy generation module includes a hierarchical underwriting decision tree unit, which is used to automatically complete underwriting based on multi-dimensional risk factors. The decision tree is constructed using the improved C5.0 algorithm, and the splitting criterion is: Where GainRatio(S,A) is the information gain rate of feature A for dataset S, InfoGain(S,A) is the reduction of information entropy of feature A for dataset S, SplitInfo(S,A) is the complexity of feature A partitioning dataset S, γ(A) is the risk factor correlation correction coefficient, and the risk factor set F = {VehicleAge, DrivingExperience, ClaimsFrequency, CreditScore}. The weights of each factor are obtained through gradient boosting tree training. The initial weights are set by insurance actuaries according to the risk pricing model. The tree leaf nodes of the decision tree correspond to the underwriting results. The pruning strategy adopts cost-complexity pruning based on the minimum description length.

5. The intelligent robot car insurance billing system according to claim 1 is characterized in that: The multi-channel intelligent interaction module includes a dialogue strategy optimization algorithm unit to improve the efficiency of human-computer interaction. The algorithm is built based on a deep reinforcement learning framework. The state space S includes customer information integrity, question type, and historical interaction record features. The action space A includes information guidance, solution recommendation, and FAQ operations. The reward function is designed as follows: R=α·CR+β·RS-γ·ID Among them, α, β, and γ are weighting coefficients, CR is the information collection completion rate, RS is based on the customer sentiment analysis score, and ID is the conversation turn.

6. The intelligent robot car insurance billing system according to claim 1 is characterized in that: The data synchronization module adopts a distributed transaction consistency protocol based on blockchain; The protocol flow is as follows: After the customer completes the payment, the payment gateway generates an event message containing the transaction hash; The system broadcasts the message to the insurance company blockchain node and the cloud auto insurance platform node through smart contracts; Each node verifies the validity of the transaction through the PBFT consensus algorithm and updates the status after reaching consensus; The status update results are fed back to the front-end interface and underwriting module through the event bus.

7. The intelligent robot car insurance billing system according to claim 1 is characterized in that: The electronic policy management module adopts a hybrid encryption mechanism that is resistant to quantum attacks; The mechanism consists of the following steps: Based on the SM9 identification cryptographic algorithm, the system generates a unique identification key for each customer. This key is bound to the customer's identity information, enabling identity authentication without the need for traditional certificate management. The AES-256 symmetric encryption algorithm is used to encrypt the policy text and generate a session key. The SM2 elliptic curve algorithm is also used to re-encrypt the session key. The encryption public key comes from the insurance company's digital certificate. Embed a timestamp generated based on the SM9 algorithm during the encryption process; The final encrypted policy data consists of three parts: AES-encrypted policy content, SM2-encrypted session key, and SM9-generated authentication tag.

8. The intelligent robot car insurance billing system according to claim 1 is characterized in that: The system also includes an online learning and performance optimization module, which collects multi-source data such as customer operation logs, underwriting results, and complaints and suggestions in real time, automatically identifies and annotates abnormal cases through natural language processing technology, and constructs a dynamic training data set. Based on a micro-batch learning strategy, the system regularly conducts incremental training on information verification, insurance type matching, and underwriting rule models. Update trigger conditions include: The deviation between the customer's operation path and the predicted path exceeds the threshold for N consecutive times (N≥5); The amount of newly annotated data reaches more than 5% of the historical data set; The conversion rate of specific business scenarios continues to be lower than the benchmark value.

Images