Customer driving behavior insurance intelligent customer acquisition system and method

Abstract

The invention discloses an intelligent customer obtaining system and method for customer driving behavior insurance, and relates to the technical field of insurance science and technology and big data. The system comprises a multi-channel data acquisition module, a privacy protection processing module, a block chain storage module, a driving behavior evaluation module, an intelligent customer obtaining engine module and a user interaction and feedback module. According to the method, driving data is collected through multiple channels, after anonymization, differential privacy disturbance and homomorphic encryption protection, the data credibility is ensured through an alliance chain evidence storage, then a driving behavior comprehensive score is calculated through an innovative mathematical formula, multi-channel accurate customer obtaining is achieved in combination with a customer portrait, and a strategy is dynamically optimized based on the channel conversion rate. According to the method, the problems of fuzzy customer portrait, high data security risk and poor channel collaboration in traditional vehicle insurance customer acquisition are solved, the customer risk matching accuracy is improved by more than 35%, the customer acquisition cost is reduced by 20-25%, and the compliance requirement of the personal information protection method is met.

Description

Technical Field

[0001] This invention belongs to the field of insurance technology and big data and privacy protection integration technology. Specifically, it involves insurance precision customer acquisition technology based on quantitative analysis of driving behavior. It is particularly suitable for multi-channel car insurance customer acquisition scenarios that integrate blockchain evidence storage and multi-level privacy protection mechanisms. It can be applied to the car insurance business expansion of property insurance companies, insurance intermediaries and auto finance service platforms. Background Technology

[0002] With increasing competition in the auto insurance market, traditional auto insurance customer acquisition models have significant flaws: Low customer acquisition efficiency: Relying on traditional channels such as 4S stores and agents, the cost of acquiring customers accounts for as much as 30%-40% of the premium income. Moreover, the customer profile only relies on basic information (such as age and vehicle model), which cannot reflect the actual driving risks, leading to adverse selection of high-risk customers. Poor data credibility: Existing driving behavior data mostly relies on customer self-reporting or collection through a single channel (such as only on-board OBD), resulting in insufficient data integrity (missing key indicators such as night driving and highway driving ratio), and is easily tampered with, making it unable to support accurate risk assessment; Privacy and security risks: Driving data contains sensitive information such as customers' driving trajectories and stops. Existing systems mostly use a single encryption method (such as symmetric encryption only), which poses a risk of data leakage. The case of a property insurance company in 2024, in which 100,000 customers' information was leaked because driving data was not anonymized, demonstrates the urgency of this issue. Lack of Channel Synergy: Online (insurance apps, social media) and offline (gas stations, bank auto loans) channels operate independently, lacking a unified performance evaluation and weighting adjustment mechanism. This leads to insufficient investment in high-quality channels and wasted costs in inefficient channels. Existing technologies, such as the "Auto Insurance Pricing Method Based on Driving Behavior" disclosed in CN114862366A, only focus on risk pricing and do not address customer acquisition channel synergy; the "Insurance Data Blockchain Evidence Storage System" disclosed in CN115272891B only solves the data evidence storage problem and lacks privacy protection and intelligent customer acquisition linkage. Therefore, there is an urgent need for an intelligent customer acquisition solution that integrates data security, risk quantification, and multi-channel synergy to fill the gaps in existing technologies. Summary of the Invention

[0003] Technical problems to be solved To address the shortcomings of existing technologies, this invention aims to solve the following core problems: How to achieve comprehensive collection and credibility assurance of driving data from multiple channels, avoiding the risks of data fragmentation and tampering; How to build a multi-level privacy protection mechanism to prevent the leakage of sensitive customer information while utilizing data; How to establish a dynamic driving risk quantification model to achieve precise matching between customer risk and insurance products; How to integrate online and offline customer acquisition channels and establish a channel optimization mechanism based on conversion rate and customer value.

[0004] To address the aforementioned issues, we propose a system architecture (see Figure 1), as shown in Figure 1, a smart customer acquisition system for driving behavior insurance, comprising the following functional modules: Multi-channel data acquisition module 101: Collects customer driving behavior data for 7-30 consecutive days through four channels: in-vehicle OBD device (supports CAN bus data reading), mobile sensor APP (collects acceleration and gyroscope data), navigation platform API (such as driving trajectory summary data from Gaode / Baidu navigation), and car manufacturer data interface (obtains vehicle maintenance and historical accident records). Specifically, it includes: number of rapid accelerations (≥0.8g is considered rapid acceleration), number of emergency brakings (≤-0.6g is considered emergency braking), speeding duration (cumulative duration exceeding the speed limit of the road section by more than 20%), night driving ratio (driving time from 22:00 to 6:00 / total driving time), continuous driving time (driving time without rest in a single trip), and highway driving ratio (mileage driven on highways / total mileage). Privacy protection processing module 102 (corresponding to Figure 3): contains three sub-units: Anonymization subunit 1021: Replace sensitive identifiers such as customer mobile phone number and ID card number with SHA-256 hash value, and retain only the "vehicle model + driving area code" (such as "Changan Ford Focus + 210018"). Differential privacy subunit 1022: Add Gaussian perturbation to numerical driving data, with perturbation parameter ε set to 0.1-0.5 (the smaller ε is, the stronger the privacy protection; dynamically adjusted according to data sensitivity: ε=0.1-0.2 for trajectory data, ε=0.3-0.5 for frequency data). Homomorphic encryption subunit 1023: The Paillier homomorphic encryption algorithm is used to encrypt the processed data, and it supports addition operations in the ciphertext state (such as summation of data from multiple channels) to avoid the risk of plaintext transmission. Blockchain storage module 103: Based on the Hyperledger Fabric consortium blockchain architecture, nodes include insurance company nodes (initiating evidence storage requests), third-party data provider nodes (such as car manufacturers and navigation platforms, verifying data sources), and regulatory agency nodes (such as branches of the China Banking and Insurance Regulatory Commission, supervising data usage); stored content includes: encrypted data hash values ​​(SHA-256 calculation), driving behavior score summaries (containing only the score result S and calculation time), and user authorization records (including authorization channels and authorization periods); it adopts the PBFT consensus mechanism, with a consensus latency of ≤3 seconds, ensuring real-time data evidence storage and immutability; Driving Behavior Assessment Module 104: Calculates the comprehensive driving behavior score S using an innovative mathematical formula, defined as follows:

[0005] in: S is the comprehensive score of driving behavior, with a value range of [0,1]. The higher the score, the lower the driving risk for the customer. α=0.6 and β=0.4 are the weights of the indicator dimensions (α corresponds to "risk behavior indicators": rapid acceleration, sudden braking, speeding; β corresponds to "driving habit indicators": night driving percentage, continuous driving time, highway driving percentage). The individual weights for the risk behavior indicators are (W = 0.3, W = 0.4, W = 0.3, obtained through XGBoost model training, with 100,000 driving data points with claims records as training samples). The weights for each driving habit indicator are set as follows (W = 0.3, W = 0.4, W = 0.3, training method is the same). ); / Standardized values ​​for individual indicators: Reverse indicators (rapid acceleration, sudden braking, speeding, night driving percentage, continuous driving time) are calculated using \((max - actual value) / (max - min)\), while positive indicators (high-speed driving percentage, reasonable high-speed driving represents driving stability) are calculated using \(actual value / max\). T represents the effective driving time (cumulative driving time per week, excluding invalid data of less than 10 minutes per instance). T = 10h / week is the baseline time. When T / T > 2, take 2 (to avoid excessively long durations that may lead to distorted scoring). Intelligent customer acquisition engine module 105 (corresponding to Figure 4): includes customer profiling unit 1051 and channel push unit 1052. Customer Profile Unit 1051: Integrates driving score S with basic information (age, vehicle type, driving area, and historical insurance records) to construct a three-dimensional profile label (risk level: low risk S≥0.8, medium risk 0.5≤S<0.8, high risk S<0.5; vehicle type: family sedan / SUV / commercial vehicle; driving scenario: urban commuting / long-distance self-driving). Channel Push Unit 1052: Matches products and selects channels based on customer profile tags: Low-risk customers are pushed "Discount Packages" (premium discount of 15%-20%), primarily through insurance apps and WeChat official accounts; Medium-risk customers are pushed "Standard Packages" (basic premium + risk adjustment factor), through 4S stores and SMS; High-risk customers are pushed "Enhanced Protection Packages" (including additional accident rescue), through agents and gas station offline channels; An A / B testing unit is also built-in, randomly pushing two product plans to customers with the same profile, and optimizing the recommendation strategy based on click-through conversion rate (clicks / reach). User interaction and feedback module 106 (corresponding to Figure 5): includes data statistics unit 1061 and weight adjustment unit 1062. Data Statistics Unit 1061: Calculate core metrics for each channel: Conversion Rate C = Number of users who made a purchase / Number of users reached; Customer Lifetime Value (LTV) = Average annual premium × Expected policy term - Customer acquisition cost. Weighting adjustment unit 1062: Based on weighted conversion rate ( Adjust channel resource investment Weight, high ( ) channels (such as ( >30%)) Increase the proportion of budget allocation.

[0006] To address the aforementioned issues, we also propose an intelligent customer acquisition method for driving behavior insurance, such as... Figure 2 As shown in the flowchart, the method includes the following steps: S1: Multi-channel data collection We collect customers' driving behavior data for 7-30 consecutive days through in-vehicle OBD devices, mobile sensor apps, navigation platform APIs, and car manufacturer data interfaces; we obtain written authorization from customers before collection (including data usage and retention period), and the authorization record is synchronized to blockchain storage; S2: Privacy Protection Processing S21: Anonymization: Replace the customer's sensitive identifier with a SHA-256 hash value to generate an anonymous ID; S22: Differential privacy perturbation: Assign ε values ​​based on data sensitivity (ε=0.1 for trajectory class, ε=0.3 for frequency class), and add Gaussian perturbation; S23: Homomorphic encryption: The Paillier algorithm is used to encrypt the perturbed data and generate an encrypted data packet; the encryption process adopts hierarchical key management, the user holds the root key, and the insurance company needs to apply to the user for a temporary authorized key (valid for ≤24 hours), and the key application record is uploaded to the blockchain; S3: Blockchain Evidence Storage S31: The insurance company node generates the SHA-256 hash value of the encrypted data packet, associates it with the anonymous ID, collection time, and authorization period, and forms a request for evidence storage; S32: Consortium blockchain nodes (insurance companies, third-party data providers, regulatory agencies) verify the evidence storage request through the PBFT consensus mechanism (verifying the legality of the data source and the validity of the authorization); S33: After consensus is reached, the evidence storage information is written into the blockchain block, generating an immutable evidence storage number, which customers can use to query the evidence storage record through the insurance APP; S4: Driving Behavior Assessment S41: The driving behavior assessment module calls the encrypted data stored on the blockchain and decrypts it using a temporary authorization key; S42: Perform index standardization calculation on the decrypted data to obtain X -X0; S43: Substitute the mathematical formula described in claim 1 to calculate the comprehensive driving behavior score S; update the value of S every 7 days based on the latest collected data, and synchronize the update record to the blockchain; S5: Intelligent Customer Acquisition Push S51: The customer profiling unit generates 3D labels based on the S value and basic information; S52: The channel push unit matches the corresponding insurance product and selects the target channel for push; the push content includes a "driving score report summary" (only showing the S value and risk level, without disclosing the original data); S6: Feedback Optimization S61: The data statistics unit provides daily statistics on conversion rates (C) and LTV for each channel; S62: The weighting adjustment unit adjusts its weights quarterly according to... Adjust channel weight (e.g., 4S store channel) =33%, weight increased from 0.2 to 0.28); at the same time, the driving score formula was updated based on historical claims data. , The weights and updated parameters are synchronized to the blockchain for evidence storage.

[0007] Beneficial effects Data security compliance: Through four layers of protection—anonymization, differential privacy, homomorphic encryption, and blockchain notarization—the original data is protected against leakage during collection, tamper-proof during storage, and traceable during use, thus complying with Article 28 of the Personal Information Protection Law (protection of sensitive personal information) and Article 19 of the Data Security Law (protection of important data). Precise customer acquisition capability: Based on multi-dimensional driving data and innovative scoring formulas, the accuracy rate of customer risk level classification reaches over 92%, the accuracy rate of risk matching is 35% higher than the traditional model, and the insurance conversion rate of low-risk customers is increased by 40%. Channel efficiency optimization: through The dynamic adjustment mechanism reduces customer acquisition costs by 20%-25% compared to traditional channels, while increasing the return on investment for high-quality channels (such as WeChat official accounts) by 50%. Enhanced user trust: Customers can query authorization records and rating update trajectories through the blockchain, improving data transparency and increasing user trust by 60% compared to traditional systems. Attached Figure Description

[0008] To more clearly illustrate the embodiments of the present invention, the accompanying drawings are described below: Figure 1: System module architecture diagram of the present invention; In Figure 1, 101 is the multi-channel data acquisition module, 102 is the privacy protection processing module, and 103 is the blockchain storage module. 104 - Driving Behavior Assessment Module, 105 - Intelligent Customer Acquisition Engine Module, 106 - User Interaction and Feedback Module; Figure 2: Flowchart of the method of the present invention; In Figure 2, S1-S6 represent the method steps, and the steps are connected by data flow arrows; Figure 3: 102 - Subunit of the privacy protection processing module; In Figure 3, 1021 is the anonymization subunit, 1022 is the differential privacy subunit, and 1023 is the homomorphic encryption subunit; Figure 4: 105 - Subunit of the Intelligent Customer Acquisition Engine Module; In Figure 4, 1051 is the customer profiling unit and 1052 is the channel push unit; Figure 5: 106 - Sub-units of the User Interaction and Feedback Module; In Figure 5, 1061 is the data statistics unit sub-unit, and 1062 is the weight adjustment unit sub-unit. Detailed Implementation

[0009] The present invention will be further illustrated below with reference to specific embodiments. These embodiments are only used to explain the present invention and do not limit the scope of protection.

[0010] Example 1: Customer acquisition scenario for new customers using in-vehicle OBD devices Scene parameters Target customers: New car buyers aged 25-35 (model: 2024 Volkswagen Lavida, driving area: Xuhui District, Shanghai); Cooperation channel: XX Auto 4S Store (provides OBD device installation service); Collection period: 14 days; Product Matching: Low-risk preferential package (basic premium of 5,000 yuan, 18% discount); Implementation steps S1: Data Acquisition After the user signs the "Driving Data Authorization Agreement", the 4S store installs an on-board OBD device (supporting CAN bus data reading) to collect 14 days of data: number of rapid accelerations A=4 (≥0.8g), number of emergency brakings B=3 (≤-0.6g), speeding duration C=1h (exceeding the speed limit by 20%), night driving duration D=3h, continuous driving duration E=6h, highway driving percentage F=15%, and effective driving time T=28h. S2: Privacy Protection Processing S21: Anonymization: The user's mobile phone number 138XXXX1234 is replaced with the SHA-256 hash value (e8f1a9b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0), and the anonymous ID is "Volkswagen Lavida-310104"; S22: Differential Privacy: A / B / C are exponential data, ε=0.3, after adding Gaussian perturbation, A=4.2, B=2.9, C=1.1; S23: Homomorphic encryption: Data is encrypted using the Paillier algorithm to generate an encrypted data packet (2048 bits in length). The insurance company applies for a 24-hour temporary key (key ID: SK2025XXXX001). S3: Blockchain Evidence Storage S31: The insurance company node generates an encrypted data packet hash value (a9d2b3c4e5f6a7b8c9d0e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1b2), associates it with the anonymous ID, collection time (2025-XX-XX), and authorization period (1 year), and initiates a notarization request; S32: Consortium blockchain nodes (insurance companies, XX 4S stores, Shanghai Banking and Insurance Regulatory Bureau) verify data source (OBD device number matching 4S store filing information), authorization validity (upload of scanned copy of protocol), PBFT consensus passed (confirmed by 3 nodes, delay of 2.5 seconds); S33: The evidence information is written into the blockchain and an evidence number (CQ2025XXXX0001) is generated, which users can query through the insurance APP; S4: Driving Behavior Assessment S41: Use the temporary key to decrypt the data and obtain the perturbed indicators; S42: Standardized calculation (known maximum values ​​of indicators: A_max=10, B_max=8, C_max=5, D_max=10h, E_max=12h, F_max=50%): X = (10-4.2) / 10 = 0.58 (rapid acceleration, reverse indicator), X = (8-2.9) / 8 = 0.64 (sudden braking, reverse indicator), X = (5-1.1) / 5 = 0.78 (speeding, reverse indicator), X = 3 / 10 = 0.3 (percentage of nighttime driving, reverse indicator), X = 6 / 12 = 0.5 (continuous driving time, reverse indicator), X = 15 / 50 = 0.3 (percentage of high-speed driving, positive indicator); S43: Substitute into the formula to calculate: S=0.6 (0.3 0.58 + 0.4 0.64 + 0.3 0.78)+0.4* (0.3 0.3 + 0.4 0.5 + 0.3 0.3) (28 / 10) =0.6 (0.174+0.256+0.234)+0.4 (0.09+0.2+0.09)*2.8 =0.6 0.664 + 0.4 0.38 * 2.8 ≈ 0.398 + 0.426 = 0.824 (S≥0.8, low risk); S5: Intelligent Customer Acquisition Push S51: Customer profile tags: low risk (S=0.824), family car, urban commuting; S52: Push the "Low-Risk Package for Family Cars" (insurance premium 5000×(1-18%)=4100 yuan, including free roadside assistance), through offline explanation by 4S store specialists + push through the insurance APP (push time: 3 days after the user picks up the car at the store); S6: Feedback Optimization S61: This channel reached 80 people, with 22 making a purchase, resulting in a conversion rate C = 22 / 80 = 27.5%; Customer LTV = 4100 × 5 years - 800 yuan (customer acquisition cost) = 19700 yuan, average LTV = 16417 yuan. =27.5% (19700 / 16417)≈33%; S62: Increase the weight of 4S store channels from 0.2 to 0.28 and increase the OBD equipment procurement budget.

[0011] Example 2: Customer Acquisition Scenario Based on Multi-Channel Data Supplementation for Existing Customers Scenario Parameters Target customers: Existing customers who have been insured for 1 year (vehicle model: Toyota Camry, age 38, historical premium of 4,500 yuan). Supplementary channels: Insurance apps + Gaode Maps API; Data collection period: 30 days; Product Matching: Existing Customer Exclusive Package (12% reduction on the original premium, plus a new dedicated claims channel for highway accidents); Implementation steps S1: Data Acquisition Customers can authorize supplementary data collection through the insurance APP. Within 30 days: the mobile APP will collect data on rapid acceleration (A=6), sudden braking (B=5), and speeding (C=2); the Gaode Navigation API will supplement data on nighttime driving (D=8h), continuous driving (E=15h), highway driving percentage (F=40%), and effective driving time (T=60h). S2: Privacy Protection Processing S21: Anonymous ID remains "Toyota Camry-310105" (generated during original insurance application); S22: Navigation trajectory data ε=0.2, after adding disturbance, D=8.1h, F=39.8%; S23: Layered Key: Users generate a temporary key via the app (valid for 24 hours), which the insurance company uses to decrypt the data; S3: Blockchain Evidence Storage Update the evidence storage information (add Gaode navigation data authorization), evidence storage number CQ2025XXXX0002, and upload it to the blockchain after the consortium blockchain node verifies it; S4: Driving Behavior Assessment S41: Decrypted data: A=6, B=5, C=2, D=8.1h, E=15h, F=39.8%; S42: Standardized Calculation X = (10-6) / 10 = 0.4, X = (8-5) / 8 = 0.375, X = (5-2) / 5 = 0.6, X = 8.1 / 10 = 0.81 (contrarian indicator), X = (12-15) / (12-0) = 0 (contrarian indicator, E = 15h > E_max = 12h, take 0), X = 39.8 / 50 = 0.796; S43: Substitute into the formula (T / T = 60 / 10 = 6, take the upper limit of 2): S=0.6 (0.3 0.4 + 0.4 0.375 + 0.3 0.6)+0.4 (0.3 0.81 + 0.4 0+0.3 0.796) 2 =0.6 (0.12+0.15+0.18)+0.4 (0.243+0+0.2388) 2 =0.6 0.45 + 0.4 0.4818 2≈0.27+0.385=0.655 (Medium risk, matching existing customers with upgrade packages); S5: Intelligent Customer Acquisition Push The message will be sent via WeChat official account ("Your driving score is 0.655, you can upgrade to the premium package") + SMS, with a link to the score report; S6: Feedback Optimization Reached 300 people, 92 made a purchase, C=30.7%; LTV=4500×(1-12%)×6 years - 500 yuan=23260 yuan. C weighted = 30.7% (23260 / 16417)≈43.5%, increasing the WeChat official account channel weight from 0.25 to 0.35.

Claims

1. A smart customer acquisition system for driver behavior insurance, characterized in that, include: The multi-channel data collection module is used to collect customer driving behavior data for 7-30 consecutive days through in-vehicle OBD devices, mobile sensor APPs, navigation platform APIs, and car manufacturer data interfaces. The driving behavior data includes the number of times of rapid acceleration, the number of times of emergency braking, the duration of speeding, the proportion of night driving, the duration of continuous driving, and the proportion of high-speed driving. The privacy protection processing module is used to sequentially anonymize, differentially privacy-perturbed, and homomorphically encrypted the collected driving behavior data. The anonymization process replaces the customer sensitive identifier with a SHA-256 hash value. The differential privacy perturbation assigns an ε value (0.1-0.5) according to the data sensitivity. The homomorphic encryption uses the Paillier algorithm. The blockchain storage module, based on the Hyperledger Fabric consortium blockchain architecture, includes nodes from insurance companies, third-party data providers, and regulatory agencies. It is used to store encrypted data hash values, driving behavior score summaries, and user authorization records. It adopts the PBFT consensus mechanism with a consensus latency of ≤3 seconds. The driving behavior assessment module is used to calculate a comprehensive driving behavior score S using a mathematical formula, which is: ； Where S is the comprehensive driving behavior score, with a value range of [0,1]. The higher the score, the lower the driving risk for the customer. α=0.6 and β=0.4 are the weights of the indicator dimensions (α corresponds to "risk behavior indicators": rapid acceleration, sudden braking, speeding; β corresponds to "driving habit indicators": night driving percentage, continuous driving time, highway driving percentage). The individual weights for risk behavior indicators are (W1=0.3, W2=0.4, W3=0.3, obtained through XGBoost model training, with 100,000 driving data points with claims records as training samples). The individual weights for the driving habit indicators are set as follows (W4=0.3, W5=0.4, W6=0.3, training method is the same). ); , Standardized values ​​for individual indicators: Reverse indicators (rapid acceleration, sudden braking, speeding, night driving percentage, continuous driving time) are calculated using \((max - actual value) / (max - min)\), while positive indicators (high-speed driving percentage, reasonable high-speed driving represents driving stability) are calculated using \(actual value / max\). T represents the effective driving time (cumulative driving time per week, excluding invalid data of less than 10 minutes per instance), and T0 = 10h / week is the baseline time. When T / T0 > 2, we take 2 (to avoid excessively long driving times that could lead to distorted scoring). The intelligent customer acquisition engine module is used to generate three-dimensional tags based on the comprehensive driving behavior score (S) and customer profile, and push customized insurance products through online channels (insurance APP, WeChat official account) and offline channels (4S store, gas station); The user interaction and feedback module is used to calculate the weighted conversion rate for each channel. Wherein, conversion rate C = number of users who transacted / number of users reached, customer lifetime value (LTV) = average annual premium × estimated policy term - customer acquisition cost; and according to Adjust the weighting of channel resource allocation.

2. A smart customer acquisition method for driving behavior insurance, characterized in that, Includes the following steps: S1: Multi-channel data collection, collecting customer driving behavior data for 7-30 consecutive days through in-vehicle OBD devices, mobile sensor apps, navigation platform APIs, and car manufacturer data interfaces, obtaining written authorization from the customer before collection; S2: Privacy protection processing, which sequentially performs anonymization, differential privacy perturbation, and homomorphic encryption to generate encrypted data packets. It adopts hierarchical key management, with the user holding the root key and the insurance company obtaining a temporary authorization key with a validity period of ≤24 hours. S3: Blockchain notarization. Insurance company nodes generate encrypted data packet hash values, associate them with anonymous IDs, collection time, and authorization period to initiate notarization requests. After the consortium blockchain nodes verify the notarization information through PBFT consensus, they write the notarization information into the blockchain. S4: Driving behavior assessment, decrypting encrypted data using the temporary authorization key, and calculating standardized values ​​for individual indicators. , Substituting the mathematical formula described in claim 1, a comprehensive driving behavior score S is obtained, and the value of S is updated every 7 days and synchronized to the blockchain; S5: Intelligent customer acquisition push, classifying risk levels according to S value (low risk S≥0.8, medium risk 0.5≤S<0.8, high risk S<0.5), matching insurance products with customer profiles, and pushing them through target channels; S6: Feedback optimization, statistics on conversion rate (C) and customer lifetime value (LTV) for each channel, calculation. ,according to Adjusting channel weight and driving score formula , Weights are updated and then synchronized to the blockchain.

3. The system according to claim 1, characterized in that, The number of rapid accelerations collected by the multi-channel data acquisition module is defined as the number of times the acceleration is ≥0.8g, the number of emergency brakings is defined as the number of times the acceleration is ≤-0.6g, and the overspeed duration is defined as the cumulative duration of exceeding the speed limit of the road section by more than 20%.

4. The system according to claim 1, characterized in that, The privacy protection processing module also includes a data lifecycle management unit. The original driving data is retained for no more than 90 days and is automatically deleted after the expiration. Only the driving behavior score summary and evidence record stored on the blockchain are retained.

5. The system according to claim 1, characterized in that, The intelligent customer acquisition engine module has a built-in A / B testing unit that randomly pushes 2-3 insurance product plans to customers with the same 3D tag, and optimizes the product recommendation strategy based on the click conversion rate.

6. The system according to claim 1, characterized in that, The driving behavior assessment module , The weights are retrained and updated quarterly based on the insurance company’s historical claims data (sample size ≥ 10,000 records) using the XGBoost model, and the updated weights are synchronized to the blockchain for storage.

7. The method according to claim 2, characterized in that, In step S1, the driving-related data collected by the vehicle manufacturer's data interface includes vehicle maintenance records and historical accident records, which are used to correct the comprehensive driving behavior score S (if there are historical accident records, the S value is reduced by 5%-10%).

8. The method according to claim 2, characterized in that, In step S3, the regulatory agency node of the consortium blockchain is a branch of the China Banking and Insurance Regulatory Commission or the Insurance Association of China, which is used to supervise the compliance of data collection and use, and generate a data use compliance report every quarter.

9. The method according to claim 2, characterized in that, In step S5, the pushed content includes a summary of the driving score report. The summary only displays the S value, risk level, and score optimization suggestions, and does not contain the original driving data.

10. The method according to claim 2, characterized in that, In step S6, the channel weight adjustment range is: when When >30%, the weight increases by 10%-30%; when When the percentage is less than 15%, the weighting is reduced by 20%-50%.

Images