Electric vehicle survey task intelligent management method based on multi-dimensional data collaboration

Abstract

The invention provides an electric vehicle survey task intelligent management method based on multi-dimensional data collaboration. Belongs to the technical field of task management. A surveyor submits information to the system, and a franchise maintenance shop pushes a preliminary maintenance demand according to the system; constructing an automatic quotation model; performing peripheral price comparison to generate a quotation ranking; an optimal franchising maintenance shop recommendation is automatically generated; the system detects an abnormal cooperative behavior through a historical quotation fluctuation rate, and identifies and prevents a collusion quotation behavior between franchise maintenance stores; for the franchise maintenance shop with the suspected collusion quotation, the system marks the franchise maintenance shop with the suspected collusion quotation as a high risk, and focuses on follow-up quotation; and after the maintenance is completed, the maintenance result is accepted, and order settlement is completed after the maintenance result is confirmed to be correct. And a floating threshold is dynamically adjusted by adopting a sliding window mechanism and Monte Carlo simulation, so that false screening caused by a static threshold is avoided, and the rationality of quotation is improved.

Description

Technical Field

[0001] The present invention proposes an intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration, belonging to the technical field of task management. Background Art

[0002] In the field of electric vehicle repair services, the traditional survey task management method has many deficiencies. On the one hand, the allocation of surveyors and affiliated repair shops often relies on manual operations, which is inefficient and error-prone; and it will cause some surveyors to be overburdened, while some surveyors are idle, resulting in a waste of resources. On the other hand, the quotation process among affiliated repair shops lacks transparency and fairness, and situations such as over-quoting, under-quoting, or unreasonable quoting often occur, resulting in damage to the interests of consumers, and at the same time affecting the overall quality and efficiency of electric vehicle repair services. In addition, when affiliated repair shops submit quotations, their data often lacks a protection mechanism and is easily obtained by other affiliated repair shops, thus triggering unfair behaviors such as malicious competition and collusive quoting. These problems not only damage the interests of consumers, but also affect the reputation and sustainable development of electric vehicle repair services. Finally, different affiliated repair shops may have differences in the calculation logic and standards of fees, resulting in difficult-to-unify and inaccurate control of the results of combined price calculations, which has a certain impact on the claim reduction of insurance companies. Summary of the Invention

[0003] The present invention provides an intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration to solve the problems mentioned in the above background art:

[0004] The intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration proposed by the present invention includes:

[0005] S1. The surveyor submits information to the system, and the affiliated repair shop according to the preliminary repair requirements pushed by the system;

[0006] S2. Build an automatic quotation model;

[0007] S3. Conduct surrounding price comparison and generate a quotation ranking; automatically generate a recommendation for the best affiliated repair shop;

[0008] S4. The system detects abnormal collaborative behaviors through historical quotation volatility, identifies and prevents collusive quoting behaviors among affiliated repair shops; for affiliated repair shops suspected of collusive quoting, the system marks them as high-risk and focuses on them in subsequent quotations;

[0009] S5. After the repair is completed, accept the repair result. After confirmation, complete the order settlement.

[0010] The intelligent management system for electric vehicle survey tasks based on multi-dimensional data collaboration proposed by the present invention includes a memory, a processor, and a computer program stored on the memory and executable on the memory. The processor executes the program to implement the intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration as described in any one of the above.

[0011] Advantages of the present invention: By means of processes such as automated order allocation, quotation comparison, and repair process monitoring, the processing efficiency of survey tasks is improved; an automatic quotation model is constructed, and combined with historical repair data and real-time market dynamics, a benchmark quotation range is generated to provide a reasonable quotation reference for affiliated repair shops, improving the accuracy of quotations; by analyzing the SHAP value to quantify the influence weight of each feature on the quotation, the transparency and interpretability of the quotation model are enhanced; a sliding window mechanism and Monte Carlo simulation are used to dynamically adjust the floating threshold, avoiding mis-screening caused by static thresholds and improving the reasonableness of quotations; the quotation data of affiliated repair shops is encrypted and transmitted throughout the process to ensure the security of data during transmission; the quotation hash value is stored on the alliance chain to prevent data tampering and provide a traceability basis for subsequent disputes; through technologies such as time series analysis and graph neural networks, collusive quotation behaviors among affiliated repair shops are detected, high-risk affiliated repair shops are marked in a timely manner and given key attention, reducing human manipulation and malicious collusion, and improving the accuracy and efficiency of detection; a dynamic credit score mechanism is introduced to deduct points for collusive behaviors and restrict their order-taking permissions, effectively curbing collusive quotation behaviors; the parts list is automatically associated with the repair order, the compatibility of parts is verified through a knowledge graph, and a compensation plan is automatically generated for false reporting of parts, safeguarding the rights and interests of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a flowchart of the method steps of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0013] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not used to limit the present invention.

[0014] An embodiment of the present invention, as Figure 1 shown, is an intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration. The method includes:

[0015] S1. The general administrator creates new surveyor and franchised repair shop accounts on the web side and assigns corresponding permissions; the order administrator creates new electric vehicle survey orders on the web side, and the system automatically assigns surveyors according to the case information and the distribution of franchised repair shops; the surveyor receives the order using the APP and goes to the site for survey; the surveyor uploads information such as vehicle photos, vehicle location information, and preliminary repair requirements to the system; the franchised repair shop submits a preliminary quotation using the APP according to the preliminary repair requirements pushed by the system.

[0016] S2. The system constructs an automatic quotation model based on historical repair data; compares the preliminary quotation submitted by the franchised repair shop with the quotation generated by the system's automatic quotation model; the system screens out franchised repair shops with reasonable quotations according to the preset floating range.

[0017] S3. Encrypts the quotation data of the franchised repair shop throughout the transmission process; after the system screens out franchised repair shops with reasonable quotations, conducts a comparison of prices in the vicinity, generates a quotation ranking; the system applies a multi-objective decision-making model to automatically generate a recommendation for the best franchised repair shop; the order administrator confirms the winning franchised repair shop according to the system's recommendation.

[0018] S4. The system detects abnormal collaborative behaviors through historical quotation volatility, identifies and prevents collusive quotation behaviors among franchised repair shops; for franchised repair shops suspected of collusive quotations, the system marks them as high-risk and pays key attention to them in subsequent quotations.

[0019] S5. After receiving the repair order, the franchised repair shop starts the repair work and uploads photos of the repair process in real time; after the repair is completed, the franchised repair shop uploads information such as photos of the repaired vehicle, price, and parts list to the system; the order administrator accepts the repair result, and after confirming that there is no error, completes the order settlement.

[0020] The working principle of the above technical solution is as follows: The general administrator is responsible for creating and managing the accounts of surveyors and affiliated repair shops on the web side, and allocating different operation permissions according to roles to ensure the security and controllability of the system; The order administrator creates an electric vehicle survey order on the web side, and details vehicle information, location information, repair requirements, contact information, etc., providing basic data for subsequent survey and repair work; The system automatically assigns the most suitable surveyor to go to the site for survey according to case information, the geographical location of affiliated repair shops, available resources, etc.; The surveyor receives the order using the APP, goes to the site for survey, and uploads information such as vehicle photos, location information, and preliminary repair requirements to the system, providing a basis for subsequent repair quotations; The affiliated repair shop submits a preliminary quotation using the APP according to the preliminary repair requirements pushed by the system, providing comparison data for the system's automatic quotation model; The system constructs an automatic quotation model based on historical repair data for generating relatively accurate repair quotations; The system compares the preliminary quotations submitted by the affiliated repair shops with the quotations generated by the automatic quotation model, and screens out affiliated repair shops with reasonable quotations according to a preset floating range (such as 15%); If the user has a self-selected shop (including non-affiliated repair shops), and the repair quotation of the self-selected shop fluctuates within the upper and lower range of the quotation generated by the automatic quotation model, such as 5%, then the user's self-selected shop is preferred; The quotation data of affiliated repair shops is encrypted throughout the transmission process to ensure data security; After the system screens out affiliated repair shops with reasonable quotations, it conducts price comparison in the vicinity, comprehensively considering factors such as quotation amount, repair time limit, and shop rating, and generates a quotation ranking; The system applies a multi-objective decision-making model, comprehensively considering multiple dimensions (such as a 5-dimensional evaluation system including quotation amount, repair time limit, shop rating, etc.), and automatically generates a recommendation for the best affiliated repair shop; The order administrator confirms the winning affiliated repair shop according to the system recommendation and enters the subsequent repair process; The system detects abnormal collaborative behavior through historical quotation volatility, identifies and prevents collusive quotation behavior among affiliated repair shops; For affiliated repair shops suspected of collusive quotation, the system marks them as high-risk and focuses on them in subsequent quotations to ensure the fairness and reasonableness of quotations; After receiving the repair order, the affiliated repair shop starts the repair work and uploads photos of the repair process in real time to ensure the transparency and traceability of the repair process; After the repair is completed, the affiliated repair shop uploads information such as photos of the repaired vehicle, price, and parts list to the system. The order administrator accepts the repair result and completes the order settlement after confirmation.

[0021] The effects of the above technical solution are as follows: Through the collaborative work of the web end and the APP, the rapid creation, allocation, and processing of survey orders are realized. Surveyors can quickly receive orders and go to the site for survey, greatly shortening the survey time, reducing the waiting time for survey, and optimizing the survey process; The system automatically assigns surveyors and screens affiliated repair shops with reasonable quotes, reducing manual intervention and improving work efficiency; Affiliated repair shops can upload photos of the repair process and post-repair information in real time, enabling order administrators to quickly accept the repair results and further accelerating the order settlement speed; The system constructs an automatic quotation model based on historical repair data, providing a reference standard for the preliminary quotations of affiliated repair shops. By comparing the preliminary quotations submitted by affiliated repair shops with the quotations generated by the system's automatic quotation model and setting a reasonable floating range, affiliated repair shops with reasonable quotations are effectively screened out, improving the accuracy of quotations and the competitiveness of affiliated repair shops, and reducing the risk of quotation errors; The system applies a multi-objective decision-making model, comprehensively considering multiple dimensions, and automatically generates recommendations for the best affiliated repair shops, ensuring the comprehensiveness and fairness of quotations; The quotation data of affiliated repair shops is encrypted throughout the transmission process, effectively preventing data leakage and tampering and ensuring data security; The system detects abnormal collaborative behaviors through historical quotation volatility and can identify and prevent collusive quotation behaviors among affiliated repair shops, maintaining the market order and fair competition environment; For affiliated repair shops suspected of collusive quotations, the system marks them as high-risk and pays key attention to them in subsequent quotations, effectively curbing the occurrence of collusive quotation behaviors; Through fast and accurate survey and repair services, as well as a fair and reasonable quotation mechanism, the satisfaction of customers with electric vehicle survey and repair services is improved; Uploading photos of the repair process and post-repair information in real time enables customers to understand the repair progress and results at any time, enhancing the transparency and credibility of the service.

[0022] In one embodiment of the present invention, the S1 includes:

[0023] S11. When the general administrator creates an account through the web end, based on the RBAC model, the permissions of surveyors and affiliated repair shops are dynamically allocated;

[0024] S12. Introduce the geofencing technology to bind the real-time positions of affiliated repair shops and surveyors;

[0025] S13. The system dynamically calculates the optimal surveyor allocation plan based on the reinforcement learning model, combined with the real-time load of surveyors, historical response speed, and traffic condition data;

[0026] S14. When the order administrator creates a new order, the system automatically calls the Amap / Baidu Map API to parse the vehicle position information and generate three-dimensional geographical coordinates for precise matching of the distribution of affiliated repair shops;

[0027] S15. When the surveyor uploads data through the APP, the system has a built-in image recognition module to automatically verify the integrity of the vehicle photo and extract key fields;

[0028] S16. Maintenance requirements use structured forms, and the system automatically generates maintenance priority labels.

[0029] The working principle of the above technical solution is as follows: when the general administrator creates the accounts of surveyors and franchised repair shops through the Web, permissions are dynamically assigned to different roles based on the RBAC (role-based access control) model; permissions include data access scope (for example, only orders in the region to which they belong), operation permissions (for example, quotation submission, repair progress update) and data encryption level, ensuring that each role can only access and operate data and functions within its scope of authority; introducing geo-fencing technology to bind the franchised repair shops to the real-time locations of surveyors; ensuring that when allocating orders, the system can give priority to matching available personnel within a preset range (for example, 5 kilometers) to improve response speed and survey efficiency; the system is based on a reinforcement learning model (such as DQN), combined with the surveyor's real-time load (current number of tasks), historical response speed and traffic condition data, to dynamically calculate the optimal surveyor allocation plan; through continuous learning and optimization, the system can More accurately predict the availability and response speed of surveyors, so as to make more reasonable allocation decisions; when the order administrator creates a new order, the system automatically calls the Amap / Baidu map API to parse the vehicle location information; generates three-dimensional geographic coordinates (longitude, latitude, altitude) to provide accurate location data for subsequent distribution matching of franchised repair shops and surveyor allocation; when the surveyor uploads vehicle photos through the APP, the system's built-in image recognition module (based on the YOLOv5 algorithm) automatically verifies the integrity of the photos; such as whether the license plate and frame number are clear and identifiable, and extracts key fields (such as vehicle model code) to provide accurate information for subsequent repair quotations and parts preparation; repair requirements are in the form of structured forms, such as checking options such as "expedited" and "vehicle replacement compensation"; the system automatically generates repair priority labels (such as P0-P3) based on the checked options, which are used for subsequent quotation model reference and priority arrangement of repair progress.

[0030] The effects of the above technical solutions are as follows: Through the RBAC model, the general administrator can dynamically allocate permissions to surveyors and affiliated repair shops, ensuring that each role can only access and operate the data and functions within its permission scope, improving the security and controllability of the system, and enhancing the data protection and privacy control capabilities; By introducing the geofencing technology, the real-time locations of affiliated repair shops and surveyors are bound, ensuring that available personnel within the preset range can be preferentially matched when allocating orders, shortening the response time of surveyors, improving the survey efficiency, reducing unnecessary long-distance travel, and saving resources; The system is based on a reinforcement learning model, combines the real-time load of surveyors, historical response speed, and traffic condition data, dynamically calculates the optimal surveyor allocation plan, can more accurately predict the availability and response speed of surveyors, and thus make more reasonable allocation decisions, improving the completion quality and efficiency of survey tasks; When the order administrator creates a new order, the system automatically calls the Gaode / Baidu Map API to parse the vehicle location information and generate three-dimensional geographical coordinates; This provides accurate location data for subsequent matching of the distribution of affiliated repair shops and surveyor allocation, ensuring that orders can be accurately and quickly allocated to the most suitable affiliated repair shop or surveyor; When the surveyor uploads vehicle photos through the APP, the built-in image recognition module of the system can automatically verify the integrity of the photos and extract key fields, reducing the workload of surveyors, improving the efficiency and accuracy of data upload, and providing accurate information for subsequent repair quotation and parts preparation; The repair requirements adopt a structured form, such as checking options like "urgent" and "vehicle replacement compensation", and the system automatically generates repair priority labels; This structured management method makes the repair requirements clearer and more definite, and also provides a strong reference basis for subsequent quotation models and repair schedule arrangements.

[0031] In one embodiment of the present invention, the S2 includes:

[0032] S21. Input features into the XGBoost integrated learning model and output results;

[0033] S22. The XGBoost integrated learning model quantifies the influence weights of each feature on the quotation through SHAP value analysis;

[0034] S23. Adopt a sliding window mechanism to statistically analyze the deviation distribution between the quotations of affiliated repair shops and the system benchmark price in real time, and dynamically adjust the floating threshold through Monte Carlo simulation;

[0035] S24. For quotations exceeding the threshold, trigger an exception mark and associate it with the anti-collusion detection step in S4;

[0036] S25. Combine OCR technology to parse the accessory list pictures uploaded by affiliated repair shops, compare them with the system accessory database (including SKU codes and market prices), and verify the rationality of the quotations.

[0037] The working principle of the above technical solution is as follows: Input features:

[0038] Historical maintenance types: Record the maintenance historical data of different vehicle models and different fault types, providing a reference basis for quotation.

[0039] Spare part cost: The system dynamically captures the supply chain prices to ensure the timeliness and accuracy of spare part costs;

[0040] Labor hour coefficient: Provide a reference for quotation in terms of labor hours according to the complexity and required time of the maintenance tasks;

[0041] Regional economic level: Considering the economic development levels and consumption levels in different regions, adjust the quotation appropriately;

[0042] Seasonal fluctuation factor: Considering the fluctuation of maintenance demand in different seasons, make seasonal adjustments to the quotation;

[0043] Output results:

[0044] Benchmark quotation range: The XGBoost integrated learning model outputs a reasonable benchmark quotation range according to the input features, providing a reference standard for the quotations of franchise repair shops; The XGBoost integrated learning model quantifies the influence weights of each feature on the quotation through SHAP (Shapley Additive Explanations) value analysis; The SHAP value is based on the Shapley value in game theory, which can fairly allocate the contribution of each feature to the model prediction result, helping to understand how the model makes predictions; Through SHAP value analysis, it can be clearly seen which features have the greatest impact on the quotation, providing a basis for subsequent quotation adjustment and optimization; Adopt a sliding window mechanism (for example, the window size is 30 days) to statistically analyze the deviation distribution between the quotations of franchise repair shops and the system benchmark price in real time; This helps to timely discover the changing trend of quotation deviations, providing a basis for dynamically adjusting the floating threshold; Through Monte Carlo simulation, dynamically adjust the floating threshold (for example, 10%-20%) to avoid mis-screening caused by static thresholds; For quotations exceeding the threshold, trigger an anomaly mark, which helps to quickly identify potentially problematic quotations, providing a basis for subsequent investigation and processing; Associate the anomaly-marked quotations with the anti-collusion detection step of S4; Through further analysis and investigation, judge whether there is a collusive quotation behavior among franchise repair shops; Combine OCR technology to parse the pictures of spare part lists uploaded by franchise repair shops; Compare the extracted spare part lists with the system spare part database (including SKU codes and market prices); Through comparison, it can be judged whether the quotations of franchise repair shops are reasonable and whether there are situations such as over-reporting or under-reporting.

[0045] The effects of the above technical solution are as follows: Through the XGBoost integrated learning model, multiple factors such as historical repair types, parts costs, labor-hour coefficients, regional economic levels, and seasonal fluctuation factors are comprehensively considered to output a benchmark quotation range, which can more accurately reflect the actual repair cost and improve the accuracy of quotations. At the same time, the SHAP value analysis is used to quantify the influence weights of each feature on the quotation, making the quotation process more transparent and facilitating understanding and management. The sliding window mechanism is adopted to statistically analyze the deviation distribution between the quotations of franchise repair shops and the system benchmark price in real time, which can dynamically reflect market dynamics and changes in the quotation behavior of franchise repair shops. By dynamically adjusting the floating threshold through Monte Carlo simulation, the problem of false screening that may be caused by static thresholds is avoided, making the quotation system more adaptable to market changes and enhancing the flexibility of quotations. Quotations exceeding the threshold trigger an abnormal mark and are associated with the anti-collusion detection step of S4, which can quickly identify and handle abnormal quotations and prevent unreasonable quotation behaviors from affecting the market order. This mechanism helps to promptly detect and handle collusive quotation behaviors among franchise repair shops and maintain a fair competition environment in the market. Combining OCR technology to analyze the pictures of parts lists uploaded by franchise repair shops and comparing them with the system parts database can automatically verify the reasonableness of quotations, improve the efficiency of quotation review, reduce human errors, and enhance the accuracy of quotation review. Dynamically capturing supply chain prices as input features of parts costs enables the quotation system to promptly reflect changes in parts prices and contributes to the optimization of supply chain management. At the same time, through the feedback mechanism of the quotation system, franchise repair shops can be prompted to more reasonably control parts prices and improve the efficiency of the entire repair service market.

[0046] In one embodiment of the present invention, the S23 includes:

[0047] Set the size of the sliding window according to historical data analysis and business requirements; and select the sliding step size;

[0048] Preprocess the quotation data submitted by franchise repair shops. For each franchise repair shop, calculate the relative deviation or absolute deviation between its quotation and the system benchmark price;

[0049] Within the sliding window, statistically analyze the deviation distribution of all franchise repair shops. Based on historical deviation data, use the Monte Carlo simulation method to generate a large number of possible deviation distribution samples;

[0050] For each floating threshold (e.g., 10%-20%), calculate the probability of false screening (i.e., reasonable quotations are determined to be abnormal) at this threshold; by comparing the false screening probabilities under different thresholds, select the threshold with the minimum false screening probability as the current optimal floating threshold;

[0051] According to external factors (such as market changes, seasonal factors, fluctuations in the prices of accessories, etc.), rerun the Monte Carlo simulation regularly (such as weekly or monthly) to dynamically adjust the floating threshold to adapt to the new market environment.

[0052] The working principle of the above technical solution is as follows: According to historical data analysis and business requirements, the size of the sliding window is set, for example, 30 days, which means that the system will calculate the deviation between the quotes of franchised repair shops and the system benchmark price based on the data of the past 30 days; The selection of the size of the sliding window needs to balance the sufficiency and timeliness of historical data to ensure that it can reflect recent market dynamics and is not overly affected by short-term fluctuations; The sliding step is selected to determine the frequency of window movement; If more detailed capture of deviation changes is required, a smaller sliding step is selected, such as sliding once a day or every few days; The selection of the sliding step affects the granularity and real-time nature of data analysis, and a smaller step can detect market changes faster; The quote data submitted by franchised repair shops is preprocessed to ensure the accuracy and consistency of the data; For each franchised repair shop, calculate the relative deviation or absolute deviation between its quote and the system benchmark price; The relative deviation considers the percentage difference between the quote and the benchmark price and is applicable to comparing quotes of different magnitudes; The absolute deviation is the difference between the quote and the benchmark price and is applicable to the comparison of specific values; Within the sliding window, count the deviation distribution of all franchised repair shops, including statistical indicators such as the average value, standard deviation, maximum value, and minimum value of the deviation; These statistical indicators are used to describe the overall characteristics and fluctuations of the deviation and provide basic data for subsequent Monte Carlo simulations; Based on historical deviation data, use the Monte Carlo simulation method to generate a large number of possible deviation distribution samples; The Monte Carlo simulation reflects the possible deviation situations between the quotes of franchised repair shops and the system benchmark price under different market conditions through random sampling and simulation calculations; The simulation results are used to evaluate the mis-screening probability under different floating thresholds; For each possible floating threshold (e.g., 10%-20%), calculate the probability of mis-screening (i.e., a reasonable quote is judged as abnormal) under this threshold; The mis-screening probability is an important indicator for evaluating the rationality of the threshold, and a lower mis-screening probability means higher accuracy and reliability; By comparing the mis-screening probabilities under different thresholds, select the threshold that minimizes the mis-screening probability as the current optimal floating threshold; The selection of the optimal threshold needs to comprehensively consider the mis-screening probability, actual market conditions, and business requirements; Existing pricing systems usually compare the deviation between the benchmark price and the quotes of franchised repair shops and may use a fixed threshold to judge whether the deviation is too large, resulting in the inability to capture some dynamic market fluctuations in a timely manner; Moreover, existing technologies often judge whether the price is reasonable based on static rules rather than by simulating the mis-screening probability under different floating thresholds, resulting in insufficient accuracy of quote monitoring; According to external factors (such as market changes, seasonal factors, fluctuations in parts prices, etc.), re-run the Monte Carlo simulation regularly (e.g., weekly or monthly); Regularly re-running the simulation can ensure that the floating threshold adapts to the new market environment and maintain the accuracy and effectiveness of quote management; When adjusting the floating threshold, consider the regional economic level factor and set different threshold ranges for different regions; Introduce a seasonal fluctuation factor to reflect the periodic changes in maintenance demand and adjust the threshold to adapt to seasonal fluctuations.

[0053] The effects of the above technical solution are as follows: By setting the sliding window size and statistically analyzing the deviation between the quotes of franchise repair shops and the system benchmark price based on data over a past period (e.g., 30 days), it can more accurately reflect recent market dynamics and changes in the quoting behavior of franchise repair shops, optimize the benchmark price adjustment strategy, improve the system's prediction ability and enhance the data analysis ability. By calculating the relative deviation and absolute deviation and comprehensively considering the difference between the quote and the benchmark price, it improves the comprehensiveness and accuracy of deviation identification, enhances the ability to identify abnormal price fluctuations, and increases the sensitivity to large-scale price changes. Selecting an appropriate sliding step size can more finely capture deviation changes according to business needs, such as sliding once a day or every few days, enabling the quote management system to more flexibly adapt to market changes. At the same time, it can better filter out noise and abnormal fluctuations and improve system stability. By using the Monte Carlo simulation method to generate a large number of possible deviation distribution samples, it reflects the possible deviation situations between the quotes of franchise repair shops and the system benchmark price under different market conditions, providing a more comprehensive perspective for quote management and avoiding the limitations brought by single-scenario analysis. For each possible floating threshold, calculate the probability of false screening at this threshold. By comparing the false screening probabilities under different thresholds, select the threshold with the minimum false screening probability as the current optimal floating threshold, reducing the risk of misjudging reasonable quotes as abnormal. This data-driven threshold setting method makes the floating threshold more in line with the actual market situation, improving the accuracy and fairness of quote management. According to external factors, regularly re-run the Monte Carlo simulation and dynamically adjust the floating threshold to adapt to the new market environment, enabling the quote management system to continuously and effectively respond to market changes, reducing the risks brought by market fluctuations, and avoiding delays or errors that may occur during manual adjustment. Considering the differences in regional economic levels and seasonal fluctuation factors and introducing these factors when adjusting the floating threshold makes quote management more in line with the actual situations of different regions and different seasons, improving the pertinence and effectiveness of quote management. By statistically analyzing statistical indicators such as the average value, standard deviation, maximum value, and minimum value of the deviation, as well as generating a large number of possible deviation distribution samples, it provides rich data support and decision-making basis for the decision-making level.

[0054] In one embodiment of the present invention, S3 includes:

[0055] S31. Transmit the quote data of franchise repair shops through a symmetric encryption algorithm, and the key is dynamically distributed through the SM2 asymmetric encryption algorithm;

[0056] S32. Store the quote hash value on the alliance chain;

[0057] S33. Divide geographical grids based on the GeoHash algorithm, screen franchised repair shops within a preset range, and calculate the expected value of repair timeliness in combination with real-time traffic data;

[0058] S34. Build a repair timeliness prediction model, input historical repair durations, weather data, and traffic flows, and output a time confidence interval;

[0059] S35. Define five-dimensional indicators: quoted price amount (30% weight), repair timeliness (25%), store rating (20%), parts compliance rate (15%), and user complaint rate (10%). The weights support dynamic adjustment (for example, emphasizing timeliness during accident-prone periods such as bad weather);

[0060] S36. Use the NSGA-II multi-objective genetic algorithm to generate a Pareto optimal solution set, the system recommends the top 3 candidate franchised repair shops, and visually displays the scores of each dimension.

[0061] The working principle of the above technical solution is as follows: The symmetric encryption algorithm (such as AES) is used to encrypt and transmit the quotation data submitted by the franchised repair shop to ensure the security of the data during transmission. The feature of symmetric encryption is that the same key is used for encryption and decryption, which has the advantages of public algorithm, small computational amount, fast encryption speed, high encryption efficiency, etc. The SM2 asymmetric encryption algorithm is adopted to dynamically distribute the symmetric encryption key. The public key and private key pair are generated through the SM2 algorithm. The public key is used to encrypt the symmetric key, and the private key is used to decrypt it. The hash value of the encrypted quotation data is calculated (such as using the SHA-256 algorithm). The hash value is unique, and any modification to the original data will cause a change in the hash value. The quotation hash value is stored and certified through the consortium blockchain. Once the data is tampered with, it can be discovered by comparing the stored and certified hash value. The GeoHash algorithm is used to encode the geographical location information into a short string, which represents a geographical grid. The franchised repair shops within a preset range are screened through the GeoHash algorithm to improve the service response speed and customer satisfaction. The expected value of the repair time limit is calculated in combination with real-time traffic data (such as the congestion index). The real-time traffic data reflects the current traffic conditions. By considering the traffic congestion situation, the time for the franchised repair shop to reach the repair site can be estimated more accurately. The historical repair duration, weather data, and traffic flow are selected as input features. These features have an important impact on the repair time limit. For example, bad weather may cause traffic congestion, which in turn affects the repair time limit. The long short-term memory network (LSTM) is used to construct a repair time limit prediction model. By training the LSTM network, the complex relationship between the input features and the repair time limit can be learned, so as to predict the future repair time limit. The model outputs a time confidence interval instead of a single time point. The time confidence interval represents the uncertainty range of the prediction result, which helps users to understand the repair time limit more comprehensively. Five dimensions, namely the quotation amount, repair time limit, store rating, parts compliance rate, and user complaint rate, are defined as the indicators for evaluating franchised repair shops. Each indicator has a corresponding weight, and the weight reflects the importance degree of the indicator. The weight supports dynamic adjustment. For example, during the accident-prone period such as bad weather, the time limit can be emphasized, and the weight of the repair time limit can be increased. By dynamically adjusting the weight, it can be ensured that the system can evaluate the comprehensive performance of franchised repair shops more accurately under different market environments. The NSGA-II multi-objective genetic algorithm is adopted to generate the Pareto optimal solution set. NSGA-II is a non-dominated sorting genetic algorithm, which is suitable for solving problems with multiple conflicting objectives. In this solution, the NSGA-II algorithm is used to find the optimal combination of franchised repair shops in five dimensions. The system recommends the top 3 candidate franchised repair shops from the Pareto optimal solution set. By visually displaying the scores of each dimension, users can intuitively understand the advantages and disadvantages of each candidate franchised repair shop, so as to make a more informed choice.

[0062] The effects of the above technical solution are as follows: The quotation data of franchised repair shops is transmitted through a symmetric encryption algorithm, ensuring the confidentiality of data during transmission and avoiding the risk of data leakage. At the same time, the key is dynamically distributed through the SM2 asymmetric encryption algorithm, further improving the security of key management, preventing the key from being illegally obtained or intercepted, and thus ensuring the integrity and confidentiality of the quotation data. The quotation hash value is stored on the alliance chain, and using the immutability of the blockchain effectively prevents the data from being maliciously tampered with and enhances the credibility of the data. Based on the GeoHash algorithm, geographical grids are divided to quickly screen franchised repair shops within a preset range, and the expected value of repair timeliness is calculated in combination with real-time traffic data, which can accurately estimate the arrival time of repair services, improve the service response speed, and thus enhance customer satisfaction. A repair timeliness prediction model is constructed. By inputting multi-dimensional information such as historical repair duration, weather data, and traffic flow, a time confidence interval is output. The LSTM network is good at processing time series data, can capture the complex relationships between repair timeliness and various factors, thus improving the accuracy of prediction, and can maintain good prediction stability, is not easily affected by changes in historical data and cause the model performance to decline, and can better handle the timeliness and dynamics in time series data to ensure that the prediction results are more accurate. Five-dimensional indicators of quotation amount, repair timeliness, store rating, parts compliance rate, and user complaint rate are defined, and corresponding weights are set. The multi-dimensional evaluation method can more comprehensively reflect the comprehensive strength of franchised repair shops, avoid the errors of single indicators, and improve the accuracy of evaluation. At the same time, the weights support dynamic adjustment. For example, during periods with frequent accidents such as bad weather, more emphasis can be placed on timeliness, making the evaluation system more flexible and adaptable. The NSGA-II multi-objective genetic algorithm is used to generate the Pareto optimal solution set, and the system recommends the top 3 candidate franchised repair shops from it. The NSGA-II algorithm can handle multiple conflicting objectives and find a set of balanced solutions, making the recommendation results more reasonable and reliable. At the same time, by visually displaying the scores of each dimension, users can intuitively understand the advantages and disadvantages of each candidate franchised repair shop and thus make a more informed choice.

[0063] In one embodiment of the present invention, S31 includes:

[0064] S311. Select a symmetric encryption algorithm according to data characteristics and security requirements, and set the key length;

[0065] S312. Divide the quotation data of franchised repair shops into several small pieces, and use multi-threaded or distributed computing resources to perform parallel encryption processing on the divided data;

[0066] After encryption processing, record the processing time of each data block; and, before encryption, calculate the checksum of the data and transmit the checksum together with the encrypted data; at the receiving end, calculate the checksum of the received encrypted data and compare it with the transmitted checksum; if the checksums are consistent, it indicates that the data has not been lost or incomplete during transmission.

[0067] S314. The system generates a pair of SM2 key pairs, including a public key and a private key; the public key is used to encrypt the key of the symmetric encryption algorithm, and the private key is used for decryption; the system sends the public key to the franchised repair shop through a secure channel.

[0068] S315. The franchised repair shop uses the public key to encrypt the key of the symmetric encryption algorithm and sends the encrypted key to the system; the system uses the private key to decrypt to obtain the key of the symmetric encryption algorithm for decrypting the franchised repair shop's quotation data.

[0069] S316. Predict the key expiration time based on the data block processing time and transmission delay, and dynamically generate a new key pair in advance, and distribute it to the franchised repair shop through the SM2 algorithm.

[0070] S317. When the key is about to expire or needs to be updated, the system notifies the franchised repair shop to update the key through a secure channel; after receiving the notice, the franchised repair shop obtains the new key of the symmetric encryption algorithm according to the new key distribution mechanism.

[0071] S318. During data transmission, the system monitors the transmission status in real time to detect whether there is any abnormality; if a transmission abnormality is detected, the system starts a retry mechanism to resend the data.

[0072] The working principle of the above technical solution is as follows: According to the data characteristics and security requirements, select a suitable symmetric encryption algorithm; for example, select AES (Advanced Encryption Standard) as the symmetric encryption algorithm. AES has a fast encryption speed, high security, and is suitable for encrypting and transmitting a large amount of data; set the key length of the AES algorithm, such as 128 bits, 192 bits, or 256 bits, to meet the requirements of different security levels; divide the franchise repair shop quotation data into several small pieces to utilize multi-threaded or distributed computing resources for parallel encryption processing; through parallel encryption, the encryption efficiency can be significantly improved and the encryption time can be shortened. Traditional encryption algorithms are often single-threaded or have limited parallelism, and there are performance bottlenecks when processing a large amount of data, especially in large-scale data processing environments and are not widely used; at the same time, record the processing time of each data block and use it in the subsequent key distribution strategy; before encryption, calculate the checksum of the data (such as MD5 or SHA-256 hash value) and transmit the checksum together with the encrypted data; at the receiving end, calculate the checksum of the received encrypted data and compare it with the transmitted checksum; if the checksums are consistent, it indicates that the data has not been lost or incomplete during transmission, ensuring the integrity of the data. Existing systems usually adopt a simple checksum mechanism, but may not perform a complete check between the encryption and decryption processes. The method of transmitting the checksum together with the data is relatively rare, and it relies more on a single checksum mechanism to determine whether the data is transmitted completely; the system generates a pair of SM2 key pairs, including a public key and a private key; the public key is used to encrypt the key of the symmetric encryption algorithm, and the private key is used to decrypt; the system sends the public key to the franchise repair shop through a secure channel (such as the HTTPS or SSL / TLS protocol) to ensure the security of the public key during transmission; the franchise repair shop uses the received public key to encrypt the key of the symmetric encryption algorithm and sends the encrypted key to the system; the system uses the private key to decrypt to obtain the key of the symmetric encryption algorithm for decrypting the franchise repair shop quotation data; according to the data block processing time and transmission delay, predict the key expiration time and dynamically generate a new key pair in advance, and distribute it to the franchise repair shop through the SM2 algorithm; when the key is about to expire or needs to be updated, the system notifies the franchise repair shop to update the key through a secure channel, enhancing the security of use. After receiving the notice, the franchise repair shop obtains the new key of the symmetric encryption algorithm according to the new key distribution mechanism; through the key update mechanism, it ensures that the system can continuously use secure keys for data encryption and decryption. Most existing technologies rely on static key management mechanisms. Once the key is generated, it usually will not be updated regularly or is updated manually without an automated update and expiration processing mechanism, resulting in a greatly increased risk of key leakage; during data transmission, the system monitors the transmission status in real time to detect whether there are any abnormalities (such as network interruption, data transmission timeout, etc.); if a transmission abnormality is detected, the system starts a retry mechanism to resend the data;Through transmission status monitoring and retry mechanisms, reliable data transmission and integrity are ensured.

[0073] The effects of the above technical solutions are as follows: The symmetric encryption algorithm improves the encryption speed and enhances the security performance; encrypting and transmitting a large amount of data meets the dual requirements of data transmission for efficiency and security; dividing the franchise repair shop quotation data into several small pieces and performing parallel encryption processing using multi-threaded or distributed computing resources significantly improves the encryption efficiency, can make full use of computing resources, shorten the encryption time, and improve the overall processing ability of the system. In a distributed environment, it can ensure the smooth completion of the encryption task, enhancing the reliability and fault tolerance of the system; the system can continue to run when a node fails and will not interrupt the entire encryption process due to a single-point failure, enhancing the fault tolerance of the system; calculating the checksum of the data (such as MD5 or SHA-256 hash value) before encryption and transmitting the checksum together with the encrypted data; at the receiving end, calculating the checksum of the received encrypted data and comparing it with the transmitted checksum to ensure that the data is not lost or incomplete during transmission, effectively guaranteeing the integrity and reliability of the data; dynamically distributing the key of the symmetric encryption algorithm through the SM2 asymmetric encryption algorithm, enhancing the security of key management; the system generates a pair of SM2 key pairs, the public key is used to encrypt the key of the symmetric encryption algorithm, and the private key is used to decrypt, ensuring the security of the key during transmission and preventing the key from being illegally obtained or intercepted; by predicting the key expiration time and generating a new key pair in advance, avoiding security vulnerabilities caused by long-term use or expiration of the key and reducing the possibility of attackers cracking the key; adopting the SM2 national cryptographic algorithm, based on elliptic curve cryptography (ECC), with a shorter key length and higher encryption efficiency at the same security level, effectively resisting classical attack means before quantum computing; the dynamic key generation and distribution mechanism can smoothly transition the key update process, avoiding service suspension or data encryption failure caused by key expiration; by predicting the transmission delay and planning the key distribution time in advance, ensuring that the key update can still be completed on time even when the network is unstable, reducing the key synchronization failure rate caused by network problems. By introducing two physical quantities, the data block processing time and the transmission delay, it is possible to more accurately predict the usage and expiration time of the key, thereby achieving more accurate and real-time key distribution, ensuring the accuracy and real-time nature of key distribution, and also improving the efficiency and security of the entire encryption system.

[0074] In an embodiment of the present invention, the S312 includes:

[0075] Before chunking, preprocess the franchise repair shop quotation data, and formulate a chunking strategy according to the characteristics of the data and the requirements of the encryption algorithm;

[0076] For linear data structures, such as arrays or lists, the sequential chunking algorithm is adopted to sequentially divide the data into several small pieces from beginning to end.

[0077] For complex data structures, such as hash tables or graphs, the hash block algorithm is adopted to allocate them into different blocks according to the hash values of the data;

[0078] For large files, the memory-mapped file technology is utilized to map the large files into memory and block the data in memory;

[0079] According to the parallel processing ability of the system, a thread pool or a process pool is created, and the thread pool or the process pool is used to execute parallel encryption tasks; the blocked data is allocated to each thread or process in the thread pool or the process pool for encryption processing through static allocation (i.e., each thread or process processes a fixed number of blocks) or dynamic allocation (i.e., dynamically adjusting the number of allocated blocks according to the load of the thread or process);

[0080] The symmetric encryption algorithm is encapsulated into a function that accepts a data block and a key as inputs and outputs an encrypted data block; the encryption function is executed in parallel in each thread or process to encrypt the data block allocated to itself;

[0081] The encrypted data blocks of each thread or process are merged into the complete encrypted data through sequential merging or parallel merging.

[0082] The working principle of the above technical solution is as follows: Before partitioning the quotation data of franchise repair shops, preprocessing is first performed, including removing redundant parts in the data to reduce the data volume and improve the encryption efficiency; compressing the data to further reduce the storage space occupied by the data, facilitating subsequent partitioning and encryption operations; formulating a suitable partitioning strategy according to the characteristics of the data and the requirements of the encryption algorithm; the partition size can be determined based on factors such as the size of the data, the block size of the encryption algorithm, memory limitations, and parallel processing capabilities; fixed-size partitions can be selected, or partitioning can be performed according to the natural boundaries of the data (such as records, rows, or columns) to adapt to the characteristics of different data structures; for linear data structures (such as arrays or lists), a sequential partitioning algorithm is used to sequentially divide the data into several small pieces from beginning to end; for complex data structures (such as hash tables or graphs), a hash partitioning algorithm is used to distribute the data to different blocks according to the hash value of the data; for large files, the memory-mapped file (mmap) technology is used to map the large file into memory and partition the data in memory. Although the prior art has partitioning methods using fixed block sizes or based on the natural boundaries of the data, it does not make special processing for complex data structures (such as hash tables, graphs, etc.), and existing parallel encryption is usually based on static task allocation and simple file reading methods, lacking flexible dynamic load distribution and memory optimization; moreover, the prior art rarely considers the memory-mapped file technology, usually requiring all data to be read at once and then processed, resulting in high memory occupancy and I / O overhead; according to the parallel processing capabilities of the system, a thread pool or process pool is created to execute parallel encryption tasks; the size of the thread pool or process pool can be determined according to the number of CPU cores, memory size, and parallel encryption requirements of the system; through static allocation (i.e., each thread or process processes a fixed number of blocks) or dynamic allocation (i.e., dynamically adjusting the number of blocks allocated according to the load of the thread or process) strategies, the partitioned data is allocated to each thread or process in the thread pool or process pool for encryption processing; encapsulate the symmetric encryption algorithm (such as AES) into a function that accepts a data block and a key as inputs and outputs the encrypted data block; the encryption function should have a good interface design to facilitate parallel execution in each thread or process; execute the encryption function in parallel in each thread or process to encrypt the data block allocated to itself; through parallel processing, the encryption efficiency can be significantly improved and the encryption time can be shortened; the encrypted data blocks of each thread or process are merged into the complete encrypted data through sequential merging or parallel merging (such as using the idea of merge sort); during the merging process, the order and integrity of the data blocks should be ensured so that the original data can be correctly decrypted and restored at the receiving end, while traditional encryption schemes usually directly write the encrypted data to the target file without involving a complex merging process.

[0083] The effects of the above technical solution are as follows: By removing redundant data and compressing data in the preprocessing step, the amount of data to be encrypted is reduced, thus significantly improving the efficiency of the encryption process; By formulating a reasonable chunking strategy and determining the chunk size according to data characteristics and the requirements of the encryption algorithm, the encryption process becomes more efficient, the parallel processing ability of encryption is improved, and at the same time, according to the requirements of different encryption algorithms, the memory management and data transmission performance are also improved to a certain extent; Using a thread pool or a process pool to process parallel encryption tasks makes full use of the multi-core processing ability of the system, further improving the encryption speed and reducing the latency of the encryption process; Adopting a sequential chunking algorithm for linear data structures (such as arrays or lists), a hash chunking algorithm for complex data structures (such as hash tables or graphs), and using the memory mapping file (mmap) technology to chunk large files enables this solution to adapt to various different types of data structures, improving the efficiency and response speed of data processing and optimizing the resource utilization rate; This flexibility enables this solution to perform excellently in a wider range of application scenarios; By using a static allocation or dynamic allocation strategy to allocate the chunked data to each thread or process in the thread pool or process pool for encryption processing, the utilization of system resources is optimized, the overhead of task scheduling is reduced, and the load is evenly distributed among all threads or processes, thus avoiding overloading of certain threads or processes and resulting in a decrease in processing efficiency; The dynamic allocation strategy can dynamically adjust the number of chunks allocated according to the load conditions of the threads or processes, avoiding resource idleness or overload and improving the stability and reliability of the system; Encapsulating the symmetric encryption algorithm (such as AES) into a function that accepts a data chunk and a key as inputs and outputs the encrypted data chunk simplifies the implementation of the encryption process, enables the encryption function to be executed in parallel in each thread or process, improves the code reusability and maintainability, reduces the error rate, and while improving the system maintainability and scalability, also enhances the security; Merging the encrypted data chunks of each thread or process into the complete encrypted data through sequential merging or parallel merging ensures the integrity of the data; During the merging process, methods such as checksum and hash value can be used for data integrity verification to ensure that the encrypted data has not been tampered with or damaged during transmission or storage.

[0084] In one embodiment of the present invention, S4 includes:

[0085] S41. Based on time series analysis, calculate the volatility standard deviation of the quotes of franchised repair shops. If the synchronization rate of more than 3 shops in the region exceeds 90%, it is determined as potential collusion;

[0086] S42. Based on a graph neural network, construct a relationship topology graph of franchised repair shops and detect abnormal collaboration patterns through node embedding features;

[0087] S43. Assign a dynamic credit score to high-risk franchised repair shops. The initial value is 100 points. Each time a conspiracy behavior is detected, 20 points will be deducted. When the score is below 60, restrict its order receiving permission.

[0088] S44. Update the detection rules by real-time fusing new data through an online learning algorithm.

[0089] The working principle of the above technical solution is as follows: First, use the ARIMA model to analyze the time series data of the quotes of franchised repair shops. The ARIMA model is a time series prediction model that can effectively capture trends, seasonality, and random fluctuations in time series data. Through the ARIMA model, calculate the standard deviation of the quote volatility of each franchised repair shop, which reflects the degree of quote volatility. Then, compare the volatilities of multiple shops in the region. If the synchronization rate (i.e., the consistency of the fluctuation trend) of the volatilities of more than 3 shops in the region exceeds 90%, it is determined that there is a potential conspiracy behavior among these shops. A synchronization rate exceeding 90% means that the quote fluctuation trends of these shops are highly consistent, which may involve price manipulation or coordinated behavior. Use a graph neural network (GNN) to construct a relationship topology graph of franchised repair shops. The GNN is a deep learning model specialized for processing graph-structured data that can capture complex relationships in the data. In the relationship topology graph, each franchised repair shop is regarded as a node, and the edges between nodes represent the relationships between shops (such as geographical proximity, frequent business transactions, etc.). Through the GNN, learn the node embedding features to extract feature representations that can reflect the relationships between nodes. Use these feature representations to detect abnormal collaborative patterns in the graph, such as small groups with frequently similar quotes. The shops in these small groups may have conspiracy behaviors and manipulate market prices through coordinated quotes. Use the GNN to detect abnormal collaborative patterns in the relationship network to further confirm the existence of conspiracy behaviors. Assign a dynamic credit score to the franchised repair shops determined to be at high risk (i.e., having potential conspiracy behaviors). The initial credit score is set at 100 points. Each time a conspiracy behavior is detected, 20 points will be deducted. When the credit score is below 60, restrict the order receiving permission of the franchised repair shop to prevent it from continuing to participate in market manipulation behaviors. Update the detection rules by real-time fusing new data through an online learning algorithm. The online learning algorithm can process streaming data and continuously adapt to changes in the market environment. As new data arrives, the algorithm can continuously optimize and adjust the detection rules to improve the accuracy and robustness of detection.

[0090] The effects of the above technical solutions are as follows: By calculating the volatility standard deviation of the quotes of franchised repair shops through time series analysis and comparing the volatility synchronization rate of multiple shops within a region, potential collusive behaviors can be accurately identified; this method is based on the volatility characteristics of the data itself, avoiding the inaccuracy of subjective judgment, improving the detection accuracy, and enhancing the risk warning ability; using a graph neural network (GNN) to construct a relationship topology graph of franchised repair shops can reveal the potential relationship network among the shops; detecting abnormal collaborative patterns through node embedding features, such as small groups with frequently similar quotes, can further confirm the existence of collusive behaviors; this method can handle complex data relationships, discover hidden collusive patterns, enhance the supervision ability of collusive behaviors, and improve the detection efficiency and accuracy; assigning dynamic credit scores to high-risk franchised repair shops and deducting credit scores according to the number of collusive behaviors, restricting the order-taking authority when the score is below a certain threshold, effectively constraining and punishing high-risk franchised repair shops, encouraging them to abide by market rules and maintaining a fair competition environment in the market; updating the detection rules by real-time integrating new data through an online learning algorithm can enable the detection system to continuously adapt to the changes in the market environment; as new data arrives, the detection rules can be continuously optimized and adjusted to improve the detection accuracy and robustness; this self-adaptability enables the detection system to maintain effectiveness in the long term; by accurately detecting collusive behaviors, revealing complex relationships, enhancing credit management, and improving the adaptability of detection rules, it helps to improve the overall efficiency and fairness of the market; the reduction of collusive behaviors can promote market competition and improve the efficiency of resource allocation; the enhancement of credit management can maintain market order and protect the rights and interests of consumers; the adaptability of detection rules can ensure that market rules keep pace with the times and meet the needs of market development.

[0091] In one embodiment of the present invention, step S42 includes:

[0092] S421. Collect the corresponding data among franchised repair shops and preprocess the collected corresponding data;

[0093] S422. Regard each franchised repair shop as a node in the graph, and the node contains the basic information of the franchised repair shop; define the edges between the nodes according to the relevant factors among the franchised repair shops; wherein, the weight of the edge represents the association strength between the franchised repair shops.

[0094] S423. Use the relevant knowledge of graph theory to combine the nodes and edges into a relationship topology graph of franchised repair shops.

[0095] S424. Extract initial features for each node, and use the extracted node features and the structural information of the topology graph to train the GNN model.

[0096] S425. Generate an embedding vector for each node through the trained GNN model; and perform dimensionality reduction on the generated embedding vectors to visualize the distribution of nodes in a two-dimensional or three-dimensional space;

[0097] S426. Calculate the similarity of embedding vectors between nodes through cosine similarity, and detect small groups with similar frequent quotes by using a clustering algorithm based on the similarity between nodes;

[0098] S427. Combine business knowledge and the expert knowledge base to identify abnormal collaboration patterns; and timely feedback the detection results to the system; meanwhile, use the detection results as the input of the online learning algorithm to update the detection rules in real time.

[0099] The working principle of the above technical solution is as follows: collect corresponding data among franchise repair shops, including transaction data and quotation data; these data are the basis for constructing a relationship topology graph; preprocess the collected data, such as removing outliers, filling in missing values, data standardization, etc., to ensure the quality and consistency of the data; regard each franchise repair shop as a node in the graph, and the node contains the basic information of the franchise repair shop, such as store ID, location, business scope, etc.; define the edges between nodes according to relevant factors among franchise repair shops, such as transaction relationships, quotation similarities, and interaction frequencies. The weight of the edge represents the association strength between franchise repair shops and reflects the closeness between them; use relevant knowledge of graph theory to combine nodes and edges into a relationship topology graph of franchise repair shops. This graph reflects the complex relationship network among franchise repair shops; extract initial features for each node, including the basic information of the franchise repair shop, historical transaction data, quotation patterns, etc.; use the extracted node features and the structural information of the topology graph to train a GNN model; the GNN model can capture the complex relationships between nodes and learn the embedding representations of nodes; through the trained GNN model, generate an embedding vector for each node; the embedding vector captures the local and global structural information of the node in the topology graph; perform dimensionality reduction processing on the generated embedding vector (such as PCA, t-SNE, etc.) to visualize the distribution of nodes in a two-dimensional or three-dimensional space; calculate the similarity of embedding vectors between nodes through cosine similarity; based on the similarity between nodes, use a clustering algorithm (such as K-means, DBSCAN, etc.) to detect small groups with similar frequent quotations; traditional clustering methods, such as K-means, DBSCAN and other clustering algorithms, usually have high requirements for data preprocessing and parameter setting, and it is difficult to process data with complex structures and dependencies; there may be potential collusion behaviors among the franchise repair shops in these small groups; combine business knowledge and an expert knowledge base to identify abnormal collaborative patterns. For example, if some franchise repair shops have highly similar quotations within a specific time period and there is no other reasonable explanation, this may be regarded as a potential collusion behavior; timely feedback the detection results to the system so that the system can take corresponding measures (such as restricting the order acceptance authority, conducting investigations, etc.); at the same time, use the detection results as the input of an online learning algorithm to update the detection rules in real time to adapt to changes in the market environment; traditional methods rely on rules or simple statistical methods for anomaly detection and are difficult to dynamically adapt to changes in data.

[0100] The effects of the above technical solution are as follows: By collecting transaction data and quotation data among franchised repair shops and constructing a relationship topology graph, the complex relationship network among franchised repair shops can be comprehensively revealed, which helps to understand the interaction patterns among franchised repair shops, provides a basis for subsequent anomaly detection, and improves the accuracy of anomaly detection; Initial features including basic information, historical transaction data, quotation patterns, etc. are extracted for each franchised repair shop node. These features can accurately reflect the business conditions and quotation behaviors of franchised repair shops, enhance the insight into market trends and competition situations, and reduce the misjudgment rate; Through the training of the GNN model, these features are further integrated and utilized to generate embedding vectors that can capture the local and global structural information of nodes in the topology graph; Using the embedding vectors generated by the trained GNN model, the similarity between nodes is calculated through cosine similarity, and a clustering algorithm is used to detect small groups with similar frequent quotations, which can effectively identify small groups that may have potential collusion behaviors, provide guarantee for the fair competition in the market, and improve the accuracy of collusion detection; The generated embedding vectors are processed for dimensionality reduction, and the distribution of nodes is visualized in a two-dimensional or three-dimensional space. This visual analysis helps to intuitively understand the relationship network among franchised repair shops and the distribution of abnormal collaboration patterns, and further deepen the understanding of the data; When identifying abnormal collaboration patterns, combining business knowledge and an expert knowledge base can more accurately judge whether the quotation behavior is abnormal; This method combining business practice and expert experience improves the accuracy and reliability of detection; The detection results are used as the input of an online learning algorithm to update the detection rules in real time. This enables the detection system to adapt to changes in the market environment and continuously improve the accuracy and efficiency of detection.

[0101] In one embodiment of the present invention, the S426 includes:

[0102] Before calculating the similarity, the generated embedding vectors are standardized. For each pair of nodes in the topology graph, the cosine similarity between their embedding vectors is calculated through matrix operations or loop traversal;

[0103] According to the characteristics of the data and the requirements of the clustering task, a clustering algorithm is selected;

[0104] Taking the similarity matrix as the input, the selected clustering algorithm is applied for clustering analysis;

[0105] For each cluster, the nodes within the cluster are analyzed, and whether it belongs to a small group with similar frequent quotations is determined by calculating the statistical features (such as mean, variance, etc.) of the quotation data of the nodes within the cluster;

[0106] According to the number of nodes within the cluster and the degree of quotation similarity, the scale of the small group is determined; Combining the actual situation of the electric vehicle survey business, it is analyzed whether the quotation behavior of the small group conforms to normal logic.

[0107] The working principle of the above technical solution is as follows: Before calculating the similarity, the embedding vectors generated by the GNN model are normalized. Normalization is to eliminate the dimensional differences between different dimensions, making the similarity calculation more accurate. Existing technologies usually rely on traditional similarity measurement methods (such as Euclidean distance, Manhattan distance, etc.), but these methods may be affected by data scale and noise, resulting in unstable calculation results; for each pair of nodes in the topological graph, the cosine similarity between their embedding vectors is calculated through matrix operations or loop traversal. Cosine similarity is an index to measure the direction similarity of two vectors, and its value range is between [-1, 1]. The closer the value is to 1, the more similar the two vectors are; according to the characteristics of the data and the requirements of the clustering task, a suitable clustering algorithm is selected. For example, K-means is suitable for processing spherical clusters, while DBSCAN can discover clusters of any shape and has good robustness to noisy data; parameters are set for the selected clustering algorithm. For the K-means algorithm, the number of clusters K needs to be determined; for the DBSCAN algorithm, the neighborhood radius Eps and the minimum number of samples MinPts need to be set; the similarity matrix is used as the input, and the selected clustering algorithm is applied for clustering analysis. The clustering algorithm divides the nodes into different clusters according to the similarity between the nodes; the clustering results are evaluated to determine whether the cluster division is reasonable. Indicators such as the silhouette coefficient and the Calinski-Harabasz index are used to evaluate the clustering effect. These indicators can reflect the tightness of the nodes within the cluster and the separation of the nodes between clusters; for each cluster, the nodes within the cluster are analyzed, and whether they belong to a small group with similar frequent quotes is determined by calculating the statistical characteristics (such as mean, variance, etc.) of the quote data of the nodes within the cluster; according to the number of nodes within the cluster and the degree of quote similarity, the scale of the small group is determined. Larger small groups may be more worthy of attention because they may have more obvious collusion behaviors; combined with the actual situation of the electric vehicle survey business, it is analyzed whether the quote behavior of the small group conforms to normal logic. For example, if the quotes of some franchised repair shops are highly similar during a specific period and there is no other reasonable explanation, this may be regarded as a potential collusion behavior. Most existing technologies use fixed clustering algorithms (such as K-means or DBSCAN). These methods may not consider the characteristics of different types of data, nor do they further optimize the behavior analysis after clustering. Moreover, existing methods only perform simple clustering statistics (such as the size and density of clusters) after clustering, lacking in-depth analysis and logical verification of the behaviors of the nodes within the cluster.

[0108] The effects of the above technical solution are as follows: Before calculating the similarity, the generated embedding vectors are normalized to eliminate the dimensional differences between different dimensions, making the calculation of cosine similarity more accurate and reliable. This helps to more precisely reflect the similarity degree between nodes, provides accurate basic data for subsequent clustering analysis, improves the performance and generalization ability of the model, and enhances the reliability of clustering analysis. According to the characteristics of the data and the requirements of the clustering task, a suitable clustering algorithm is selected and reasonable parameters are set. This flexible design enables the clustering algorithm to better adapt to different data distributions and clustering requirements, improving the accuracy and stability of the clustering effect. Evaluation indicators such as the silhouette coefficient and the Calinski-Harabasz index are used to evaluate the clustering results to determine whether the cluster division is reasonable. These evaluation indicators can objectively reflect the tightness of nodes within the cluster and the separation degree of nodes between clusters, providing a strong basis for optimizing the clustering effect, improving the efficiency and precision of clustering analysis, and promoting the improvement and iteration of the model. Through clustering analysis, similar nodes are divided into the same cluster, and statistical feature analysis (such as mean, variance, etc.) is performed on the nodes within each cluster to accurately identify small groups with similar frequent quotes. This helps to reveal the quote behavior patterns among franchise repair shops. According to the number of nodes within the cluster and the degree of quote similarity, the scale of the small group is determined, and key attention is paid to the larger small groups. Larger small groups may have more obvious collusive behaviors, so this attention helps to detect and handle potential collusive behaviors in a timely manner. Combining with the actual situation of the electric vehicle survey business, it is analyzed whether the quote behavior of the small group conforms to normal logic. This analysis method combined with the business reality improves the accuracy and reliability of collusive behavior identification, providing strong evidence support for subsequent investigations and handling.

[0109] In one embodiment of the present invention, the S5 includes:

[0110] S51. Based on the repair progress, obtain the video clip of the repair site in real time, and verify the authenticity of the parts replacement through the target detection model;

[0111] S52. The system compares the vehicle photos before and after the repair, calculates the image similarity using the Siamese network. If the difference is lower than the threshold, the system review is triggered;

[0112] S53. The parts list is automatically associated with the repair order, and the system calls the knowledge graph to verify the parts compatibility. For the behavior of falsely reporting parts, a compensation plan is automatically generated;

[0113] S54. After the acceptance is passed, the system automatically triggers the settlement through the smart contract, and the user evaluation data flows back to the store rating module to form a closed-loop optimization of "quote - repair - feedback".

[0114] The working principle of the above technical solution is as follows: At key nodes during the maintenance process (such as when starting maintenance and when maintenance is halfway through), the system automatically obtains video clips of the maintenance site; the system analyzes the key frames in the video clips through an object detection model (such as FasterR-CNN) to identify and verify whether the parts are truly replaced; the object detection model can accurately identify information such as the type and location of the parts and compare it with the part information in the maintenance order to ensure the authenticity of part replacement; the system obtains vehicle photos before and after maintenance and uses a Siamese network to calculate the similarity between the two photos; the Siamese network is a neural network structure used to measure the similarity between two inputs and can accurately calculate the similarity between images; if the similarity of the vehicle photos before and after maintenance is lower than a set threshold (such as 85%), the system review is triggered to further confirm the maintenance effect; the parts list is automatically associated with the maintenance order, and the system can accurately track the part information used for each maintenance order; the system calls a knowledge graph (such as constructed using Neo4j) to verify the compatibility of the parts. The compatibility relationship between vehicle models and parts is stored in the knowledge graph, such as a certain vehicle model only supports a specific battery model; for false reporting of parts, the system automatically generates a compensation plan according to preset rules, such as deducting the deposit at 3 times the market price; punish false reporting of parts to maintain market order; after the maintenance service passes the acceptance, the system automatically triggers the settlement process through a smart contract (such as using HyperledgerFabric); the smart contract is an automatically executed contract that can automatically execute preset operations (such as transferring money and deducting money) when specific conditions are met; user evaluation data flows back to the store rating module to evaluate the service quality and reputation of the store; by forming a closed-loop optimization mechanism of "quotation - maintenance - feedback", the service quality and user satisfaction are continuously improved.

[0115] The effects of the above technical solutions are as follows: By obtaining real-time video clips of the repair site and using a target detection model (such as Faster R-CNN) to verify the authenticity of parts replacement, the transparency of the repair process is significantly improved, which helps prevent false repairs or situations where parts are not replaced, enhancing consumers' trust in the repair service; The system compares the vehicle photos before and after repair and uses a Siamese network to calculate the image similarity. This objective and automated evaluation method can accurately reflect the repair effect. If the difference is lower than the threshold, the system will trigger a review to further ensure the repair quality; The parts list is automatically associated with the repair order, and the system calls the knowledge graph to verify the compatibility of parts, preventing repair failures or safety hazards caused by incompatible parts. At the same time, a compensation plan is automatically generated for false reporting of parts, maintaining the market order and consumers' rights and interests; After passing the acceptance, the system automatically triggers the settlement process through a smart contract, reducing manual intervention, improving the settlement efficiency and accuracy, and reducing the risk of human errors; The user evaluation data flows back to the store rating module, forming a closed-loop optimization mechanism of "quotation - repair - feedback". This helps the store promptly understand the service quality and user satisfaction, so as to continuously adjust and optimize the service process and improve the overall service quality; Through the implementation of the above technical solutions, consumers can more intuitively understand the repair process and have a more objective evaluation of the repair effect. At the same time, the automated settlement and closed-loop optimization mechanism also improve the user experience and trust. This helps enhance the competitiveness of the store and promote the healthy development of the market.

[0116] An embodiment of the present invention, an intelligent management system for electric vehicle survey tasks based on multi-dimensional data collaboration, includes a memory, a processor, and a computer program stored on the memory and executable on the memory. The processor executes the program to implement the intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration as described in any one of the above.

[0117] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. An intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration, characterized in that, The method includes: S1. The surveyor submits information to the system, and the franchised repair shop acts according to the preliminary repair requirements pushed by the system; S2. Build an automatic quotation model; S3. Conduct price comparison around, generate a quotation ranking; automatically generate recommendations for the best franchised repair shops; S4. The system detects abnormal collaborative behaviors through historical quotation volatility, identifies and prevents collusive quotation behaviors among franchised repair shops; for franchised repair shops suspected of collusive quotation, the system marks them as high-risk and pays attention to them in subsequent quotations; S5. After the repair is completed, accept the repair result. After confirmation, complete the order settlement.

2. The intelligent management method for the electric vehicle survey task based on multi-dimensional data collaboration according to claim 1, wherein, The said S1 includes: S11. Based on the RBAC model, dynamically allocate the permissions of the surveyor and the franchised repair shop; S12. Bind the real-time positions of the franchised repair shop and the surveyor; S13. Based on the reinforcement learning model, dynamically calculate the optimal surveyor allocation plan; S14. When the order administrator creates a new order, accurately match the distribution of franchised repair shops; S15. When the surveyor uploads data through the APP, the system's built-in image recognition module automatically checks the integrity of the vehicle photos and extracts key fields; S16. Adopt a structured form, and the system automatically generates repair priority labels.

3. The intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration according to claim 1, characterized in that The said S2 includes: S21. Input features in the XGBoost integrated learning model and output results; S22. Through SHAP value analysis, quantify the influence weights of each feature on the quotation; S23. Adopt a sliding window mechanism to statistically analyze the deviation distribution of the quotations of franchised repair shops from the system benchmark price in real time, and dynamically adjust the floating threshold through Monte Carlo simulation; S24. For quotations exceeding the threshold, trigger an abnormal mark and associate it with the anti-collusion detection step in S4; S25. Combine OCR technology to analyze the pictures of the parts list uploaded by the franchised repair shop, compare them with the system parts database, and verify the rationality of the quotation.

4. The intelligent management method for the electric vehicle survey task based on multi-dimensional data collaboration according to claim 3, wherein, The said S23 includes: Set the size of the sliding window according to historical data analysis and business requirements; and select the sliding step size; For each franchised repair shop, calculate the relative deviation or absolute deviation between its quotation and the system benchmark price; Within the sliding window, statistically analyze the deviation distribution of all franchised repair shops, and generate a large number of possible deviation distribution samples based on historical deviation data using the Monte Carlo simulation method; For each floating threshold, calculate the probability of false screening at this threshold; by comparing the false screening probabilities under different thresholds, select the threshold that minimizes the false screening probability as the current optimal floating threshold; According to external factors, regularly re-run the Monte Carlo simulation and dynamically adjust the floating threshold to adapt to the new market environment.

5. The intelligent management method for the electric vehicle survey task based on multi-dimensional data collaboration according to claim 1, wherein, The said S3 includes: S31. Transmit the quotation data of the franchised repair shop through the symmetric encryption algorithm, and the key is dynamically distributed through the SM2 asymmetric encryption algorithm; S32. Store the quotation hash value on the alliance chain; S33. Divide the geographical grid based on the GeoHash algorithm, screen the franchised repair shops within the preset range, and calculate the expected repair time based on real-time traffic data; S34. Build a maintenance timeliness prediction model, input historical maintenance duration, weather data, and traffic flow, and output a time confidence interval; S35. Define five-dimensional indicators; S36. Use the NSGA-II multi-objective genetic algorithm to generate a Pareto optimal solution set, and the system recommends the top 3 candidate franchise maintenance shops and visually displays the scores of each dimension.

6. The intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration according to claim 1, characterized in that, The above S4 includes: S41. Based on time series analysis, calculate the volatility standard deviation of the quotes of franchise maintenance shops; S42. Based on a graph neural network, construct a relationship topology graph of franchise maintenance shops, and detect abnormal collaboration patterns through node embedding features; S43. Assign dynamic credit scores to high-risk franchise maintenance shops; S44. Update the detection rules in real time by online learning algorithms to fuse new data.

7. The intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration according to claim 6, wherein, The above S42 includes: S421. Collect corresponding data between franchise maintenance shops and preprocess the collected corresponding data; S422. Regard each franchise maintenance shop as a node in the graph, and the node contains the basic information of the franchise maintenance shop; define the edges between nodes according to the relevant factors between franchise maintenance shops; among them, the weight of the edge represents the association strength between franchise maintenance shops; S423. Use relevant knowledge of graph theory to combine nodes and edges into a relationship topology graph of franchise maintenance shops; S424. Extract initial features for each node, and use the extracted node features and the structural information of the topology graph to train the GNN model; S425. Through the trained GNN model, generate embedding vectors for each node; and perform dimensionality reduction processing on the generated embedding vectors, and visualize the distribution of nodes in a two-dimensional or three-dimensional space; S426. Calculate the similarity of embedding vectors between nodes through cosine similarity, and based on the similarity between nodes, use a clustering algorithm to detect small groups with similar frequent quotes; S427. Combine business knowledge and the expert knowledge base to identify abnormal collaboration patterns; and timely feedback the detection results to the system; at the same time, use the detection results as the input of the online learning algorithm to update the detection rules in real time.

8. The intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration according to claim 7, characterized in that, The above S426 includes: Before calculating the similarity, standardize the generated embedding vectors. For each pair of nodes in the topology graph, calculate the cosine similarity between their embedding vectors through matrix operations or loop traversal; According to the characteristics of the data and the requirements of the clustering task, select a clustering algorithm; and set parameters for the selected clustering algorithm; Take the similarity matrix as the input and apply the selected clustering algorithm for clustering analysis; For each cluster, analyze the nodes within the cluster, and determine whether they belong to small groups with similar frequent quotes by calculating the statistical characteristics of the quote data of the nodes within the cluster; According to the number of nodes within the cluster and the degree of quote similarity, determine the scale of the small group; combine the actual situation of the electric vehicle survey business to analyze whether the quote behavior of the small group conforms to normal logic.

9. The intelligent management method for the electric vehicle survey task based on multi-dimensional data collaboration according to claim 1, wherein, The above S5 includes: S51. Based on the maintenance progress, obtain video clips of the maintenance site in real time, and verify the authenticity of spare part replacement through an object detection model; S52. The system compares the vehicle photos before and after maintenance, calculates the image similarity using a Siamese network. If the difference is lower than the threshold, trigger a system review; S53. The parts list is automatically associated with the repair order. The system calls the knowledge graph to verify the compatibility of parts. For the act of falsely reporting parts, a compensation plan is automatically generated. S54. After passing the acceptance, the system automatically triggers the settlement through the smart contract. The user evaluation data flows back to the store rating module to form a closed-loop optimization of "quotation - repair - feedback".

10. An intelligent management system for electric vehicle survey tasks based on multi-dimensional data collaboration, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and operable on the memory. The processor executes the program to implement the intelligent management method for electric vehicle survey tasks based on multi-dimensional data collaboration as described in any one of claims 1-9.

Images