Loss determination method, device, equipment and storage medium

Abstract

This application provides a damage assessment method, apparatus, device, and storage medium, relating to the field of vehicle technology, which can improve the timeliness of predicting damage assessment amounts and reduce assessment errors. The method includes: acquiring accident-related data of the accident vehicle and acquiring a set of parts prices corresponding to the accident vehicle, wherein the prices of parts in the parts price set are related to the market demand for the parts; extracting features from the accident-related data to obtain damage assessment correlation features of the accident vehicle; and making damage assessment predictions based on the damage assessment correlation features and the prices of damaged parts in the parts price set of the accident vehicle to obtain the predicted damage assessment amount for the accident vehicle.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and more particularly to a damage assessment method, apparatus, device, and storage medium. Background Technology

[0002] Damage assessment is a core part of auto insurance claims and vehicle repair services. Its accuracy and efficiency directly affect the insurance company's cost control, the car owner's claims experience, and the service quality of repair shops. Related techniques typically rely on manual experience and static parts price lists to assess damage to accident vehicles. However, this method is inefficient and prone to significant errors. Summary of the Invention

[0003] The purpose of this application is to provide a damage assessment method, apparatus, device, and storage medium that can obtain real-time spare parts prices, improve the timeliness of predicting damage assessment amounts, and thereby reduce damage assessment errors.

[0004] To achieve the above objectives, this application adopts the following technical solution: Firstly, this application provides a loss assessment method, the method comprising: Obtain accident-related data for the accident vehicle, as well as a price set for the corresponding parts. The prices of the parts in the price set are correlated with market demand. Feature extraction is performed on the accident-related data to obtain the damage assessment correlation features for the accident vehicle. Based on these features and the prices of the damaged parts in the price set, a damage assessment prediction is made to obtain the predicted damage amount for the accident vehicle.

[0005] The technical solution provided in this application obtains real-time, dynamic parts prices by acquiring parts prices related to market demand from a parts price collection. Then, it extracts features from accident-related data to obtain damage assessment-related features. Based on the damage assessment-related features and real-time parts prices, it predicts the damage assessment amount of the accident vehicle to obtain the predicted damage assessment amount, thereby reducing damage assessment errors.

[0006] One possible implementation involves predicting the estimated damage amount of an accident vehicle based on damage assessment correlation features and the prices of damaged parts in the parts price set. Specifically, this can be achieved by inputting the damage assessment correlation features and the prices of damaged parts in the parts price set into an accident loss model for damage assessment prediction. By directly inputting these features into the accident loss model, the predicted damage amount can be obtained, significantly reducing the reliance on manual experience in damage assessment and improving efficiency.

[0007] Another possible implementation method is that the calculation formula for loss assessment prediction using the accident loss model can be expressed as:

[0008] in, To predict the estimated loss amount, where n is the total number of damaged parts. Let be the price of the i-th damaged part in the parts price set. Let be the supply and demand fluctuation coefficient for the i-th damaged component. Let be the transportation cost coefficient for the i-th damaged part. Let be the depreciation rate for the i-th damaged component. Let the damage level weight be the i-th damaged component. As a benchmark for regional labor costs, This is the maintenance man-hour coefficient.

[0009] Another possible implementation is an accident loss model constructed using the XGBoost gradient boosting tree algorithm, the LightGBM gradient boosting tree algorithm, the CatBoost algorithm, and the ensemble decision tree random forest algorithm.

[0010] Another possible implementation of the damage assessment method provided in this application may further include: obtaining the actual damage assessment amount of the accident vehicle; and updating the accident loss model and / or parts price set based on the difference between the actual damage assessment amount and the predicted damage assessment amount. By updating the accident loss model and parts price set based on the difference between the actual damage assessment amount and the predicted damage assessment amount, the accuracy of damage assessment can be improved.

[0011] Another possible approach is to obtain the price set of parts corresponding to the accident vehicle. This can be achieved by: acquiring price correlation information for the parts, which includes at least one of the following: supply chain system data, platform transaction data, actual settlement data from repair shops, raw material market data, or inventory information. Based on this price correlation information, a price set of parts is constructed. This ensures that the parts prices reflect multiple dimensions of factors, including the supply chain, end market, raw materials, and regional differences.

[0012] Another possible implementation, the damage assessment method provided in this application, may further include: updating the parts price set when the data in the parts price set changes in their respective data sources. By updating the parts price set in real time, and then assessing the damage based on the real-time parts prices, the accuracy of damage assessment is improved.

[0013] Another possible implementation, the loss assessment method provided in this application, may further include: data preprocessing of accident-related data, wherein data preprocessing includes at least one of outlier removal, missing value imputation, or standardization. By preprocessing the accident-related data, complete and accurate input data is provided for the accident loss model.

[0014] Another possible implementation method is to include accident-related data such as vehicle accident damage data, vehicle network status data, accident liability data, and regional environmental data.

[0015] Another possible implementation method is to include damage characteristics, vehicle characteristics, market characteristics, and human factors characteristics in the loss assessment.

[0016] Secondly, a damage assessment device is provided, the device comprising: The acquisition module is used to acquire accident-related data of the accident vehicle and to acquire the price set of the parts corresponding to the accident vehicle. The price of the parts in the price set is related to the market demand for the parts. The processing module is used to extract features from accident-related data to obtain damage assessment-related features of the accident vehicles. The prediction module is used to predict the damage assessment amount of the accident vehicle based on the damage assessment correlation characteristics and the price of damaged parts in the parts price set.

[0017] In one possible implementation, the prediction module is also used to: input the damage assessment correlation features and the price of damaged parts in the parts price set into the accident loss model to predict the damage assessment amount.

[0018] Another possible implementation method is that the calculation formula for loss assessment prediction using the accident loss model can be expressed as:

[0019] in, To predict the estimated loss amount, where n is the total number of damaged parts. Let be the price of the i-th damaged part in the parts price set. Let be the supply and demand fluctuation coefficient for the i-th damaged component. Let be the transportation cost coefficient for the i-th damaged part. Let be the depreciation rate for the i-th damaged component. Let the damage level weight be the i-th damaged component. As a benchmark for regional labor costs, This is the maintenance man-hour coefficient.

[0020] Another possible implementation is an accident loss model constructed using the XGBoost gradient boosting tree algorithm, the LightGBM gradient boosting tree algorithm, the CatBoost algorithm, and the ensemble decision tree random forest algorithm.

[0021] In another possible implementation, the prediction module is also used to: obtain the actual damage assessment amount for the accident vehicle; and update the accident loss model and / or parts price set based on the difference between the actual damage assessment amount and the predicted damage assessment amount.

[0022] Another possible implementation involves the acquisition module also acquiring price-related information for spare parts. This price-related information includes at least one of the following: supply chain system data, platform transaction data, actual settlement data from repair shops, raw material market data, or inventory information. Based on this price-related information, a set of spare parts prices is constructed.

[0023] Another possible implementation is that the acquisition module is also used to update the parts price set when the data in the parts price set changes in their respective data sources.

[0024] In another possible implementation, the acquisition module is also used to: perform data preprocessing on accident-related data, including at least one of outlier removal, missing value imputation, or standardization.

[0025] Another possible implementation method is to include accident-related data such as vehicle accident damage data, vehicle network status data, accident liability data, and regional environmental data.

[0026] Another possible implementation method is to include damage characteristics, vehicle characteristics, market characteristics, and human factors characteristics in the loss assessment.

[0027] The technical effects of any implementation method in the second aspect can be found in the technical effects of any implementation method in the first aspect mentioned above, and will not be repeated here.

[0028] Thirdly, a computer device is provided, comprising: a processor and a memory, wherein the memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor to implement the damage assessment method described above.

[0029] Fourthly, a computer-readable storage medium is provided, wherein at least one computer program is stored in the computer-readable storage medium, and the at least one computer program is loaded and executed by a processor to implement the damage assessment method described above.

[0030] The solutions provided in the third and fourth aspects above are used to implement the method provided in the first aspect above, and their specific implementations will not be elaborated further. The technical effects corresponding to any implementation method in the solutions provided in the third and fourth aspects above can be found in the technical effects corresponding to any implementation method in the first aspect above, and will not be elaborated further here.

[0031] It should be noted that any of the possible implementations of any of the above aspects can be combined, provided that the solutions do not contradict each other. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 A schematic diagram of the system architecture of a damage assessment method provided in an embodiment of this application; Figure 2 A schematic flowchart of a damage assessment method provided in an embodiment of this application; Figure 3 A schematic flowchart illustrating another damage assessment method provided in this application embodiment; Figure 4 A schematic flowchart illustrating another damage assessment method provided in this application embodiment; Figure 5 This is a schematic diagram of the structure of a damage assessment system provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of a data acquisition module provided in an embodiment of this application; Figure 7 A schematic diagram of the architecture of an accessory price set provided in an embodiment of this application; Figure 8 A principle block diagram of an accident loss model provided in an embodiment of this application; Figure 9 A schematic diagram illustrating a loss assessment calculation logic provided in an embodiment of this application; Figure 10 This is a schematic diagram of the structure of a damage assessment device provided in an embodiment of this application; Figure 11 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0034] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] In the description of this application, it should be understood that the terms "upper," "lower," "left," "right," "front," "rear," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or relative positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and for simplification, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. Unless otherwise specified, the above-mentioned orientational descriptions can be flexibly set in practical applications, provided that the relative positional relationships shown in the accompanying drawings are satisfied.

[0036] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0037] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," and "communication" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection. They can refer to a direct connection or an indirect connection through an intermediate medium, or a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0038] In embodiments of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, article, or apparatus that includes that element.

[0039] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0040] In the embodiments of this application, at least one can also be described as one or more, and multiple can be two, three, four or more, and this application does not impose any restrictions.

[0041] In the description of this specification, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0042] It should be noted that all information (including but not limited to equipment information, personal information of the target), data (including but not limited to data used for analysis, stored data, and displayed data), and signals involved in this application have been authorized by the target or fully authorized by all parties, and the collection, use, and processing of related data must comply with relevant laws, regulations, and standards. For example, accident-related data, parts price lists, etc., involved in this application were obtained with full authorization.

[0043] The industry-standard loss assessment method mainly adopts a combination of static price lists and manual adjustments, which will be briefly explained below.

[0044] Pricing is based on the original manufacturer's parts price list, which is typically updated every 30 days. Basic information such as the damaged parts and types of damage to the accident vehicle is collected, and combined with fixed regional labor cost standards, a preliminary damage assessment is calculated using a simple formula. Finally, the damage assessment amount is manually adjusted by the assessment personnel based on their experience.

[0045] However, the above technical solutions rely on the experience of loss adjusters to adjust the loss assessment amount, which is inefficient and easily affected by subjective factors. In addition, the price update cycle of parts is long, and it is impossible to reflect the changes in parts prices in real time. This results in a large difference between the price of parts used in the loss assessment and the actual repair cost, thus leading to a large error in the loss assessment.

[0046] Based on this, this application provides a damage assessment method, which obtains real-time and dynamic parts prices by acquiring parts prices related to market demand from parts price sets, then extracts features from accident-related data to obtain damage assessment correlation features, and predicts the damage assessment amount of the accident vehicle based on the damage assessment correlation features and real-time parts prices, thereby reducing damage assessment errors.

[0047] The solutions provided by the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0048] Figure 1 A schematic diagram of the system architecture of a damage assessment method provided in this application embodiment is shown below. Figure 1 As shown, the system includes: a computer device 101, which can be implemented as a terminal or a server.

[0049] When the computer device 101 is implemented as a server, the computer device 101 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0050] When the computer device 101 is implemented as a terminal, the computer device 101 can be a smartphone, tablet computer, laptop computer, desktop computer, etc.

[0051] Optionally, the system described above includes one or more computer devices 101. This application embodiment does not limit the number of computer devices 101.

[0052] Specifically, in Figure 1 In the illustrated system architecture, computer device 101 acquires accident-related data of the accident vehicle and the price set of corresponding parts for the accident vehicle. Then, it extracts features from the accident-related data to obtain the damage assessment-related features of the accident vehicle. Finally, based on the damage assessment-related features and the prices of damaged parts in the parts price set, it performs damage assessment prediction to obtain the predicted damage assessment amount for the accident vehicle. The specific implementation of computer device 101 is described in the following method embodiment.

[0053] Figure 2 This is a schematic flowchart of a damage assessment method provided in an embodiment of this application. The method can be executed by a computer device. The computer device can be, for example,... Figure 1 The computer device 101 shown is shown.

[0054] like Figure 2 As shown, the damage assessment method provided in this application embodiment may include: Step S202: The computer equipment obtains the accident-related data of the accident vehicle and the price set of the corresponding parts for the accident vehicle.

[0055] Among them, accident-related data is used to indicate data related to accidents involving the vehicles involved in the accidents.

[0056] Optionally, accident-related data includes, but is not limited to, vehicle accident damage data, vehicle network status data, accident liability data, and regional environmental data.

[0057] Vehicle accident damage data describes the damage to the vehicles involved in the accident. For example, vehicle accident damage data may include the damaged parts of the vehicle, the degree of damage (e.g., minor, moderate, severe), the type of component damage (e.g., fracture, deformation, wear), and a list of related damaged components.

[0058] Specifically, computer equipment can acquire vehicle accident damage data by collecting data from vehicle-to-everything (V2X) collision sensors, dashcam footage, and repair shop damage assessment photos. Based on this data, the damage status of the accident vehicle can be accurately described, providing a data basis for subsequent damage degree weighting, parts replacement, and parts repair.

[0059] For new energy vehicles (pure electric / hybrid), computer equipment can also obtain battery pack information and motor / electronic control system information, and determine the degree of damage to the accident vehicle based on the battery pack information and motor / electronic control system information.

[0060] Specifically, the methods by which computer equipment determines the degree of damage to new energy vehicles include any of the following, but are not limited thereto, and the embodiments of this application do not specifically limit this.

[0061] Method 1: Computer equipment can acquire the battery pack voltage data of the accident vehicle and use vehicle network self-check data such as battery pack voltage stability and cell consistency as the basis for determining the degree of damage to the accident vehicle. For example, if the battery pack voltage fluctuation of the accident vehicle exceeds ±5%, the degree of damage to the accident vehicle will be determined as moderate damage.

[0062] Method 2: Computer equipment can also acquire post-accident operational status data of the vehicle's motor / electronic control system, such as motor speed stability and electronic control signal transmission delay. This data can then be used to help determine the extent of damage to the vehicle, avoiding errors caused by assessing damage solely based on its appearance, thus improving accuracy and reducing subsequent repair risks.

[0063] Method 3: Computer equipment can also acquire image features of the battery pack of the accident vehicle (such as images of battery pack shell deformation), and combine image features and data features (such as voltage stability data) to determine the damage to the accident vehicle, thereby improving the accuracy of damage determination.

[0064] Vehicle network status data is used to describe the operational status information of the vehicle involved in the accident.

[0065] Specifically, computer equipment can acquire vehicle network status data by collecting information such as the age of the accident vehicle, mileage, collision intensity (e.g., acceleration value), collision angle, and post-accident component self-inspection data (e.g., battery pack voltage, motor status).

[0066] Accident liability data is used to indicate relevant data on various factors that cause an accident involving a vehicle. For example, accident liability data may include the percentage of responsibility of the vehicle involved in the accident (such as full responsibility, primary and secondary responsibility, no responsibility), and rules for exemption from liability.

[0067] Specifically, computer devices can connect to insurance company systems via Application Programming Interfaces (APIs) to obtain accident liability data.

[0068] Regional environmental data is used to describe labor and parts costs in the area where the accident vehicle is located, such as regional labor cost benchmarks, parts transportation cost coefficients, and the degree of supply and demand tension in the region.

[0069] Specifically, computer equipment can obtain regional environmental data by connecting to data from regional maintenance industry associations.

[0070] The parts price set is used to indicate a data set that integrates parts prices from different data sources. The prices of parts in the parts price set are related to the market demand for the parts.

[0071] In some embodiments, the computer device obtains price association information of accessories, and then constructs a set of accessory prices based on the price association information.

[0072] Optionally, the price-related information may include at least one of the following: supply chain system data, platform transaction data, actual settlement data of repair shops, raw material market data, or inventory information.

[0073] Supply chain system data, also known as original equipment manufacturer (OEM) parts supply chain system data, ensures the accuracy of OEM parts price benchmarks by obtaining real-time ex-warehouse prices for parts.

[0074] The platform's transaction data can be transaction data from third-party parts e-commerce platforms, such as the unit price and transaction volume of transactions. The platform's transaction data can reflect the prices of parts in the end market.

[0075] The end market refers to sales venues that directly face consumers (such as owners of vehicles involved in accidents).

[0076] The actual settlement data of the repair shop can be the sum of the parts purchase price and gross profit of the authorized repair shop. The actual settlement data of the repair shop can reflect the actual cost of the repair scenario.

[0077] Raw material market data can include futures market data for raw materials such as metals and plastics, including market prices for copper, aluminum, lithium, and cobalt. This data can provide a basis for understanding price fluctuations in metal components and new energy vehicle battery packs.

[0078] Inventory information can include regional parts inventory data, such as the quantity of parts in stock and inventory turnover days in the city or region where the accident vehicle was located. This inventory information reflects the degree of supply and demand tension in a region, providing a basis for calculating the supply and demand fluctuation coefficient.

[0079] Furthermore, the parts price set is updated when the data in the parts price set changes in their respective data sources.

[0080] Specifically, when data from sources such as supply chain system data, platform transaction data, repair shop actual settlement data, raw material market data, or inventory information changes, the computer equipment can update the parts price set based on these changes. For example, when the inventory quantity of parts in the city where the accident vehicle is located changes, the corresponding inventory information for the parts in the parts price set is updated synchronously; when the unit price of parts collected on a third-party parts platform changes (such as an increase or decrease), the platform transaction data for the corresponding parts in the parts price set is updated. By updating the parts price set, it is possible to quickly respond to short-term fluctuations in parts ex-factory prices and inventory supply and demand.

[0081] For new energy vehicles, computer equipment can incorporate specialized sub-models for core components of new energy vehicles into deep learning models, taking into account lithium and cobalt futures data, battery pack self-test timing data, and motor operating status data, and outputting the cost of core components through specialized sub-models.

[0082] By designing a dedicated price fluctuation model and damage assessment rules for new energy vehicles, the system meets the precise damage assessment needs of these vehicles.

[0083] Step S204: The computer equipment extracts features from the accident-related data to obtain the damage assessment-related features of the accident vehicle.

[0084] Among them, damage assessment related features refer to features associated with the damage assessment results of the accident vehicle. Optionally, damage assessment related features include, but are not limited to, damage features, vehicle features, market features, and human factors features.

[0085] Damage characteristics may include damage severity weights and component breakage type coefficients. The damage severity weight reflects the impact of damage on the value of the parts. For example, when the damage to the accident vehicle is minor, the damage severity weight is 0.3; when the damage to the accident vehicle is moderate, the damage severity weight is 0.6; and when the damage to the accident vehicle is severe, the damage severity weight is 1.0.

[0086] Specifically, damage features can be obtained by extracting features from vehicle accident damage data in accident-related data.

[0087] Vehicle characteristics may include a depreciation rate. The depreciation rate reflects the wear and tear on an accident-damaged vehicle. For example, the depreciation rate can be calculated by multiplying the vehicle's age by 10%, with a minimum depreciation rate of 0.5. For instance, if the vehicle's age is 6 years, the depreciation rate is 6 × 10% = 0.6; if the vehicle's age is 3 years, the depreciation rate is calculated as age × 10%, resulting in 0.3. However, since the minimum depreciation rate is 0.5, the actual depreciation rate is 0.5.

[0088] Specifically, vehicle characteristics can be obtained by extracting features from the vehicle network status data in the accident-related data.

[0089] Market characteristics may include regional supply and demand volatility coefficients, metal futures correlation coefficients, and transportation cost coefficients. The regional supply and demand volatility coefficient reflects the impact of market supply and demand, while the transportation cost coefficient reflects regional logistical differences.

[0090] Specifically, by extracting features from regional environmental data in accident-related data, market characteristics (regional supply and demand fluctuation coefficient, transportation cost coefficient) can be obtained.

[0091] Labor characteristics may include regional labor cost benchmarks and repair time coefficients. The repair time coefficient is determined based on the type of repair and is calibrated using industry repair standards and historical data. For example, the repair time coefficient for replacing a headlight is 0.5 hours; for sheet metal repair, it is 1.2 hours; and for battery pack casing repair, it is 2.0 hours.

[0092] Specifically, artificial features can be obtained by extracting features from regional environmental data in accident-related data.

[0093] In addition, loss assessment related characteristics can also include price characteristics and auxiliary characteristics.

[0094] Price features may include real-time prices for accessories. These real-time prices are derived from the accessory price set.

[0095] Specifically, price features can be obtained by extracting features from the data of component price sets.

[0096] Auxiliary features may include the warranty period coefficient for parts and the collision intensity coefficient for accidents.

[0097] Specifically, by extracting features from the vehicle network status data in the accident-related data, auxiliary features (accident collision intensity coefficient) can be obtained; by extracting features from the data of parts price sets, auxiliary features (parts warranty period coefficient) can be obtained.

[0098] Furthermore, by extracting features from the data on component prices, market characteristics (correlation coefficients of metal futures) can be obtained.

[0099] The aforementioned loss assessment correlation features cover five core dimensions: price, damage, vehicle, market, and labor. This ensures that the accident loss model can comprehensively capture the key factors affecting the predicted loss assessment amount and avoid prediction bias caused by the lack of loss assessment correlation features.

[0100] Step S206: The computer equipment performs damage assessment prediction based on the damage assessment correlation characteristics and the price of damaged parts in the parts price set of the accident vehicle, and obtains the predicted damage assessment amount of the accident vehicle.

[0101] In one possible implementation, the computer equipment inputs the damage assessment correlation characteristics and the price of damaged parts in the parts price set into the accident loss model to predict the damage assessment amount.

[0102] The accident loss model is constructed using the XGBoost algorithm, the LightGBM algorithm, the CatBoost algorithm, and the Random Forest algorithm. This application's embodiment of the accident loss model supports multiple gradient boosting algorithms, allowing the selection of the optimal model based on the scenario.

[0103] XGBoost has strong nonlinear fitting capabilities and can effectively handle the interaction of multiple loss assessment features (such as the linkage between real-time prices of spare parts and regional supply and demand coefficients). Compared with traditional linear models and single decision tree models, the XGBoost algorithm has stronger generalization capabilities and can adapt to complex loss assessment scenarios.

[0104] LightGBM employs techniques such as histogram optimization and gradient unilateral sampling, resulting in faster training speed and lower memory usage. It also supports multi-loss assessment related feature interaction and complex scene fitting, adapting to the high-dimensional and heterogeneous characteristics of loss assessment data, and outputting accurate parts costs and predicted loss assessment amounts.

[0105] CatBoost can automatically process category features (such as accessory type and damage level) without the need for additional one-hot encoding, reducing preprocessing steps. It also has strong generalization ability and can effectively capture the impact of damage assessment correlation features such as dynamic price and regional coefficient on the predicted damage assessment amount.

[0106] The random forest algorithm outputs results through voting among multiple decision trees, exhibiting strong stability and outstanding resistance to overfitting. By adjusting the number and depth of decision trees, it can adapt to the feature complexity of the damage assessment scenario and also achieve comprehensive calculation of multi-dimensional damage assessment related features.

[0107] The XGBoost, LightGBM, CatBoost, and Random Forest algorithms all possess strong nonlinear fitting capabilities and can perform complex loss assessment correlation feature interactions and nonlinear calculations.

[0108] In some embodiments, the calculation formula for loss assessment prediction using the accident loss model can be expressed as: .

[0109] in, To predict the estimated loss amount, where n is the total number of damaged parts. Let be the price of the i-th damaged part in the parts price set. Let be the supply and demand fluctuation coefficient for the i-th damaged component. Let be the transportation cost coefficient for the i-th damaged part. Let be the depreciation rate for the i-th damaged component. Let the damage level weight be the i-th damaged component. As a benchmark for regional labor costs, This is the maintenance man-hour coefficient.

[0110] The above calculation formula is clear and the parameters are quantifiable. Those skilled in the art can directly write program code based on the formula to achieve automated calculation.

[0111] Specifically, the computer equipment can match the listed prices of corresponding damaged parts from the parts price list based on the list of damaged parts and the Vehicle Identification Number (VIN) of the accident vehicle. Based on the listed prices and the damage assessment characteristics of the accident vehicle, the unit cost of all damaged parts is calculated.

[0112] The cost per component is the cost of repairing or replacing a single damaged component.

[0113] Specifically, the cost of a single component of a computer device can be calculated using the following formula: Cost per component = .

[0114] in, Let be the price of the i-th damaged part in the parts price set. Let be the supply and demand fluctuation coefficient for the i-th damaged component. Let be the transportation cost coefficient for the i-th damaged part. Let be the depreciation rate for the i-th damaged component. Let be the damage weight of the i-th damaged component.

[0115] By combining spare parts prices with multi-dimensional loss assessment correlation characteristics, the accuracy of single spare parts cost accounting is ensured.

[0116] After calculating the cost of a single component, the costs of all damaged components are summed to obtain the total cost of each component (i.e., the total cost of all components).

[0117] The computer equipment can then calculate the total labor cost using the following formula: Total labor cost = L × R.

[0118] Where L is the regional labor cost benchmark and R is the maintenance man-hour coefficient.

[0119] The total labor cost is calculated by combining the regional labor cost benchmark with the maintenance time coefficient. This method can adapt to the differences in labor costs in different maintenance scenarios and avoid the deviation in labor cost accounting caused by uniform working hours.

[0120] After obtaining the total cost of individual parts and the total labor cost, the preliminary loss assessment amount is calculated by summing the total cost of individual parts and the total labor cost.

[0121] Finally, based on the accident liability ratio (e.g., full liability multiplied by 100%, secondary liability multiplied by 30%) and the insurance company's deductible rules, the preliminary loss assessment amount is adjusted to obtain the predicted loss assessment amount. This ensures that the predicted loss assessment amount aligns with the actual business rules of insurance claims, guaranteeing its feasibility and direct integration with the claims process.

[0122] In summary, by obtaining parts prices related to market demand from parts price collections, real-time and dynamic parts prices can be obtained. Then, feature extraction is performed on accident-related data to obtain damage assessment correlation features. Based on the damage assessment correlation features and real-time parts prices, the damage assessment amount of accident vehicles can be predicted, thereby reducing damage assessment errors.

[0123] After the computer device obtains the accident-related data of the accident vehicle, the computer device can also perform data preprocessing on the accident-related data, that is, step S203 is included between step S202 and step S204.

[0124] Figure 3 This is a schematic flowchart of another damage assessment method provided in an embodiment of this application. This method can be executed by a computer device.

[0125] Step S203: The computer equipment performs data preprocessing on the accident-related data. Data preprocessing includes at least one of outlier removal, missing value filling, or standardization.

[0126] Outlier removal refers to the identification and removal of abnormal data in accident-related data and parts price clusters. For example, parts prices may be clustered at prices that are 10 times higher than the industry average.

[0127] Specifically, the computer equipment may perform outlier removal on accident-related data and parts price data in any of the following ways, but is not limited thereto, and this application embodiment does not specifically limit this.

[0128] Method 1: Computer equipment can use the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm to identify and remove logically contradictory abnormal data (such as accident-related data where the collision intensity is 0 but serious damage is reported, or parts prices that are concentrated and far exceed the industry average by 10 times).

[0129] Method 2: Computer equipment can use the Interquartile Range (IQR) rule to remove outliers from accident-related data and parts price sets. Specifically, outliers can be filtered using Q3 + 1.5IQR and Q1 - 1.5IQR. The IQR rule does not require a pre-defined data distribution.

[0130] Method 3: Computer equipment can use the Z-score thresholding method to remove outliers from accident-related data and spare parts price sets. The Z-score thresholding method measures the degree of deviation of a data point from the population mean, with the unit being standard deviation. Specifically, the threshold can be set to 3 standard deviations; values ​​deviating from the mean by 3 standard deviations are considered outliers and removed. The Z-score thresholding method is suitable for normally distributed data.

[0131] By removing outliers from accident-related data and parts price sets, invalid and erroneous data can be filtered out, ensuring the reliability of the input feature set of the accident loss model and preventing outliers from interfering with the training and prediction accuracy of the accident loss model.

[0132] Missing value imputation refers to filling in missing data in accident-related data and parts price sets. For example, the parts price set may lack inventory information for parts (such as headlights).

[0133] Specifically, the computer equipment may fill in missing values ​​for accident-related data and parts price data in any of the following ways, but is not limited thereto, and the embodiments of this application do not specifically limit this.

[0134] Method 1: Computer equipment can use the K-Nearest Neighbors (KNN) algorithm to fill in missing parts prices or missing damage parameters in accident-related data based on historical data of similar vehicle models and similar damage types to the accident vehicle.

[0135] Method 2: Computer equipment can use the mean / median imputation method to fill in missing data in accident-related data and parts price sets. This method is suitable for scenarios where the accident-related data and parts price sets are evenly distributed.

[0136] Method 3: Computer equipment can use the random forest imputation method to fill in the missing data in the accident-related data and parts price set. The random forest imputation method predicts missing values ​​based on the correlation between features, thereby filling in the missing values ​​in the accident-related data and parts price set. This method is suitable for scenarios with complex feature correlations.

[0137] By filling in missing values ​​in accident-related data and parts price data, the problem of incomplete data collection can be solved, and the bias in model prediction caused by missing values ​​can be avoided.

[0138] Standardization is a process of converting data of different dimensions or magnitudes into a unified standard to facilitate data comparison, analysis, and weighting.

[0139] Accident-related data and parts price data collection include both continuous and discrete data, and computer equipment uses different methods to standardize continuous and discrete data.

[0140] For continuous data such as mileage and service life, computer equipment can map the continuous data to the [0, 1] interval through standardization using StandardScore (Z-Score), Min-Max standardization, and StandardScaler, thereby achieving standardization of continuous data and eliminating the impact of dimensional differences on accident loss models.

[0141] For discrete data such as damage level and component type, computer equipment can perform one-hot encoding on the discrete data to transform it into numerical features that the model can recognize, thereby achieving standardized processing of discrete data.

[0142] One-hot encoding is a technique that converts categorical variables into binary vectors. For example, if the degree of damage includes three levels: minor damage, moderate damage, and severe damage, one-hot encoding of the degree of damage can yield the following encoding results: minor damage is [1, 0, 0], moderate damage is [0, 1, 0], and severe damage is [0, 0, 1].

[0143] By standardizing accident-related data and parts price data, we can unify the data format and scale, ensuring that multi-source heterogeneous data can be collaboratively input into the accident loss model, thereby improving the training efficiency and prediction accuracy of the accident loss model.

[0144] In summary, by preprocessing accident-related data and parts price data, we ensure that the data input into the accident loss model meets the model's input requirements. Furthermore, by automatically performing outlier removal, missing value imputation, and standardization through algorithms, we can avoid the subjectivity and inefficiency of manual screening.

[0145] Furthermore, after obtaining the predicted damage assessment amount for the accident vehicle, the computer equipment can also obtain the actual damage assessment amount for the accident vehicle, and then update the accident loss model and / or parts price set based on the difference between the actual damage assessment amount and the predicted damage assessment amount. That is, after step S206, steps S207 and S208 are also included.

[0146] Figure 4 This is a schematic flowchart of another damage assessment method provided in an embodiment of this application. This method can be executed by a computer device.

[0147] Step S207: The computer equipment obtains the actual damage assessment amount of the accident vehicle.

[0148] The actual damage assessment amount refers to the amount that the insurance company actually pays to the vehicle involved in the accident.

[0149] In some embodiments, the actual loss assessment amount may also be referred to as the actual compensation amount. This application does not limit the name of the actual loss assessment amount.

[0150] Specifically, computer equipment can obtain the actual damage assessment amount for accident vehicles by connecting to the insurance company's system.

[0151] Step S208: The computer equipment updates the accident loss model and / or spare parts price set based on the difference between the actual loss assessment amount and the predicted loss assessment amount.

[0152] In one possible implementation, after the computer equipment obtains the predicted and actual damage assessment amounts for the accident vehicles, it calculates the difference between the predicted and actual damage assessment amounts, and updates the accident loss model based on this difference.

[0153] Specifically, after the computer equipment obtains the predicted damage assessment amount, it acquires the actual damage assessment amount of the accident vehicle and calculates the difference between the predicted damage assessment amount and the actual damage assessment amount. Then, the difference between the predicted damage assessment amount and the actual damage assessment amount is marked as the optimization target. Using historical damage assessment data as samples, the weight allocation of damage assessment-related features is adjusted through a Bayesian optimization algorithm, with a focus on optimizing the weights of real-time parts prices and regional supply and demand coefficients.

[0154] The iteration cycle of the weights of the loss assessment associated features can be set according to actual needs. For example, it can be set to iterate once every 7 days. Each iteration can be based on the loss assessment feedback data of the past 30 days. This application does not limit this.

[0155] Damage assessment feedback data can include the actual damage assessment amount, user complaint records, and feedback from repair shops.

[0156] By dynamically updating the weights of damage assessment-related features based on the Bayesian optimization algorithm, the traditional method of manually setting weights is replaced. This enables the model to learn autonomously and adapt dynamically, and can cope with scenarios such as market price fluctuations, the launch of new parts, and changes in damage types without the need for technical personnel intervention, thereby continuously improving the accuracy of damage assessment.

[0157] In addition, computer equipment can also dynamically update the weights of loss assessment related features in the following ways.

[0158] Method 1: Computer equipment can dynamically update the weights of loss assessment-related features through grid search. Specifically, a candidate range (e.g., 0.3-0.6) and step size for the weights of loss assessment-related features are preset. All combinations are traversed, and the optimal weights are selected based on historical loss assessment deviations (i.e., the deviation between historical predicted loss assessment amounts and actual loss assessment amounts). This method requires no complex probability models, is simple and intuitive, and can complete a weight iteration every 7 days, ensuring the adaptability of the accident loss model.

[0159] Method 2: Computer equipment can dynamically update the weights of loss assessment-related features through random search. Specifically, random sampling combinations are performed within the candidate weight interval, and the optimal solution is selected through statistical significance. This method is more efficient than grid search, suitable for optimizing high-dimensional feature weights, and can quickly find a weight allocation method that fits the current market for loss assessment-related features.

[0160] Method 3: Computer equipment can dynamically update the weights of damage assessment-related features using a genetic algorithm. Specifically, by simulating the selection, crossover, and mutation processes of biological evolution, and using the minimization of damage assessment deviation as the fitness function, the weights of damage assessment-related features are iteratively optimized. This method has strong global search capabilities and is suitable for handling complex changes such as sudden market price fluctuations and the launch of new parts.

[0161] For new energy vehicles, computer equipment can dynamically update the weights of damage assessment-related features in the following way.

[0162] Method 1: Computer equipment can adjust the weighting of the components based on the battery pack and motor / electronic control system of the accident vehicle. For example, the battery pack price is linked to lithium and cobalt futures data, increasing the weighting of the metal futures correlation coefficient to 0.6. The motor / electronic control system price is linked to copper and aluminum futures data, setting the weighting of the metal futures correlation coefficient to 0.4, adapting to cost fluctuations in metal components such as motor windings and electronic control housings. By designing exclusive weighting rules based on the characteristics of core components of new energy vehicles, the price drivers of core components of new energy vehicles can be accurately captured, avoiding the damage assessment errors caused by using the pricing logic of traditional fuel vehicles.

[0163] Method 2: In the accident loss model, set up independent loss assessment related feature branches for battery pack, motor, and electronic control, and input raw material futures data (such as lithium, cobalt, and copper), component self-inspection data, and add a new energy vehicle-specific historical loss assessment dataset (including core component damage cases) during model training. Optimize the weights of loss assessment related features through specialized training. Without designing a separate price fluctuation model, accurate loss assessment of core components can still be achieved.

[0164] Furthermore, computer equipment can reconstruct the model annually by incorporating data on new components (such as new electronic control components for new energy vehicles), new vehicle models (such as newly added hybrid models), and new damage types (such as damage related to intelligent driving systems), updating the feature set and algorithm parameters. This ensures that the accident loss model can adapt to industry technological developments (such as new vehicle technologies and new components) over the long term, preventing model aging from leading to a decline in damage assessment accuracy.

[0165] By updating the accident loss model, we can overcome the shortcomings of existing models where feature weights are manually set and cannot be dynamically adjusted. We can build a weight self-learning mechanism to enable the model to automatically iterate and optimize based on historical loss assessment deviations, the launch of new parts, market changes, and other factors, thereby continuously improving the accuracy of loss assessment.

[0166] In another possible implementation, after the computer equipment obtains the predicted and actual damage assessment amounts for the accident vehicle, it calculates the difference between the predicted and actual damage assessment amounts, and updates the parts price set based on the difference between the predicted and actual damage assessment amounts.

[0167] Specifically, computer equipment can update the accessory price set in the following ways.

[0168] Method 1: Computer equipment can update the parts price set by combining real-time incremental updates, daily full verification, and manual review of anomalies.

[0169] Specifically, the computer equipment can retrieve supply chain system data at fixed intervals (e.g., every 2 hours) to ensure the real-time availability of prices for original equipment manufacturer (OEM) parts and core components. It can also retrieve price quotes from e-commerce platforms and repair shops at fixed intervals (e.g., early morning each day), compare these quotes, and adjust the prices of components in the price set based on the comparison results to correct price discrepancies and ensure the accuracy of the pricing in the price set. Furthermore, the computer equipment can calculate the price difference between the current component price and the historical average component price. When the price difference exceeds a threshold (e.g., ±20%), manual intervention is triggered to verify the price's reasonableness (e.g., whether it's a new component or if there's a supply chain shortage).

[0170] Method Two: Computer equipment can update the parts price set by combining high-frequency incremental updates with time-sharing full verification. Specifically, the computer equipment can incrementally synchronize core dynamic data sources such as the supply chain and futures every hour to improve the timeliness of the parts price set data. Every 12 hours, it can perform full verification of quotations from e-commerce platforms and repair shops to reduce synchronization costs. Abnormal prices will still trigger manual review if they deviate from the average by ±20%, thus enabling the parts price set to respond to market fluctuations in real time.

[0171] Method 3: Computer equipment can update the parts price set by combining tiered incremental updates of parts prices with multiple rounds of full verification daily. Specifically, the data sources are divided into core types (supply chain, futures data) and auxiliary types (e-commerce, inventory data). Core data sources are incrementally synchronized every 3 hours, and auxiliary data sources are incrementally synchronized every 6 hours. Full verification is performed on both core and auxiliary data sources once each in the morning and evening to ensure that the update efficiency and accuracy of different types of data are matched.

[0172] By updating the parts price list, we can ensure that the parts price list is in sync with the market in the long term, and avoid loss assessment errors caused by the lag in parts prices.

[0173] Figure 5 This is a schematic diagram of the structure of a damage assessment system provided in an embodiment of this application.

[0174] like Figure 5 As shown, the damage assessment system includes a data acquisition module 501, a data preprocessing module 502, a parts price set 503, an accident loss model 504, a damage assessment amount calculation module 505, an output module 506, and an iterative feedback module 507.

[0175] The data acquisition module 501 is used to collect accident-related data and data from parts price databases. The data acquisition module 501 includes five data acquisition sub-modules.

[0176] Specifically, such as Figure 6 As shown, the data acquisition module includes five data acquisition sub-modules: accident damage data acquisition sub-module 601, vehicle network data acquisition sub-module 602, liability data acquisition sub-module 603, regional environment data acquisition sub-module 604, and parts price data acquisition sub-module 605. The above five data acquisition sub-modules obtain accident-related data and parts price data from different data sources.

[0177] The accident damage data acquisition submodule 601 is used to collect vehicle damage information through vehicle-to-everything (V2X) collision sensors, driving recorders, etc. The V2X data acquisition submodule 602 is used to obtain data such as the vehicle status and driving data of the accident vehicle. The liability data acquisition submodule 603 is used to obtain accident liability information (such as the liability percentage of the accident vehicle) from the insurance company system. The regional environmental data acquisition submodule 604 is used to obtain regional repair costs, transportation costs, etc. The parts price data acquisition submodule 605 is used to obtain parts price data from the supply chain, e-commerce platforms, etc.

[0178] The data preprocessing module 502 is used to clean and standardize the collected raw data (accident-related data and data from parts price sets), including three steps: missing value filling, outlier removal, and data standardization, to ensure that the data quality meets the input requirements of the accident loss model.

[0179] Parts Price Set 503, also known as the dynamic parts price database, is used to store and update parts price information in real time. It adopts a mechanism that combines real-time updates and full verification to ensure the timeliness and accuracy of the prices of parts in Parts Price Set 503.

[0180] The accident loss model 504 is the core intelligent computing module. It adopts machine learning algorithms and has the ability to self-learn the weights of loss assessment related features. It can automatically optimize the accident loss model parameters based on loss assessment feedback data (such as the deviation between the predicted loss assessment amount and the actual loss assessment amount).

[0181] The damage assessment amount calculation module 505 is used to calculate the predicted damage assessment amount based on the data output by the accident loss model 504 and the parts price set 503.

[0182] The output module 506 is used to output the predicted loss assessment amount for subsequent processes such as insurance claims and repair quotations.

[0183] The iterative feedback module 507 is used to collect actual damage assessment feedback data, which drives the optimization and updating of the parts price set 503 and the accident loss model 504, forming a closed loop of continuous improvement.

[0184] Figure 7 This is a schematic diagram of the architecture of an accessory price set provided in an embodiment of this application.

[0185] In some embodiments, the accessory price set may also be referred to as a dynamic accessory price library. This application does not limit the name of the accessory price set.

[0186] like Figure 7 As shown, the core of the accessory price set is constructed from the basic attribute layer 702, the price dynamic layer 704, and the update mechanism layer 703.

[0187] The basic attribute layer 702 is used to store static attributes such as part model, compatible vehicle model, manufacturer, part type, and warranty period.

[0188] The part model number is the unique identifier for the part. The compatible vehicle model refers to the type of vehicle the part is compatible with. Specifically, this can be linked to the vehicle's unique identification information via the VIN code. Part types can include original equipment manufacturer (OEM), aftermarket, and used parts. The manufacturer can be the information of the part's manufacturer. The warranty period is the quality guarantee period.

[0189] Price Dynamics Layer 704 is used to update the ex-factory price, channel price, and terminal retail price of components in real time, and to simultaneously calculate the regional price difference coefficient, supply and demand fluctuation coefficient, and metal futures correlation coefficient. Price Dynamics Layer 704 includes real-time price groups and dynamic coefficient groups.

[0190] The regional price difference coefficient can be set according to the regional economic level and logistics convenience. For example, the regional price difference coefficient for first-tier cities can be set to 1.0, while that for remote areas can be set to 1.2. The regional price difference coefficient reflects regional price differences.

[0191] The supply and demand volatility coefficient can be calculated based on inventory turnover days. For example, when the inventory turnover days are less than 7 days, the supply and demand volatility coefficient is set to 1.1, and when the inventory turnover days are greater than 30 days, the supply and demand volatility coefficient is set to 0.9. The supply and demand volatility coefficient can reflect the impact of supply and demand on the price of parts.

[0192] Inventory turnover days can be understood as the average number of days it takes for a company to go from acquiring inventory of parts to selling them.

[0193] The correlation coefficient of metal futures is for metal parts and components. It can be calculated by multiplying the increase in futures prices by 0.3. The correlation coefficient of metal futures can be used to correlate with the fluctuation of raw material costs.

[0194] The update mechanism layer 703 is used to update the data in the price dynamic layer 704. Specifically, it can use an incremental update mechanism to synchronize the latest changes in component prices in real time, and a full-volume verification mechanism to verify the accuracy of the data in the price dynamic layer (such as the aforementioned component ex-factory price, channel price, terminal retail price, regional price difference coefficient, supply and demand fluctuation coefficient, and metal futures correlation coefficient). At the same time, it can detect and issue early warnings for abnormal fluctuations in component prices. Through daily full-volume verification and abnormal early warning mechanisms, it ensures price accuracy while enabling rapid adaptation to emergencies such as raw material futures fluctuations and regional component shortages.

[0195] The multi-source data source 701 may include at least one of the following: supply chain system data, platform transaction data, actual settlement data of repair shops, raw material market data, or inventory information.

[0196] Computer equipment can obtain parts prices and dynamic coefficients (such as regional price difference coefficients, supply and demand fluctuation coefficients, and metal futures correlation coefficients) from the parts price central query through the price query interface 705. This is used to provide real-time and accurate parts prices for accident loss models. At the same time, the price query interface 705 allows users to independently select different parts types such as original, aftermarket, or used parts for classification and query. Based on the part type selected by the user, it automatically switches to the corresponding price system, thereby realizing a unique mapping between parts prices and vehicles and damaged parts, while meeting the parts selection needs of different users (such as quality requirements and cost requirements).

[0197] Specifically, the multi-source data source 701 provides multi-dimensional real-time prices of parts to the price dynamic layer 704, ensuring that parts prices can reflect the latest market conditions. The basic attribute layer 702 provides matching basis for parts price data, ensuring that parts prices correspond accurately with parts models and vehicle types. The update mechanism layer 703 is used to update the data in the price dynamic layer 704 to ensure data timeliness.

[0198] Compared to traditional static price lists, the timeliness of parts prices is improved by hundreds of times. It can accurately cover scenarios such as parts premiums in remote areas, short-term supply chain shortages, and raw material price fluctuations, ensuring that the assessed price always matches the actual market cost.

[0199] Figure 8 This is a principle block diagram of an accident loss model provided in an embodiment of this application.

[0200] like Figure 8 As shown, the core process modules of the accident loss model include: input layer 801, weight self-learning unit 802, model calculation layer 803, and output layer 804.

[0201] Input layer 801 is used to input multi-dimensional damage assessment related features (such as the price features, damage features, vehicle features, market features, human factors features, and auxiliary features mentioned above).

[0202] The weight self-learning unit 802 is used to dynamically optimize the weights of loss assessment-related features. The weight self-learning unit 802 may include optimization algorithms (such as Bayesian optimization, grid search, genetic algorithms, etc.) and a weight storage pool. The core weight range in the weight storage pool is 0.3-0.6, and the weights of price features and market features are adjusted primarily during weight self-learning.

[0203] The model calculation layer 803 is used to calculate the cost of a single part, the total cost of the parts, and the preliminary loss assessment amount based on the loss assessment correlation features and the weights corresponding to the loss assessment correlation features.

[0204] Among them, the cost per part is the cost of repairing / replacing a single damaged part, the total cost of parts is the sum of the costs of all damaged parts, and the preliminary damage assessment amount is the preliminary sum of the total cost of parts and labor costs.

[0205] Output layer 804 is used to output the calculation results of model calculation layer 803 (cost of a single component, total cost of components, and preliminary loss assessment amount).

[0206] Specifically, the computer equipment can input damage assessment correlation features into the accident loss model through the input layer 801. After obtaining multi-dimensional damage assessment correlation features, the weight self-learning unit 802 is used to assign weights to each damage assessment correlation feature. Then, the weighted damage assessment correlation features are input into the model calculation layer to calculate the cost of a single part, the total cost of parts, and the preliminary damage assessment amount, which intuitively reflects the severity of damage and repair costs of the accident vehicle. Finally, the preliminary damage assessment amount is output through the output layer 804.

[0207] After the output layer 804 outputs the preliminary loss assessment amount, the difference between the preliminary loss assessment amount and the actual loss assessment amount is calculated. This difference is used as loss assessment feedback data and input into the weight self-learning unit 802 to provide an objective function for weight optimization, drive the continuous improvement of the accident loss model, and form a continuous improvement cycle of prediction-feedback-optimization.

[0208] For example, Figure 9 This is a schematic diagram of a loss assessment calculation logic provided in an embodiment of this application.

[0209] like Figure 9As shown, the following five units can be combined to calculate the assessed loss amount: single component cost calculation unit 901, total component cost calculation unit 902, labor cost calculation unit 903, preliminary assessment loss amount unit 904, and rule adjustment unit 905.

[0210] The single-component cost calculation unit 901 is used to calculate the cost of a single damaged component, i.e., the single-component cost. Specifically, the single-component cost can be calculated using the following formula: Cost per component = .

[0211] in, Real-time prices for parts This is the supply and demand fluctuation coefficient. This is the transportation cost coefficient. For depreciation rate, Weights are assigned to the degree of damage.

[0212] The total parts cost calculation unit 902 is used to calculate the total cost of all damaged parts of the accident vehicle, i.e., the total parts cost. Specifically, the total parts cost can be obtained by summing the costs of all individual parts.

[0213] The labor cost calculation unit 903 is used to calculate the labor costs required to repair the accident vehicle, i.e., the total labor cost. Specifically, the total labor cost can be calculated by multiplying the regional labor cost benchmark by the repair hour coefficient.

[0214] The preliminary damage assessment unit 904 is used to calculate the preliminary damage assessment amount for the accident vehicle. Specifically, the preliminary damage assessment amount is calculated by summing the total parts cost and the total labor cost.

[0215] The rule adjustment unit 905 is used to adjust the preliminary damage assessment amount based on the liability ratio of the accident vehicle and the insurance company's deductible rules to obtain the predicted damage assessment amount. Specifically, based on the liability ratio of the accident vehicle, the preliminary damage assessment amount is multiplied by the corresponding percentage to obtain the damage assessment amount adjusted according to the liability ratio of the accident vehicle. Then, the corresponding amount is deducted according to the insurance company's deductible rules to obtain the predicted damage assessment amount.

[0216] For example, if the vehicle involved in the accident is fully responsible, the preliminary damage assessment amount is multiplied by 100%; if the vehicle is primarily responsible, the preliminary damage assessment amount is multiplied by 70%; if the vehicle is secondarily responsible, the preliminary damage assessment amount is multiplied by 30%; and if the vehicle is not responsible, the preliminary damage assessment amount is multiplied by 0%. The rule adjustment unit 905 converts the preliminary damage assessment amount into a predicted damage assessment amount that conforms to insurance business rules, ensuring the feasibility of the damage assessment result.

[0217] Furthermore, the technical solution provided in this application can also be implemented through an integrated solution combining a dynamic price engine, a deep learning model, a rule engine, and a continuous iteration mechanism. This solution can also be executed by a computer device.

[0218] The dynamic pricing engine integrates multi-source data (the same data collected from the parts price set). The dynamic pricing engine employs a three-layer architecture: data layer, computation layer, and application layer.

[0219] The data layer stores the collected raw parts price data; the computing layer dynamically generates coefficients such as regional price differences, supply and demand fluctuations, and futures correlations through real-time stream computing (such as Flink); and the application layer provides a standardized price query interface (supporting categorized calls for original equipment manufacturer (OEM), aftermarket, and used parts).

[0220] Simultaneously, an event-triggered incremental update and a timed full-scale verification mechanism can be used to update the data in the data layer. That is, when the supply chain price or futures price changes exceed a threshold (e.g., 5%), an immediate incremental update is triggered. At the same time, a full-scale verification of the data in the data layer is performed at regular intervals (e.g., at midnight every day). This provides the accident loss model with real-time and accurate dynamic component prices, solving the problem of static pricing.

[0221] After obtaining parts prices from the dynamic pricing engine, the computer equipment can perform data preprocessing on the accident-related data and parts price set by using the random forest algorithm to fill missing values, using the IQR rule to remove outliers, and using the Min-Max algorithm to standardize the data.

[0222] After data preprocessing, computer equipment can calculate the predicted damage amount of the accident vehicle by combining deep learning models with rule engines.

[0223] Deep learning models can employ a combination of Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks. CNNs are used to process image data such as damage assessment photos (identifying the location and extent of damage), while LSTMs are used to process vehicle-to-everything (V2X) time-series data (such as changes in collision intensity and battery pack voltage fluctuations).

[0224] Specifically, a preliminary damage assessment amount is calculated using a deep learning model. After obtaining the preliminary damage assessment amount, it is adjusted according to the compensation rules to obtain the predicted damage assessment amount. Then, the actual damage assessment amount of the accident vehicle is obtained, and the difference between the actual damage assessment amount and the predicted damage assessment amount is calculated. Finally, based on the difference between the actual damage assessment amount and the predicted damage assessment amount, the parameters in the deep learning model are automatically updated using the backpropagation algorithm of the deep learning model, and the deep learning model is iteratively updated.

[0225] Among them, compensation rules refer to the deterministic rules built into the rules engine, such as accident liability ratio, deductible rules, and regional labor cost standards.

[0226] In addition, computer equipment can use event-triggered and daily full update mechanisms to update the data in the parts pricing engine; and it can update the parameters of the deep learning model at fixed intervals (such as every 10 days); furthermore, it can reconstruct the deep learning model based on new vehicle models and parts data, thereby ensuring that the solution can adapt to market and industry development in the long term and avoid model aging.

[0227] In summary, the technical solution provided in this application obtains real-time, dynamic parts prices by acquiring parts prices related to market demand from a parts price database. Then, it extracts features from accident-related data to obtain damage assessment correlation features. Based on these features and real-time parts prices, it predicts the damage assessment amount for the accident vehicle, thereby reducing assessment errors. This shortens the parts price update delay to within a few hours, reduces manual intervention, improves assessment efficiency and consistency, lowers additional compensation costs for insurance companies and reduces claims disputes for vehicle owners, and provides an efficient, accurate, and universal intelligent damage assessment solution for insurance companies, OEM after-sales platforms, and third-party damage assessment agencies.

[0228] The technical solution provided in this application can be directly applied to various scenarios such as auto insurance claims by insurance companies, after-sales repair quotations by OEMs, assessments by third-party damage assessment agencies, and used car damage value calculation. Furthermore, the damage assessment method in this application can shorten the damage assessment cycle from the traditional several days to several hours, significantly improving claims efficiency, reducing the labor costs of insurance companies for damage assessment and the waiting time for car owners, while also reducing the cost of secondary corrections due to human error.

[0229] The foregoing mainly describes the solution provided in this application. Accordingly, this application also provides a damage assessment device for implementing the above-described method embodiments.

[0230] like Figure 10The schematic diagram of the damage assessment device shown indicates that the device may include an acquisition module 1001, a processing module 1002, and a prediction module 1003. The acquisition module 1001 is used to perform... Figure 2 , Figure 3 and Figure 4 The operation of step S202 in the illustrated method and Figure 3 The operation of step S203; the processing module 1002 is used to execute Figure 2 , Figure 3 and Figure 4 The illustrated method includes step S204; the prediction module 1003 is used to perform this operation. Figure 2 , Figure 3 and Figure 4 The operation of step S206 in the illustrated method and Figure 4 The operations of steps S207 and S208.

[0231] In some embodiments, the damage assessment device includes hardware structures and / or software modules corresponding to the execution of each function in order to achieve the above-described functions. Those skilled in the art will readily recognize that, based on the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0232] This application embodiment can divide the damage assessment device into functional modules according to the above method embodiment. For example, each function can be divided into a separate functional module, or two or more functions can be integrated into one damage assessment module. The integrated module can be implemented in hardware or as a software functional module. It should be noted that the module division in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0233] like Figure 11 As shown, the computer device provided in this application embodiment may include a processor 1101, a bus 1102, a communication interface 1103, and a memory 1104. The processor 1101, the memory 1104, and the communication interface 1103 communicate with each other via the bus 1102. It should be understood that this application does not limit the number of processors and memories in the network device.

[0234] Bus 1102 can be a PCI bus, an Extended Industry Standard Architecture (EISA) bus, or a UB bus, etc. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 11 The bus 1102 may be represented by a single line, but this does not mean that there is only one bus or one type of bus. The bus 1102 may include a path for transmitting information between various components of the network device (e.g., memory 1104, processor 1101, communication interface 1103).

[0235] Processor 1101 may include any one or more processors such as CPU, graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).

[0236] The memory 1104 may include volatile memory, such as random access memory (RAM). The processor 1101 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0237] The communication interface 1103 uses transceiver modules such as, but not limited to, network interface cards and transceivers to enable communication between network devices and other devices or communication networks.

[0238] The memory 1104 stores executable program code, and the processor 1101 executes the executable program code to implement the functions of the aforementioned method embodiments. That is, the memory 1104 stores instructions for performing the aforementioned damage assessment method.

[0239] On the other hand, a damage assessment system is provided, which includes multiple modules that work together to implement the damage assessment method described above.

[0240] In another aspect, a computer-readable storage medium is provided, which stores at least one computer program, which is loaded and executed by a processor to implement the damage assessment method provided in the above-described method embodiments.

[0241] In another aspect, a computer program product is provided, which includes a computer program or instructions that, when executed by a processor, implement the damage assessment method provided in the above-described method embodiments.

[0242] Through the above description of the implementation methods, those skilled in the art will clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the module can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, modules, and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0243] Since the damage assessment module, computer-readable storage medium, and computer program product in the embodiments of the present invention can be applied to the above method, the technical effects obtained can also be referred to the above method embodiments. The embodiments of the present invention will not be repeated here.

[0244] The method steps in this embodiment can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can reside in an ASIC. Alternatively, the ASIC can reside in a network device. Of course, the processor and storage medium can also exist as discrete components in the network device.

[0245] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer programs or instructions. When a computer program or instruction is loaded and executed on a computer, the processes or functions of the embodiments of this application are performed, in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, or other programmable module. The computer program or instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, a computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; it can also be an optical medium, such as a digital video disc (DVD); or it can be a semiconductor medium, such as a solid-state drive (SSD).

[0246] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for damage assessment, characterized in that, The method includes: Obtain accident-related data of the accident vehicle, and obtain the price set of the parts corresponding to the accident vehicle, wherein the price of the parts in the price set is related to the market demand of the parts; Feature extraction is performed on the accident-related data to obtain the damage assessment-related features of the accident vehicle; Based on the damage assessment correlation features and the prices of the damaged parts in the accident vehicle in the parts price set, a damage assessment prediction is made to obtain the predicted damage assessment amount for the accident vehicle.

2. The method according to claim 1, characterized in that, The method of predicting the damage assessment amount for the accident vehicle based on the damage assessment correlation features and the listed prices of the damaged parts in the parts price set includes: The damage assessment correlation features and the price of the damaged parts in the parts price set are input into the accident loss model to perform the damage assessment prediction, and the predicted damage assessment amount is obtained.

3. The method according to claim 2, characterized in that, The calculation formula for the loss assessment prediction using the accident loss model is expressed as follows: in, The predicted loss amount is n, where n is the total number of damaged parts. Let be the price listed in the price set of the ith damaged component. Let be the supply and demand fluctuation coefficient for the i-th damaged component. Let be the transportation cost coefficient for the i-th damaged part. Let be the depreciation rate for the i-th damaged component. Let the damage level weight be the i-th damaged component. As a benchmark for regional labor costs, This is the maintenance man-hour coefficient.

4. The method according to claim 3, characterized in that, The accident loss model is constructed using the XGBoost algorithm, the LightGBM algorithm, the CatBoost algorithm, and the ensemble decision tree random forest algorithm.

5. The method according to claim 2, characterized in that, The method further includes: Obtain the actual damage assessment amount for the accident vehicle; The accident loss model and / or the set of parts prices are updated based on the difference between the actual loss assessment amount and the predicted loss assessment amount.

6. The method according to any one of claims 1 to 5, characterized in that, The step of obtaining the price set of parts corresponding to the accident vehicle includes: Obtain price-related information for spare parts, including at least one of the following: supply chain system data, platform transaction data, actual settlement data of repair shops, raw material market data, or inventory information; Based on the price association information, the accessory price set is constructed.

7. The method according to claim 6, characterized in that, The method further includes: If the data in the accessory price set changes in its respective data source, the accessory price set shall be updated.

8. The method according to any one of claims 1 to 5, characterized in that, The method further includes: The accident-related data is preprocessed, including at least one of outlier removal, missing value imputation, or standardization.

9. The method according to any one of claims 1 to 5, characterized in that, The accident-related data includes: vehicle accident damage data, vehicle network status data, accident liability data, and regional environmental data.

10. The method according to any one of claims 1 to 5, characterized in that, The damage assessment related features include: damage features, vehicle features, market features, and human factors features.

11. A damage assessment device, characterized in that, The device includes: The acquisition module is used to acquire accident-related data of the accident vehicle and to acquire the price set of the parts corresponding to the accident vehicle. The price of the parts in the price set is related to the market demand of the parts. The processing module is used to extract features from the accident-related data to obtain the damage assessment-related features of the accident vehicle. The prediction module is used to predict the damage assessment amount of the accident vehicle based on the damage assessment correlation features and the prices of the damaged parts in the parts price set in the parts price set.

12. A computer device, characterized in that, The computer device includes a processor and a memory, wherein the memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor to implement the loss assessment method as described in any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one computer program, which is loaded and executed by a processor to implement the loss assessment method as described in any one of claims 1 to 10.

Images