WO2022267456A1 - 车辆定损方法、装置、设备及存储介质 - Google Patents

车辆定损方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2022267456A1
WO2022267456A1 PCT/CN2022/071483 CN2022071483W WO2022267456A1 WO 2022267456 A1 WO2022267456 A1 WO 2022267456A1 CN 2022071483 W CN2022071483 W CN 2022071483W WO 2022267456 A1 WO2022267456 A1 WO 2022267456A1
Authority
WO
WIPO (PCT)
Prior art keywords
risk
compensation
target
factor data
vehicle
Prior art date
Application number
PCT/CN2022/071483
Other languages
English (en)
French (fr)
Inventor
张霖
朱磊
徐赛奕
王遥
朱艳乔
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2022267456A1 publication Critical patent/WO2022267456A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/08—Insurance

Definitions

  • the present application relates to the technical field of neural networks, and in particular to a vehicle damage assessment method, device, equipment and storage medium.
  • unsaturated data refers to this type of data, which is only owned by some users, for example: driving behavior data, only a part of vehicles and user authorization models use driving behavior data.
  • driving behavior data only a part of vehicles
  • user authorization models use driving behavior data.
  • this part of unsaturated data has a strong predictability to the risk of accident.
  • the inventor realized that if the unsaturated data is added to the traditional model according to the traditional method, since most users lack this type of data, the overall model It does not have a strong effect, resulting in low accuracy in predicting the amount of compensation for vehicle damage.
  • the present application provides a vehicle damage determination method, device, equipment and storage medium, which are used to determine the amount of vehicle damage compensation through a preset actuarial model and a trained residual network model, and improve the accuracy of vehicle damage compensation prediction.
  • the first aspect of the present application provides a vehicle damage assessment method, including: receiving a vehicle damage assessment request, and obtaining a target factor data set according to the vehicle damage assessment request, and the target factor data set includes driver factor data and vehicle factor data , risk factor data and driving behavior factor data; carry out risk prediction on the target factor data set through the preset actuarial model, and obtain the initial compensation risk prediction result; use the trained residual network model to carry out the described driving behavior factor data
  • the risk identification process obtains the initial compensation risk residual value; calculates the target compensation risk actual result according to the initial compensation risk prediction result and the initial compensation risk residual value, and determines the vehicle damage compensation amount based on the target compensation risk actual result.
  • the second aspect of the present application provides a vehicle damage assessment device, including a memory, a processor, and computer-readable instructions stored in the memory and operable on the processor, and the processor executes the computer-readable instructions.
  • the following steps are implemented when the instruction is read: receive a vehicle damage assessment request, and obtain a target factor data set according to the vehicle damage assessment request, and the target factor data set includes driver factor data, vehicle factor data, risk factor data, and driving behavior factor data Carry out risk prediction on the target factor data set through the preset actuarial model to obtain the initial compensation risk prediction result; carry out risk identification processing on the driving behavior factor data through the trained residual network model to obtain the initial compensation risk residual difference; calculating the actual target compensation risk result according to the initial compensation risk prediction result and the initial compensation risk residual value, and determining the vehicle damage compensation amount based on the target compensation risk actual result.
  • the third aspect of the present application provides a computer-readable storage medium, wherein computer instructions are stored in the computer-readable storage medium, and when the computer instructions are run on the computer, the computer is made to perform the following steps: receiving a vehicle damage assessment request According to the vehicle damage assessment request, the target factor data set is obtained, and the target factor data set includes driver factor data, vehicle factor data, risk factor data and driving behavior factor data; Perform risk prediction on the data set to obtain the initial compensation risk prediction result; carry out risk identification processing on the driving behavior factor data through the trained residual network model to obtain the initial compensation risk residual value; according to the initial compensation risk prediction result and The initial compensation risk residual value calculates the target compensation risk actual result, and the vehicle damage compensation amount is determined based on the target compensation risk actual result.
  • the fourth aspect of the present application provides a vehicle damage assessment device, wherein the vehicle damage assessment device includes: an acquisition module, configured to receive a vehicle damage assessment request, and obtain a target factor data set according to the vehicle damage assessment request, the The target factor data set includes driver factor data, vehicle factor data, risk factor data and driving behavior factor data; the prediction module is used to perform risk prediction on the target factor data set through a preset actuarial model to obtain an initial compensation risk prediction Result; the identification module is used to carry out risk identification processing on the driving behavior factor data through the trained residual network model to obtain the initial compensation risk residual value; the calculation module is used to predict the initial compensation risk according to the result and the obtained Calculate the target compensation risk actual result based on the initial compensation risk residual value, and determine the vehicle damage compensation amount based on the target compensation risk actual result.
  • the vehicle damage assessment device includes: an acquisition module, configured to receive a vehicle damage assessment request, and obtain a target factor data set according to the vehicle damage assessment request, the The target factor data set includes driver factor data, vehicle factor data, risk factor
  • the vehicle damage assessment request is received, and the target factor data set is obtained according to the vehicle damage assessment request, and the target factor data set includes driver factor data, vehicle factor data, risk factor data and driving behavior factors Data; carry out risk prediction on the target factor data set through the preset actuarial model, and obtain the initial compensation risk prediction result; carry out risk identification processing on the driving behavior factor data through the trained residual network model, and obtain the initial compensation risk Residual value; calculate the actual result of the target compensation risk according to the initial compensation risk prediction result and the initial compensation risk residual value, and determine the vehicle damage compensation amount based on the actual target compensation risk result.
  • the target factor data set is obtained according to the vehicle damage assessment request, and the target factor data set includes the driver factor data; the risk prediction is performed on the target factor data set through the preset actuarial model, and the initial compensation risk prediction result is obtained; through The trained residual network model performs risk identification processing on the driving behavior factor data to obtain the initial compensation risk residual value; the vehicle damage compensation amount is determined according to the initial compensation risk prediction result and the initial compensation risk residual value.
  • the amount of vehicle damage compensation is determined through the preset actuarial model and the trained residual network model, which improves the accuracy of vehicle damage compensation prediction and the accuracy of risk prediction.
  • Fig. 1 is a schematic diagram of an embodiment of the vehicle damage determination method in the embodiment of the present application
  • Fig. 2 is a schematic diagram of another embodiment of the vehicle damage determination method in the embodiment of the present application.
  • Fig. 3 is a schematic diagram of an embodiment of the vehicle damage determination device in the embodiment of the present application.
  • Fig. 4 is a schematic diagram of another embodiment of the vehicle damage determination device in the embodiment of the present application.
  • Fig. 5 is a schematic diagram of an embodiment of a vehicle damage assessment device in an embodiment of the present application.
  • the embodiment of the present application provides a vehicle damage assessment method, device, equipment, and storage medium, which are used to determine the amount of vehicle damage compensation through a preset actuarial model and a trained residual network model, and improve the accuracy of risk prediction.
  • an embodiment of the vehicle damage determination method in the embodiment of the present application includes:
  • the 101 Receive a vehicle damage assessment request, and obtain a target factor data set according to the vehicle damage assessment request.
  • the target factor data set includes driver factor data, vehicle factor data, risk factor data, and driving behavior factor data.
  • the target factor data set is used to indicate the factor data related to the user's driving risk.
  • the target factor data set includes driver factor data, vehicle factor data, risk factor data and driving behavior factor data, wherein the driver factor data includes age, Gender, driving age and other factor data; vehicle factor data include vehicle model and vehicle condition, among which, vehicle condition refers to the technical condition of the vehicle, which is used to indicate the safety performance, power performance, operability, exhaust emission, vehicle appearance, etc. of the vehicle.
  • the risk factor data includes the number of historical accidents and the type of historical accidents. For example, the number of historical accidents is 1, and the type of historical accidents is that the claim information has not passed the audit; the driving behavior factor data can include data on rapid acceleration, rapid deceleration, and sharp turns.
  • the driving behavior factor data may also include driving duration data, driving mileage and driving speed, and the driving behavior factor data may also include other data, which is not specifically limited here.
  • the server receives the vehicle damage assessment request, analyzes the vehicle damage assessment request, and obtains the identity of the risk user and the license plate number of the risk user; the server queries the preset database according to the identity of the risk user and the license plate number of the risk user, and obtains the target factor data set , the target factor data set includes driver factor data, vehicle factor data, risk factor data and driving behavior factor data.
  • the preset database is a graph database (for example, neo4j), a data warehouse hive, an in-memory database (for example, SQLite or a remote service dictionary redis) or a relational database (for example, mysql), which is not limited here.
  • the server stores the target factor data set in the blockchain database, which is not limited here.
  • the subject of execution of the present application may be a vehicle damage assessment device, and may also be a terminal or a server, which is not specifically limited here.
  • the embodiment of the present application is described by taking the server as an execution subject as an example.
  • the preset actuarial model is used to indicate that the initial compensation risk prediction result is determined according to the target factor data set.
  • the value range of the initial compensation risk prediction result is greater than or equal to 0, and the initial compensation risk prediction result is based on the actual application scenario. It can be a benchmark payout ratio (that is, a probability value that conforms to a normal standard distribution table), or it can be benchmark payout amount data (greater than 0), which is not specifically limited here.
  • the server acquires initial vehicle compensation sample data, and the server performs claim data labeling processing on the initial vehicle compensation sample data to obtain target vehicle compensation sample data; the server builds a preset actuarial model based on the target vehicle compensation sample data.
  • the preset actuarial model can be a pre-trained linear regression model, a pre-trained polynomial regression model, or other pre-trained predictive user compensation risk model, which is not limited here.
  • the server performs feature screening on the target factor data set through a preset actuarial model to obtain multiple actuarial index feature data, and the server determines the weight value corresponding to each actuarial index feature data; Calculate the mean value of the weight value to obtain the initial compensation risk prediction result.
  • the initial compensation risk prediction result may be 0.9.
  • the initial compensation risk prediction result may be 800.
  • the initial compensation risk residual value is used to indicate the relative error value between the initial compensation risk prediction result and the target compensation risk actual result.
  • the initial compensation risk residual value can be the modified compensation ratio, or the modified compensation amount data. Wherein, the value range of the corrected payout ratio is greater than 0, and the corrected payout amount data can be a positive number, a negative number, or 0.
  • the initial compensation risk residual value is used to correct the initial compensation risk prediction result, so the initial compensation risk residual value is consistent with the initial compensation risk prediction result. For example, when the initial compensation risk prediction result is the benchmark compensation ratio, the initial compensation risk residual value is the modified compensation ratio, and the initial compensation risk residual value can be 1.125.
  • the initial compensation risk residual value is the corrected compensation amount data, and the initial compensation risk residual value can be +20, -20, or 0, which is not limited here.
  • the server inputs the driving behavior factor data into the trained residual network model; the server converts the driving behavior factor data into multiple initial feature vectors through the trained residual network model; the server numerically calculates the multiple initial feature vectors Normalization processing to obtain multiple normalized eigenvectors; the server performs identification and classification processing based on the multiple normalized eigenvectors to obtain the initial compensation risk residual value.
  • the value range of the initial compensation risk residual value is greater than or equal to 0, if the initial compensation risk prediction result can be in the value range between 0 and 1, that is, the initial compensation risk prediction result is a risk prediction probability value, then the server Multiply the initial compensation risk prediction result and the initial compensation risk residual value to obtain the target compensation risk actual result, the value range of the target compensation risk actual result is between 0 and 1; the server determines the vehicle damage according to the target compensation risk actual result Amount of compensation. For example, further, where the initial compensation risk prediction result is the basic compensation amount, and the initial compensation risk residual value is a positive or negative compensation difference, the server adds the initial compensation risk prediction result and the initial compensation risk residual value, Get the amount of compensation for vehicle damage.
  • the target factor data set is obtained according to the vehicle damage assessment request, and the target factor data set includes the driver factor data; the risk prediction is performed on the target factor data set through the preset actuarial model, and the initial compensation risk prediction result is obtained; through The trained residual network model performs risk identification processing on the driving behavior factor data to obtain the initial compensation risk residual value; the vehicle damage compensation amount is determined according to the initial compensation risk prediction result and the initial compensation risk residual value.
  • the amount of vehicle damage compensation is determined through the preset actuarial model and the trained residual network model, which improves the accuracy of vehicle damage compensation prediction and the accuracy of risk prediction.
  • FIG. 2 another embodiment of the vehicle damage determination method in the embodiment of the present application includes:
  • the 201 Receive a vehicle damage assessment request, and acquire a target factor data set according to the vehicle damage assessment request.
  • the target factor data set includes driver factor data, vehicle factor data, risk factor data, and driving behavior factor data.
  • step 201 The execution process of step 201 is similar to the execution process of step 101, and details will not be repeated here.
  • the server obtains a historical driving risk data set, and performs feature cleaning on the historical driving risk data set to obtain a driving behavior sample data set, each driving behavior sample data has a corresponding sample identifier; that is, the server Perform feature selection on the historical driving risk data set (that is, all underlying factors for modeling or preliminary processing factors), mainly based on saturation and correlation, and delete low-saturation features; server removal and Y label Highly correlated features (signatures that may have data breaches).
  • the server sequentially performs numerical feature normalization, data binning, and discrete feature numerical processing on the driving behavior sample data set to obtain the target sample data set; among them, numerical feature normalization improves the stability of the model; data binning can avoid The model is overfitting; the server realizes the numericalization of discrete features through one-hot encoding or target encoding.
  • the server predicts and identifies the target sample data set through the preset actuarial model, and obtains multiple sample prediction results; the server obtains the real result corresponding to each sample prediction result, and divides each sample prediction result by the corresponding The real results of the real results, get multiple ratios, and set multiple ratios as output variables; the server sets the target sample data set as an input variable, and generates a decision tree model according to the input variables and output variables through the gradient boosting decision tree algorithm, and according to the minimum The loss function of the simplified decision tree is used to perform pruning training on the decision tree model to obtain a trained residual network model.
  • the decision tree model includes multiple decision trees. After pruning training, applying the trained residual model can reduce energy consumption and improve risk prediction speed.
  • step 202 The execution process of step 202 is similar to the execution process of step 102, and details will not be repeated here.
  • the initial compensation risk residual value is used to indicate the relative error value between the initial compensation risk prediction result and the target compensation risk actual result.
  • the server performs feature extraction and feature vector normalization processing on the driving behavior factor data through the trained residual network model to obtain multiple normalized feature vectors.
  • the trained residual network model includes multiple decision trees ; The server identifies and classifies multiple normalized feature vectors based on multiple decision trees, obtains the prediction results corresponding to each decision tree, and accumulates the prediction results corresponding to each decision tree to obtain the initial compensation risk residual value.
  • each normalized feature vector belongs to [0,1] interval.
  • the initial compensation risk prediction result of the actuarial model is optimized. It can improve the prediction accuracy of the actuarial model without affecting the prediction results of the actuarial model.
  • the residual network model For users who have driving behavior factor data, use the residual network model to adjust the risk, and use the model to predict the results more accurately.
  • the versatility of the actuarial model is enhanced.
  • the residual network model can be used to improve the accuracy of risk prediction.
  • the server judges whether the initial compensation risk prediction result belongs to a preset numerical range; wherein, the preset numerical range is greater than or equal to 0 and less than or equal to 1. Further, the server judges whether the initial compensation risk prediction result is greater than or equal to 0 and less than or equal to 1. For example, if the initial compensation risk prediction result is 0.75, the server determines that the initial compensation risk prediction result belongs to the preset value range, that is, , the initial compensation risk prediction result is the base compensation ratio; if the initial compensation risk prediction result is 4000, the server determines that the initial compensation risk prediction result does not belong to the preset numerical range.
  • the server will multiply the initial compensation risk prediction result and the initial compensation risk residual value to obtain the target compensation risk actual result; for example, the initial compensation risk prediction result is 0.780, and the initial The residual value of the compensation risk is 0.975, and the initial compensation risk prediction result is multiplied by the residual value of the initial compensation risk, which is 0.78*0.975, and the actual result of the target compensation risk is 0.800.
  • the server queries the target business risk compensation rate from the preset risk compensation configuration table based on the actual result of the target compensation risk, and determines the vehicle damage compensation amount according to the target business risk compensation rate and the preset compensation benchmark amount.
  • the server performs an addition calculation on the initial compensation risk prediction result and the initial compensation risk residual value to obtain the target compensation risk actual result; for example, the initial compensation risk prediction result is 1250, the initial compensation risk residual value is -105, the initial compensation risk prediction result is added to the initial compensation risk residual value, that is, 1250+(-105), and the actual result of the target compensation risk is 1145.
  • the server queries the target business risk level from the preset risk level configuration table based on the actual result of the target compensation risk. When the target business risk level is not empty, the server sets the actual result of the target compensation risk as the vehicle damage compensation amount.
  • the server obtains the review order generation request; the server parses the review order generation request to obtain the business type, target reviewer ID, and user compensation information, and the user compensation information includes the amount of vehicle damage compensation; the server obtains order template information according to the business type; the server based on The template information is filled with the order according to the user's payment information to obtain the order content information; the server obtains the reviewer information according to the target reviewer ID, and converts the reviewer information and order content information into an order to be reviewed, and the pending order has a unique order ID ; The server writes the order to be reviewed to the preset message queue, and sends the order to be reviewed to the target terminal through the preset message queue according to the information of the reviewer.
  • the server receives the audit notification message sent by the target terminal, and extracts the order ID, audit result, and audit opinion information from the audit notification message.
  • the server updates the order status corresponding to the order to be audited according to the order ID and audit result.
  • the order status includes pending Review status, review status, review pass status and review fail status; and send the review results and review opinion information to the target users through preset notification methods, including SMS and message notification methods, specifically here
  • the server obtains the business type according to the order ID, and invokes the corresponding business process (for example, the vehicle compensation business process) according to the business type.
  • the target factor data set is obtained according to the vehicle damage assessment request, and the target factor data set includes the driver factor data; the risk prediction is performed on the target factor data set through the preset actuarial model, and the initial compensation risk prediction result is obtained; through The trained residual network model performs risk identification processing on the driving behavior factor data to obtain the initial compensation risk residual value; the vehicle damage compensation amount is determined according to the initial compensation risk prediction result and the initial compensation risk residual value.
  • the amount of vehicle damage compensation is determined through the preset actuarial model and the trained residual network model, which improves the accuracy of vehicle damage compensation prediction and the accuracy of risk prediction.
  • An embodiment of the vehicle loss assessment device in the embodiment of the application includes:
  • the acquisition module 301 is configured to receive a vehicle damage assessment request, and acquire a target factor data set according to the vehicle damage assessment request, where the target factor data set includes driver factor data, vehicle factor data, risk factor data, and driving behavior factor data;
  • a prediction module 302 configured to perform risk prediction on the target factor data set through a preset actuarial model to obtain an initial claim payment risk prediction result
  • the identification module 303 is used to perform risk identification processing on the driving behavior factor data through the trained residual network model to obtain the initial compensation risk residual value;
  • the calculation module 304 is configured to calculate the actual result of the target compensation risk according to the initial compensation risk prediction result and the initial compensation risk residual value, and determine the vehicle damage compensation amount based on the actual target compensation risk result.
  • target factor data set is stored in the blockchain database, which is not limited here.
  • the target factor data set is obtained according to the vehicle damage assessment request, and the target factor data set includes the driver factor data; the risk prediction is performed on the target factor data set through the preset actuarial model, and the initial compensation risk prediction result is obtained; through The trained residual network model performs risk identification processing on the driving behavior factor data to obtain the initial compensation risk residual value; the vehicle damage compensation amount is determined according to the initial compensation risk prediction result and the initial compensation risk residual value.
  • the amount of vehicle damage compensation is determined through the preset actuarial model and the trained residual network model, which improves the accuracy of vehicle damage compensation prediction and the accuracy of risk prediction.
  • FIG. 4 another embodiment of the vehicle damage determination device in the embodiment of the present application includes:
  • the acquisition module 301 is configured to receive a vehicle damage assessment request, and acquire a target factor data set according to the vehicle damage assessment request, where the target factor data set includes driver factor data, vehicle factor data, risk factor data, and driving behavior factor data;
  • a prediction module 302 configured to perform risk prediction on the target factor data set through a preset actuarial model to obtain an initial claim payment risk prediction result
  • the identification module 303 is used to perform risk identification processing on the driving behavior factor data through the trained residual network model to obtain the initial compensation risk residual value;
  • the calculation module 304 is configured to calculate the actual result of the target compensation risk according to the initial compensation risk prediction result and the initial compensation risk residual value, and determine the vehicle damage compensation amount based on the actual target compensation risk result.
  • the identification module 303 can also be specifically used for:
  • the trained residual network model includes multiple decision trees;
  • Identify and classify multiple normalized feature vectors based on multiple decision trees obtain the corresponding prediction results of each decision tree, and accumulate the corresponding prediction results of each decision tree to obtain the initial compensation risk residual value.
  • the computing module 304 may also include:
  • Judging unit 3041 configured to judge whether the initial compensation risk prediction result belongs to a preset numerical range
  • the first calculation unit 3042 is configured to, if the initial compensation risk prediction result belongs to the preset numerical range, perform multiplication operation on the initial compensation risk prediction result and the initial compensation risk residual value to obtain the target compensation risk actual result;
  • the determining unit 3043 is configured to query the target business risk compensation rate from the preset risk compensation configuration table based on the actual result of the target compensation risk, and determine the vehicle damage compensation amount according to the target business risk compensation rate and the preset compensation benchmark amount.
  • the computing module 304 may also include:
  • the second calculation unit 3044 is used to add the initial compensation risk prediction result and the initial compensation risk residual value to obtain the actual result of the target compensation risk if the initial compensation risk prediction result does not belong to the preset numerical range;
  • the setting unit 3045 is used to query the target business risk level from the preset risk level configuration table based on the actual result of the target compensation risk, and when the target business risk level is not empty, set the actual result of the target compensation risk as the vehicle damage compensation amount .
  • the vehicle damage assessment device may also include:
  • the cleaning module 305 is used to obtain a historical driving risk data set, and perform feature cleaning on the historical driving risk data set to obtain a driving behavior sample data set, and each driving behavior sample data has a corresponding sample identification;
  • the processing module 306 is used to sequentially perform numerical feature normalization, data binning and discrete feature numerical processing on the driving behavior sample data set to obtain the target sample data set;
  • the predictive identification module 307 is used to predict and identify the target sample data set through a preset actuarial model to obtain multiple sample forecast results;
  • the division module 308 is used to obtain the real result corresponding to the predicted result of each sample, respectively divide the predicted result of each sample by the real result corresponding to the predicted result of each sample, obtain multiple predicted ratios, and set the multiple predicted ratios as output variable;
  • the pruning training module 309 is used to set the target sample data set as an input variable, generate a decision tree model according to the input variable and output variable through the gradient boosting decision tree algorithm, and perform the decision tree model according to the loss function of the minimized decision tree Pruning training, the trained residual network model is obtained, and the decision tree model includes multiple decision trees.
  • the vehicle damage assessment device may also include:
  • the labeling module 310 is used to obtain initial vehicle compensation sample data, and perform claim data labeling processing on the initial vehicle compensation sample data to obtain target vehicle compensation sample data;
  • the construction module 311 is configured to construct a preset actuarial model based on the target vehicle compensation sample data, and the preset actuarial model is a trained linear regression model or a trained polynomial regression model.
  • the vehicle damage assessment device may also include:
  • a generation module 312, configured to generate an order to be reviewed according to the vehicle damage compensation amount, and send the order to be reviewed to the target terminal through a preset message queue;
  • the trigger module 313 is configured to receive an audit notification message sent by the target terminal, and trigger a corresponding business process according to the audit notification message.
  • the target factor data set is obtained according to the vehicle damage assessment request, and the target factor data set includes the driver factor data; the risk prediction is performed on the target factor data set through the preset actuarial model, and the initial compensation risk prediction result is obtained; through The trained residual network model performs risk identification processing on the driving behavior factor data to obtain the initial compensation risk residual value; the vehicle damage compensation amount is determined according to the initial compensation risk prediction result and the initial compensation risk residual value.
  • the amount of vehicle damage compensation is determined through the preset actuarial model and the trained residual network model, which improves the accuracy of vehicle damage compensation prediction and the accuracy of risk prediction.
  • Fig. 5 is a schematic structural diagram of a vehicle damage assessment device provided by an embodiment of the present application.
  • the vehicle damage assessment device 500 may have relatively large differences due to different configurations or performances, and may include one or more than one processor (central processing units) , CPU) 510 (eg, one or more processors) and memory 520, one or more storage media 530 (eg, one or more mass storage devices) for storing application programs 533 or data 532.
  • the memory 520 and the storage medium 530 may be temporary storage or persistent storage.
  • the program stored in the storage medium 530 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations for the vehicle damage assessment device 500 .
  • the processor 510 may be configured to communicate with the storage medium 530 , and execute a series of instruction operations in the storage medium 530 on the vehicle damage assessment device 500 .
  • the vehicle loss assessment device 500 can also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input and output interfaces 560, and/or, one or more operating systems 531, such as Windows Server , Mac OS X, Unix, Linux, FreeBSD, etc.
  • operating systems 531 such as Windows Server , Mac OS X, Unix, Linux, FreeBSD, etc.
  • the present application also provides a computer-readable storage medium.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium.
  • the computer-readable storage medium may also be a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium, and when the instructions are run on the computer, the computer is made to execute the steps of the vehicle damage assessment method.
  • the present application also provides a vehicle damage assessment device.
  • the vehicle damage assessment equipment includes a memory and a processor. Instructions are stored in the memory. The steps of the vehicle damage assessment method.
  • the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required by at least one function, etc.; The data created using the node, etc.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with each other using cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify its Validity of information (anti-counterfeiting) and generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
  • the technical solution of the present application is essentially or part of the contribution to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (read-only memory, ROM), random access memory (random access memory, RAM), magnetic disk or optical disc and other media that can store program codes. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Technology Law (AREA)
  • General Business, Economics & Management (AREA)
  • Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)

Abstract

本申请涉及人工智能技术领域，公开了一种车辆定损方法、装置、设备及存储介质，用于提高车辆定损的准确率。车辆定损方法包括：接收车辆定损请求，按照车辆定损请求获取目标因子数据集，目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；通过预设的精算模型对目标因子数据集进行风险预测，得到初始赔付风险预测结果；通过训练好的残差网络模型对驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；根据初始赔付风险预测结果和初始赔付风险残差值计算目标赔付风险实际结果，基于目标赔付风险实际结果确定车辆损伤赔付金额。此外，本申请还涉及区块链技术，目标因子数据集可存储于区块链节点中。

Description

车辆定损方法、装置、设备及存储介质
本申请要求于2021年06月24日提交中国专利局、申请号为202110701943.7、发明名称为“车辆定损方法、装置、设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在申请中。
技术领域
本申请涉及神经网络技术领域，尤其涉及一种车辆定损方法、装置、设备及存储介质。
背景技术
随着大数据的发展与普及，用户大数据画像中能采集的数据越来越多，也越来越全面。如何将用户数据运用到传统的车险精算模型中一直是行业关注的重点。
传统的车险精算模型，如果加入用户风险因子，则需要该用户风险因子具有明确的风险相关性，同步则要求数据饱和度高。面对大数据维度多，数据稀疏性强的特征，无法将不饱和因子(也就是稀疏因子)加入到传统的模型中。
其中，不饱和数据指的是该类型的数据，只在一部分用户拥有，例如：驾驶行为数据，只有一部分车辆和用户授权模型使用驾驶行为数据。而这部分不饱和数据对出险风险具有很强的预测性，发明人意识到，如果按照传统方法，将不饱和数据加入到传统模型中，由于大部分用户缺失该类型的数据，使得在整体模型并不能起到很强的效果，导致预测车辆损伤赔付金额的准确性低。
发明内容
本申请提供了一种车辆定损方法、装置、设备及存储介质，用于通过预设的精算模型和训练好的残差网络模型确定车辆损伤赔付金额，提高车辆算损伤赔偿预测的准确性。
本申请第一方面提供了一种车辆定损方法，包括：接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；通过预设的精算模型对所述目标因子数据集进行风险预测，得到初始赔付风险预测结果；通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额。
本申请第二方面提供了一种车辆定损设备，包括存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机可读指令，所述处理器执行所述计算机可读指令时实现如下步骤：接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；通过预设的精算模型对所述目标因子数据集进行风险预测，得到初始赔付风险预测结果；通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额。
本申请的第三方面提供了一种计算机可读存储介质，所述计算机可读存储介质中存储计算机指令，当所述计算机指令在计算机上运行时，使得计算机执行如下步骤：接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；通过预设的精算模型对所述目标因子数据集进行风险预测，得到初始赔付风险预测结果；通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额。
本申请第四方面提供了一种车辆定损装置，其中，所述车辆定损装置包括：获取模块，用于接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；预测模块，用于通过预设的精算模型对所述目标因子数据集进行风险预测，得到初始赔付风险预测结果；识别模块，用于通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；计算模块，用于根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额。
本申请提供的技术方案中，接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；通过预设的精算模型对所述目标因子数据集进行风险预测，得到初始赔付风险预测结果；通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额。本申请实施例中，按照车辆定损请求获取目标因子数据集，目标因子数据集包括驾驶人员因子数据；通过预设的精算模型对目标因子数据集进行风险预测，得到初始赔付风险预测结果；通过训练好的残差网络模型对驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；根据初始赔付风险预测结果和初始赔付风险残差值确定车辆损伤赔付金额。通过预设的精算模型和训练好的残差网络模型确定车辆损伤赔付金额，提高了车辆算损伤赔偿预测的准确性和风险预测的准确性。
附图说明
图1为本申请实施例中车辆定损方法的一个实施例示意图；
图2为本申请实施例中车辆定损方法的另一个实施例示意图；
图3为本申请实施例中车辆定损装置的一个实施例示意图；
图4为本申请实施例中车辆定损装置的另一个实施例示意图；
图5为本申请实施例中车辆定损设备的一个实施例示意图。
具体实施方式
本申请实施例提供了一种车辆定损方法、装置、设备及存储介质，用于通过预设的精算模型和训练好的残差网络模型确定车辆损伤赔付金额，提高风险预测的准确性。
本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”、“第三”、“第四”等(如果存在)是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的实施例能够以除了在这里图示或描述的内容以外的顺序实施。此外，术语“包括”或“具有”及其任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
为便于理解，下面对本申请实施例的具体流程进行描述，请参阅图1，本申请实施例中车辆定损方法的一个实施例包括：
101、接收车辆定损请求，按照车辆定损请求获取目标因子数据集，目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据。
其中，目标因子数据集用于指示与用户驾驶风险相关的因子数据，目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据，其中，驾驶人员因子数据包括年龄、性别、驾龄等因子数据；车辆因子数据包括汽车型号和车况等，其 中，车况是指车辆的技术状况，用于指示汽车的安全性能、动力性能、操作性能、尾气排放、车容车貌等多项指标；风险因子数据包括历史出险次数和历史出险类型，例如，历史出险次数为1，历史出险类型为理赔信息未通过审核；驾驶行为因子数据可以包括急加速数据、急减速数据、急转弯数据，驾驶行为因子数据还可以包括驾驶时长数据、驾驶里程和驾驶速度，驾驶行为因子数据也可以包括其他数据，具体此处不做限定。
具体的，服务器接收车辆定损请求，服务器解析车辆定损请求，得到风险用户身份标识和风险用户车牌号；服务器根据风险用户身份标识和风险用户车牌号查询预设的数据库，得到目标因子数据集，目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据。其中，预设的数据库为图数据库(例如，neo4j)、数据仓库hive、内存数据库(例如，SQLite或远程服务字典redis)或关系型数据库(例如，mysql)，具体此处不做限定。进一步地，服务器将目标因子数据集存储于区块链数据库中，具体此处不做限定。
可以理解的是，本申请的执行主体可以为车辆定损装置，还可以是终端或者服务器，具体此处不做限定。本申请实施例以服务器为执行主体为例进行说明。
102、通过预设的精算模型对目标因子数据集进行风险预测，得到初始赔付风险预测结果。
需要说明的是，预设的精算模型用于指示根据目标因子数据集确定初始赔付风险预测结果，初始赔付风险预测结果的取值范围为大于或者等于0，初始赔付风险预测结果按照实际应用场景，可以为基准赔付率(也就是，概率值，符合正态标准分布表)，也可以为基准赔付金额数据(大于0)，具体此处不做限定。进一步地，在步骤101之前，服务器获取初始车辆赔付样本数据，服务器对初始车辆赔付样本数据进行赔付数据标注处理，得到目标车辆赔付样本数据；服务器基于目标车辆赔付样本数据构建预设的精算模型。其中，预设的精算模型可以为预先训练好的线性回归模型，也可以为预先训练好的多项式回归模型，还可以为其他预先训练好的预测用户赔付风险模型，具体此处不做限定。
具体的，服务器通过预设的精算模型对目标因子数据集进行特征筛选，得到多个精算指标特征数据，服务器确定每个精算指标特征数据对应的权重值；服务器根据每个精算指标特征数据对应的权重值进行均值计算，得到初始赔付风险预测结果。例如，当初始赔付风险预测结果为基准赔付率时，初始赔付风险预测结果可以为0.9。当初始赔付风险预测结果为基准赔付金额数据时，初始赔付风险预测结果可以为800。
103、通过训练好的残差网络模型对驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值。
需要说明的是，初始赔付风险残差值用于指示初始赔付风险预测结果与目标赔付风险实际结果之间的相对误差值。初始赔付风险残差值可以为修正赔付率，也可以为修正赔付金额数据。其中，修正赔付率的取值范围大于0，修正赔付金额数据可以为正数，也可以为负数，还可以为0。初始赔付风险残差值用于修正初始赔付风险预测结果，因而初始赔付风险残差值与初始赔付风险预测结果具备一致性。例如，当初始赔付风险预测结果为基准赔付率时，初始赔付风险残差值为修正赔付率，初始赔付风险残差值可以为1.125，当初始赔付风险预测结果为基准赔付金额数据时，初始赔付风险残差值为修正赔付金额数据，初始赔付风险残差值可以为+20，也可以为-20，还可以为0，具体此处不做限定。具体的，服务器将驾驶行为因子数据输入至训练好的残差网络模型；服务器通过训练好的残差网络模型将驾驶行为因子数据转换为多个初始特征向量；服务器对多个初始特征向量进行数值归一化处理，得到多个归一化特征向量；服务器基于多个归一化特征向量进行识别分类处理，得到初始赔付风险残差值。
104、根据初始赔付风险预测结果和初始赔付风险残差值计算目标赔付风险实际结果，基于目标赔付风险实际结果确定车辆损伤赔付金额。
其中，初始赔付风险残差值的取值范围大于或等于0，若初始赔付风险预测结果可以为0至1之间的取值范围，也就是初始赔付风险预测结果为风险预测概率值，则服务器对初始赔付风险预测结果和初始赔付风险残差值进行乘法运算，得到目标赔付风险实际结果，目标赔付风险实际结果的取值范围为0至1之间；服务器根据目标赔付风险实际结果确定车辆损伤赔付金额。例如，进一步地，其中，初始赔付风险预测结果为基本赔偿金额，初始赔付风险残差值为正的或者负的赔偿差额，服务器将初始赔付风险预测结果与初始赔付风险残差值进行加法计算，得到车辆损伤赔付金额。
本申请实施例中，按照车辆定损请求获取目标因子数据集，目标因子数据集包括驾驶人员因子数据；通过预设的精算模型对目标因子数据集进行风险预测，得到初始赔付风险预测结果；通过训练好的残差网络模型对驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；根据初始赔付风险预测结果和初始赔付风险残差值确定车辆损伤赔付金额。通过预设的精算模型和训练好的残差网络模型确定车辆损伤赔付金额，提高了车辆算损伤赔偿预测的准确性和风险预测的准确性。
请参阅图2，本申请实施例中车辆定损方法的另一个实施例包括：
201、接收车辆定损请求，按照车辆定损请求获取目标因子数据集，目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据。
该步骤201的执行过程与步骤101步骤的执行过程相似，具体此处不再赘述。
进一步地，在步骤201之前，服务器获取历史驾驶风险数据集，并对历史驾驶风险数据集进行特征清洗，得到驾驶行为样本数据集，每个驾驶行为样本数据具有对应的样本标识；也就是，服务器对历史驾驶风险数据集(也就是，所有用于建模的底层因子或者初步加工的因子)进行特征选择，主要基于饱和度和相关性进行筛选，删除饱和度低的特征；服务器去除与Y标签相关性过高的特征(可能存在数据泄露的特征)。服务器对驾驶行为样本数据集依次进行数值特征归一化、数据分箱和离散特征数值化处理，得到目标样本数据集；其中，数值特征归一化提高了模型的稳定性；数据分箱能避免模型过拟合；服务器通过独热编码one-hot或目标编码target encoding实现离散特征数值化。服务器通过预设的精算模型对目标样本数据集进行预测识别，得到多个样本预测结果；服务器获取每个样本预测结果对应的真实结果，分别将每个样本预测结果除以每个样本预测结果对应的真实结果，得到多个比值，将多个比值设置为输出变量；服务器将目标样本数据集设置为输入变量，通过梯度提升决策树算法，按照输入变量和输出变量生成决策树模型，并根据最小化决策树的损失函数对决策树模型进行剪枝训练，得到训练好的残差网络模型，决策树模型包括多棵决策树。通过剪枝训练以后，应用训练好的残差模型能够降低能源消耗以及提高风险预测速度。
202、通过预设的精算模型对目标因子数据集进行风险预测，得到初始赔付风险预测结果。
该步骤202的执行过程与步骤102步骤的执行过程相似，具体此处不再赘述。
203、通过训练好的残差网络模型对驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值。
其中，初始赔付风险残差值用于指示初始赔付风险预测结果与目标赔付风险实际结果之间的相对误差值。可选的，服务器通过训练好的残差网络模型对驾驶行为因子数据进行特征提取和特征向量归一化处理，得到多个归一化特征向量，训练好的残差网络模型包括多棵决策树；服务器基于多棵决策树对多个归一化特征向量进行识别分类处理，得到每棵 决策树对应的预测结果，累加每棵决策树对应的预测结果，得到初始赔付风险残差值。
其中，每个归一化特征向量属于[0,1]区间。通过预设的精算模型和训练好的残差网络模型，优化精算模型的初始赔付风险预测结果。既可以提升精算模型预测的准确性，也不会影响精算模型的预测结果。针对有驾驶行为因子数据的用户，使用残差网络模型调整风险，使用模型预测结果更为精准。精算模型通用性增强，针对其他不饱和数据，例如用户位置服务数据、新能源车辆数据，均可以使用残差网络模型提升风险预测的准确性。
204、根据初始赔付风险预测结果和初始赔付风险残差值计算目标赔付风险实际结果，基于目标赔付风险实际结果确定车辆损伤赔付金额。
具体的，服务器判断初始赔付风险预测结果是否属于预设的数值范围；其中，预设的数值范围为大于或等于0，并且小于或等于1。进一步地，服务器判断初始赔付风险预测结果是否大于或等于0，并且小于或等于1，例如，若初始赔付风险预测结果为0.75，则服务器确定初始赔付风险预测结果属于预设的数值范围，也就是，初始赔付风险预测结果为基准赔付率；若初始赔付风险预测结果为4000，则服务器确定初始赔付风险预测结果不属于预设的数值范围。若初始赔付风险预测结果属于预设的数值范围，则服务器对初始赔付风险预测结果与初始赔付风险残差值进行乘法运算，得到目标赔付风险实际结果；例如，初始赔付风险预测结果为0.780，初始赔付风险残差值为0.975，初始赔付风险预测结果和初始赔付风险残差值相乘，也就是0.78*0.975，得到目标赔付风险实际结果为0.800。服务器基于目标赔付风险实际结果从预设的风险赔偿配置表中查询目标业务风险赔偿率，并根据目标业务风险赔偿率和预设的赔偿基准额度确定车辆损伤赔付金额。
进一步地，若初始赔付风险预测结果不属于预设的数值范围，则服务器对初始赔付风险预测结果与初始赔付风险残差值进行加法计算，得到目标赔付风险实际结果；例如，初始赔付风险预测结果为1250，初始赔付风险残差值为-105，初始赔付风险预测结果和初始赔付风险残差值相加，也就是1250+(-105)，得到目标赔付风险实际结果为1145。服务器基于目标赔付风险实际结果从预设的风险等级配置表中查询目标业务风险等级，当目标业务风险等级不为空值时，服务器将目标赔付风险实际结果设置为车辆损伤赔付金额。
205、按照车辆损伤赔付金额生成待审核订单，通过预设的消息队列将待审核订单发送至目标终端。
具体的，服务器获取审核订单生成请求；服务器解析审核订单生成请求，得到业务类型、目标审核人标识和用户赔付信息，用户赔付信息包括车辆损伤赔付金额；服务器根据业务类型获取订单模板信息；服务器基于模板信息按照用户赔付信息进行订单填充处理，得到订单内容信息；服务器根据目标审核人标识获取审核人信息，并将审核人信息和订单内容信息转换为待审核订单，待审核订单具有唯一的订单标识；服务器将待审核订单写入至预设的消息队列，并通过预设的消息队列按照审核人信息将待审核订单发送至目标终端。
206、接收目标终端发送的审核通知消息，按照审核通知消息触发对应的业务流程。
具体的，服务器接收目标终端发送的审核通知消息，并从审核通知消息中提取订单标识、审核结果和审核意见信息，服务器根据订单标识和审核结果更新待审核订单对应的订单状态，订单状态包括待审核状态、审核中状态、审核通过状态和审核未通过状态；并通过预设的通知方式将审核结果和审核意见信息送至目标用户，预设的通知方式包括短信方式和消息通知方式，具体此处不做限定；当审核结果为审核通过时，服务器按照订单标识获取业务类型，并按照业务类型调用对应的业务流程(例如，车辆赔偿业务流程)。
本申请实施例中，按照车辆定损请求获取目标因子数据集，目标因子数据集包括驾驶人员因子数据；通过预设的精算模型对目标因子数据集进行风险预测，得到初始赔付风险预测结果；通过训练好的残差网络模型对驾驶行为因子数据进行风险识别处理，得到初始 赔付风险残差值；根据初始赔付风险预测结果和初始赔付风险残差值确定车辆损伤赔付金额。通过预设的精算模型和训练好的残差网络模型确定车辆损伤赔付金额，提高了车辆算损伤赔偿预测的准确性和风险预测的准确性。
上面对本申请实施例中车辆定损方法进行了描述，下面对本申请实施例中车辆定损装置进行描述，请参阅图3，本申请实施例中车辆定损装置的一个实施例包括：
获取模块301，用于接收车辆定损请求，按照车辆定损请求获取目标因子数据集，目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；
预测模块302，用于通过预设的精算模型对目标因子数据集进行风险预测，得到初始赔付风险预测结果；
识别模块303，用于通过训练好的残差网络模型对驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；
计算模块304，用于根据初始赔付风险预测结果和初始赔付风险残差值计算目标赔付风险实际结果，基于目标赔付风险实际结果确定车辆损伤赔付金额。
进一步地，将目标因子数据集存储于区块链数据库中，具体此处不做限定。
本申请实施例中，按照车辆定损请求获取目标因子数据集，目标因子数据集包括驾驶人员因子数据；通过预设的精算模型对目标因子数据集进行风险预测，得到初始赔付风险预测结果；通过训练好的残差网络模型对驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；根据初始赔付风险预测结果和初始赔付风险残差值确定车辆损伤赔付金额。通过预设的精算模型和训练好的残差网络模型确定车辆损伤赔付金额，提高了车辆算损伤赔偿预测的准确性和风险预测的准确性。
请参阅图4，本申请实施例中车辆定损装置的另一个实施例包括：
获取模块301，用于接收车辆定损请求，按照车辆定损请求获取目标因子数据集，目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；
预测模块302，用于通过预设的精算模型对目标因子数据集进行风险预测，得到初始赔付风险预测结果；
识别模块303，用于通过训练好的残差网络模型对驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；
计算模块304，用于根据初始赔付风险预测结果和初始赔付风险残差值计算目标赔付风险实际结果，基于目标赔付风险实际结果确定车辆损伤赔付金额。
可选的，识别模块303还可以具体用于：
通过训练好的残差网络模型对驾驶行为因子数据进行特征提取和特征向量归一化处理，得到多个归一化特征向量，训练好的残差网络模型包括多棵决策树；
基于多棵决策树对多个归一化特征向量进行识别分类处理，得到每棵决策树对应的预测结果，累加每棵决策树对应的预测结果，得到初始赔付风险残差值。
可选的，计算模块304还可以包括：
判断单元3041，用于判断初始赔付风险预测结果是否属于预设的数值范围；
第一运算单元3042，用于若初始赔付风险预测结果属于预设的数值范围，则对初始赔付风险预测结果与初始赔付风险残差值进行乘法运算，得到目标赔付风险实际结果；
确定单元3043，用于基于目标赔付风险实际结果从预设的风险赔偿配置表中查询目标业务风险赔偿率，并根据目标业务风险赔偿率和预设的赔偿基准额度确定车辆损伤赔付金额。
可选的，计算模块304还可以包括：
第二运算单元3044，用于若初始赔付风险预测结果不属于预设的数值范围，则对初始 赔付风险预测结果与初始赔付风险残差值进行加法计算，得到目标赔付风险实际结果；
设置单元3045，用于基于目标赔付风险实际结果从预设的风险等级配置表中查询目标业务风险等级，当目标业务风险等级不为空值时，将目标赔付风险实际结果设置为车辆损伤赔付金额。
可选的，车辆定损装置还可以包括：
清洗模块305，用于获取历史驾驶风险数据集，并对历史驾驶风险数据集进行特征清洗，得到驾驶行为样本数据集，每个驾驶行为样本数据具有对应的样本标识；
处理模块306，用于对驾驶行为样本数据集依次进行数值特征归一化、数据分箱和离散特征数值化处理，得到目标样本数据集；
预测识别模块307，用于通过预设的精算模型对目标样本数据集进行预测识别，得到多个样本预测结果；
除法模块308，用于获取每个样本预测结果对应的真实结果，分别将每个样本预测结果除以每个样本预测结果对应的真实结果，得到多个预测比值，将多个预测比值设置为输出变量；
剪枝训练模块309，用于将目标样本数据集设置为输入变量，通过梯度提升决策树算法，按照输入变量和输出变量生成决策树模型，并根据最小化决策树的损失函数对决策树模型进行剪枝训练，得到训练好的残差网络模型，决策树模型包括多棵决策树。
可选的，车辆定损装置还可以包括：
标注模块310，用于获取初始车辆赔付样本数据，对初始车辆赔付样本数据进行赔付数据标注处理，得到目标车辆赔付样本数据；
构建模块311，用于基于目标车辆赔付样本数据构建预设的精算模型，预设的精算模型为训练好的线性回归模型或训练好的多项式回归模型。
可选的，车辆定损装置还可以包括：
生成模块312，用于按照车辆损伤赔付金额生成待审核订单，通过预设的消息队列将待审核订单发送至目标终端；
触发模块313，用于接收目标终端发送的审核通知消息，按照审核通知消息触发对应的业务流程。
本申请实施例中，按照车辆定损请求获取目标因子数据集，目标因子数据集包括驾驶人员因子数据；通过预设的精算模型对目标因子数据集进行风险预测，得到初始赔付风险预测结果；通过训练好的残差网络模型对驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；根据初始赔付风险预测结果和初始赔付风险残差值确定车辆损伤赔付金额。通过预设的精算模型和训练好的残差网络模型确定车辆损伤赔付金额，提高了车辆算损伤赔偿预测的准确性和风险预测的准确性。
上面图3和图4从模块化的角度对本申请实施例中的车辆定损装置进行详细描述，下面从硬件处理的角度对本申请实施例中车辆定损设备进行详细描述。
图5是本申请实施例提供的一种车辆定损设备的结构示意图，该车辆定损设备500可因配置或性能不同而产生比较大的差异，可以包括一个或一个以上处理器(central processing units，CPU)510(例如，一个或一个以上处理器)和存储器520，一个或一个以上存储应用程序533或数据532的存储介质530(例如一个或一个以上海量存储设备)。其中，存储器520和存储介质530可以是短暂存储或持久存储。存储在存储介质530的程序可以包括一个或一个以上模块(图示没标出)，每个模块可以包括对车辆定损设备500中的一系列指令操作。更进一步地，处理器510可以设置为与存储介质530通信，在车辆定损设备500上执行存储介质530中的一系列指令操作。
车辆定损设备500还可以包括一个或一个以上电源540，一个或一个以上有线或无线网络接口550，一个或一个以上输入输出接口560，和/或，一个或一个以上操作系统531，例如Windows Serve，Mac OS X，Unix，Linux，FreeBSD等等。本领域技术人员可以理解，图5示出的车辆定损设备结构并不构成对车辆定损设备的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。
本申请还提供一种计算机可读存储介质，该计算机可读存储介质可以为非易失性计算机可读存储介质，该计算机可读存储介质也可以为易失性计算机可读存储介质，所述计算机可读存储介质中存储有指令，当所述指令在计算机上运行时，使得计算机执行所述车辆定损方法的步骤。
本申请还提供一种车辆定损设备，所述车辆定损设备包括存储器和处理器，存储器中存储有指令，所述指令被处理器执行时，使得处理器执行上述各实施例中的所述车辆定损方法的步骤。
进一步地，所述计算机可读存储介质可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需的应用程序等；存储数据区可存储根据区块链节点的使用所创建的数据等。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统，装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain)，本质上是一个去中心化的数据库，是一串使用密码学方法相关联产生的数据块，每一个数据块中包含了一批次网络交易的信息，用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(read-only memory，ROM)、随机存取存储器(random access memory，RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，以上实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围。

Claims (20)

  1. 一种车辆定损方法，其中，所述车辆定损方法包括：
    接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；
    通过预设的精算模型对所述目标因子数据集进行风险预测，得到初始赔付风险预测结果；
    通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；
    根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额。
  2. 根据权利要求1所述的车辆定损方法，其中，所述通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值包括：
    通过训练好的残差网络模型对所述驾驶行为因子数据进行特征提取和特征向量归一化处理，得到多个归一化特征向量，所述训练好的残差网络模型包括多棵决策树；
    基于所述多棵决策树对所述多个归一化特征向量进行识别分类处理，得到每棵决策树对应的预测结果，累加每棵决策树对应的预测结果，得到初始赔付风险残差值。
  3. 根据权利要求1所述的车辆定损方法，其中，所述根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额包括：
    判断所述初始赔付风险预测结果是否属于预设的数值范围；
    若所述初始赔付风险预测结果属于预设的数值范围，则对所述初始赔付风险预测结果与所述初始赔付风险残差值进行乘法运算，得到目标赔付风险实际结果；
    基于所述目标赔付风险实际结果从预设的风险赔偿配置表中查询目标业务风险赔偿率，并根据所述目标业务风险赔偿率和预设的赔偿基准额度确定车辆损伤赔付金额。
  4. 根据权利要求3所述的车辆定损方法，其中，在所述基于所述目标赔付风险实际结果从预设的风险赔偿配置表中查询目标业务风险赔偿率，并根据所述目标业务风险赔偿率和预设的赔偿基准额度确定车辆损伤赔付金额之后，所述车辆定损方法还包括：
    若所述初始赔付风险预测结果不属于预设的数值范围，则对所述初始赔付风险预测结果与所述初始赔付风险残差值进行加法计算，得到目标赔付风险实际结果；
    基于所述目标赔付风险实际结果从预设的风险等级配置表中查询目标业务风险等级，当所述目标业务风险等级不为空值时，将所述目标赔付风险实际结果设置为车辆损伤赔付金额。
  5. 根据权利要求1-4中任意一项所述的车辆定损方法，其中，在所述接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据之前，所述车辆定损方法还包括：
    获取历史驾驶风险数据集，并对所述历史驾驶风险数据集进行特征清洗，得到驾驶行为样本数据集，每个驾驶行为样本数据具有对应的样本标识；
    对所述驾驶行为样本数据集依次进行数值特征归一化、数据分箱和离散特征数值化处理，得到目标样本数据集；
    通过预设的精算模型对所述目标样本数据集进行预测识别，得到多个样本预测结果；
    获取每个样本预测结果对应的真实结果，分别将每个样本预测结果除以每个样本预测结果对应的真实结果，得到多个预测比值，将所述多个预测比值设置为输出变量；
    将所述目标样本数据集设置为输入变量，通过梯度提升决策树算法，按照所述输入变 量和所述输出变量生成决策树模型，并根据最小化决策树的损失函数对所述决策树模型进行剪枝训练，得到所述训练好的残差网络模型，所述决策树模型包括多棵决策树。
  6. 根据权利要求1-4中任意一项所述的车辆定损方法，其中，在所述接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据之前，所述车辆定损方法还包括：
    获取初始车辆赔付样本数据，对初始车辆赔付样本数据进行赔付数据标注处理，得到目标车辆赔付样本数据；
    基于所述目标车辆赔付样本数据构建所述预设的精算模型，所述预设的精算模型为训练好的线性回归模型或训练好的多项式回归模型。
  7. 根据权利要求1-4中任一项所述的车辆定损方法，其中，在所述根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额之后，所述车辆定损方法还包括：
    按照所述车辆损伤赔付金额生成待审核订单，通过预设的消息队列将所述待审核订单发送至目标终端；
    接收所述目标终端发送的审核通知消息，按照所述审核通知消息触发对应的业务流程。
  8. 一种车辆定损设备，包括存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机可读指令，所述处理器执行所述计算机可读指令时实现如下步骤：
    接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；
    通过预设的精算模型对所述目标因子数据集进行风险预测，得到初始赔付风险预测结果；
    通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；
    根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额。
  9. 根据权利要求8所述的车辆定损设备，其中，所述通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值包括：
    通过训练好的残差网络模型对所述驾驶行为因子数据进行特征提取和特征向量归一化处理，得到多个归一化特征向量，所述训练好的残差网络模型包括多棵决策树；
    基于所述多棵决策树对所述多个归一化特征向量进行识别分类处理，得到每棵决策树对应的预测结果，累加每棵决策树对应的预测结果，得到初始赔付风险残差值。
  10. 根据权利要求8所述的车辆定损设备，其中，所述根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额包括：
    判断所述初始赔付风险预测结果是否属于预设的数值范围；
    若所述初始赔付风险预测结果属于预设的数值范围，则对所述初始赔付风险预测结果与所述初始赔付风险残差值进行乘法运算，得到目标赔付风险实际结果；
    基于所述目标赔付风险实际结果从预设的风险赔偿配置表中查询目标业务风险赔偿率，并根据所述目标业务风险赔偿率和预设的赔偿基准额度确定车辆损伤赔付金额。
  11. 根据权利要求10所述的车辆定损设备，其中，在所述基于所述目标赔付风险实际结果从预设的风险赔偿配置表中查询目标业务风险赔偿率，并根据所述目标业务风险赔偿率和预设的赔偿基准额度确定车辆损伤赔付金额之后，所述车辆定损方法还包括：
    若所述初始赔付风险预测结果不属于预设的数值范围，则对所述初始赔付风险预测结 果与所述初始赔付风险残差值进行加法计算，得到目标赔付风险实际结果；
    基于所述目标赔付风险实际结果从预设的风险等级配置表中查询目标业务风险等级，当所述目标业务风险等级不为空值时，将所述目标赔付风险实际结果设置为车辆损伤赔付金额。
  12. 根据权利要求8-11中任意一项所述的车辆定损设备，其中，在所述接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据之前，所述车辆定损方法还包括：
    获取历史驾驶风险数据集，并对所述历史驾驶风险数据集进行特征清洗，得到驾驶行为样本数据集，每个驾驶行为样本数据具有对应的样本标识；
    对所述驾驶行为样本数据集依次进行数值特征归一化、数据分箱和离散特征数值化处理，得到目标样本数据集；
    通过预设的精算模型对所述目标样本数据集进行预测识别，得到多个样本预测结果；
    获取每个样本预测结果对应的真实结果，分别将每个样本预测结果除以每个样本预测结果对应的真实结果，得到多个预测比值，将所述多个预测比值设置为输出变量；
    将所述目标样本数据集设置为输入变量，通过梯度提升决策树算法，按照所述输入变量和所述输出变量生成决策树模型，并根据最小化决策树的损失函数对所述决策树模型进行剪枝训练，得到所述训练好的残差网络模型，所述决策树模型包括多棵决策树。
  13. 根据权利要求8-11中任意一项的车辆定损设备，其中，在所述接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据之前，所述车辆定损方法还包括：
    获取初始车辆赔付样本数据，对初始车辆赔付样本数据进行赔付数据标注处理，得到目标车辆赔付样本数据；
    基于所述目标车辆赔付样本数据构建所述预设的精算模型，所述预设的精算模型为训练好的线性回归模型或训练好的多项式回归模型。
  14. 根据权利要求8-11中任意一项所述的车辆定损设备，其中，在所述根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额之后，所述车辆定损方法还包括：
    按照所述车辆损伤赔付金额生成待审核订单，通过预设的消息队列将所述待审核订单发送至目标终端；
    接收所述目标终端发送的审核通知消息，按照所述审核通知消息触发对应的业务流程。
  15. 一种计算机可读存储介质，所述计算机可读存储介质中存储计算机指令，当所述计算机指令在计算机上运行时，使得计算机执行如下步骤：
    接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；
    通过预设的精算模型对所述目标因子数据集进行风险预测，得到初始赔付风险预测结果；
    通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；
    根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额。
  16. 根据权利要求15所述的计算机可读存储介质，其中，所述通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值包括：
    通过训练好的残差网络模型对所述驾驶行为因子数据进行特征提取和特征向量归一化处理，得到多个归一化特征向量，所述训练好的残差网络模型包括多棵决策树；
    基于所述多棵决策树对所述多个归一化特征向量进行识别分类处理，得到每棵决策树对应的预测结果，累加每棵决策树对应的预测结果，得到初始赔付风险残差值。
  17. 根据权利要求15所述的计算机可读存储介质，其中，所述根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额包括：
    判断所述初始赔付风险预测结果是否属于预设的数值范围；
    若所述初始赔付风险预测结果属于预设的数值范围，则对所述初始赔付风险预测结果与所述初始赔付风险残差值进行乘法运算，得到目标赔付风险实际结果；
    基于所述目标赔付风险实际结果从预设的风险赔偿配置表中查询目标业务风险赔偿率，并根据所述目标业务风险赔偿率和预设的赔偿基准额度确定车辆损伤赔付金额。
  18. 根据权利要求17所述的计算机可读存储介质，其中，在所述基于所述目标赔付风险实际结果从预设的风险赔偿配置表中查询目标业务风险赔偿率，并根据所述目标业务风险赔偿率和预设的赔偿基准额度确定车辆损伤赔付金额之后，所述车辆定损方法还包括：
    若所述初始赔付风险预测结果不属于预设的数值范围，则对所述初始赔付风险预测结果与所述初始赔付风险残差值进行加法计算，得到目标赔付风险实际结果；
    基于所述目标赔付风险实际结果从预设的风险等级配置表中查询目标业务风险等级，当所述目标业务风险等级不为空值时，将所述目标赔付风险实际结果设置为车辆损伤赔付金额。
  19. 根据权利要求15-18中任意一项所述的计算机可读存储介质，其中，在所述接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据之前，所述车辆定损方法还包括：
    获取历史驾驶风险数据集，并对所述历史驾驶风险数据集进行特征清洗，得到驾驶行为样本数据集，每个驾驶行为样本数据具有对应的样本标识；
    对所述驾驶行为样本数据集依次进行数值特征归一化、数据分箱和离散特征数值化处理，得到目标样本数据集；
    通过预设的精算模型对所述目标样本数据集进行预测识别，得到多个样本预测结果；
    获取每个样本预测结果对应的真实结果，分别将每个样本预测结果除以每个样本预测结果对应的真实结果，得到多个预测比值，将所述多个预测比值设置为输出变量；
    将所述目标样本数据集设置为输入变量，通过梯度提升决策树算法，按照所述输入变量和所述输出变量生成决策树模型，并根据最小化决策树的损失函数对所述决策树模型进行剪枝训练，得到所述训练好的残差网络模型，所述决策树模型包括多棵决策树。
  20. 一种车辆定损装置，其中，所述车辆定损装置包括：
    获取模块，用于接收车辆定损请求，按照所述车辆定损请求获取目标因子数据集，所述目标因子数据集包括驾驶人员因子数据、车辆因子数据、风险因子数据和驾驶行为因子数据；
    预测模块，用于通过预设的精算模型对所述目标因子数据集进行风险预测，得到初始赔付风险预测结果；
    识别模块，用于通过训练好的残差网络模型对所述驾驶行为因子数据进行风险识别处理，得到初始赔付风险残差值；
    计算模块，用于根据所述初始赔付风险预测结果和所述初始赔付风险残差值计算目标 赔付风险实际结果，基于所述目标赔付风险实际结果确定车辆损伤赔付金额。
PCT/CN2022/071483 2021-06-24 2022-01-12 车辆定损方法、装置、设备及存储介质 WO2022267456A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110701943.7A CN113255833B (zh) 2021-06-24 2021-06-24 车辆定损方法、装置、设备及存储介质
CN202110701943.7 2021-06-24

Publications (1)

Publication Number Publication Date
WO2022267456A1 true WO2022267456A1 (zh) 2022-12-29

Family

ID=77189430

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/071483 WO2022267456A1 (zh) 2021-06-24 2022-01-12 车辆定损方法、装置、设备及存储介质

Country Status (2)

Country Link
CN (1) CN113255833B (zh)
WO (1) WO2022267456A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116452007A (zh) * 2023-06-15 2023-07-18 深圳市迪博企业风险管理技术有限公司 一种基于胶囊网络的企业税收合规风险评估方法
CN116452008A (zh) * 2023-06-16 2023-07-18 山东四季车网络科技有限公司 一种基于多项式建模的二手车风险预测方法及系统

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113255833B (zh) * 2021-06-24 2021-10-12 平安科技（深圳）有限公司 车辆定损方法、装置、设备及存储介质
CN114493903B (zh) * 2022-02-17 2024-04-09 平安科技（深圳）有限公司 人伤风险评估中估损模型优化方法及相关设备

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110103822A (zh) * 2019-05-17 2019-08-09 深圳市尼欧科技有限公司 一种驾驶行为习惯智能检测及安全驾驶评分的装置和方法
CN111127228A (zh) * 2019-12-30 2020-05-08 山东坤达诚经济咨询有限公司 风险因子识别方法、系统及终端
CN111260487A (zh) * 2020-01-20 2020-06-09 北京中科泽达科技有限公司 一种车险理赔的风险控制方法及装置
CN113255833A (zh) * 2021-06-24 2021-08-13 平安科技（深圳）有限公司 车辆定损方法、装置、设备及存储介质

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8489433B2 (en) * 2010-07-29 2013-07-16 Insurance Services Office, Inc. System and method for estimating loss propensity of an insured vehicle and providing driving information
CN108960431A (zh) * 2017-05-25 2018-12-07 北京嘀嘀无限科技发展有限公司 指标的预测、模型的训练方法及装置
CN110599353A (zh) * 2018-06-13 2019-12-20 百度在线网络技术（北京）有限公司 一种车辆出险和赔付率预测方法、装置、设备及介质
CN108986249B (zh) * 2018-06-26 2021-08-03 杭州车厘子智能科技有限公司 基于全景环视图像的车辆远程定损方法和系统

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110103822A (zh) * 2019-05-17 2019-08-09 深圳市尼欧科技有限公司 一种驾驶行为习惯智能检测及安全驾驶评分的装置和方法
CN111127228A (zh) * 2019-12-30 2020-05-08 山东坤达诚经济咨询有限公司 风险因子识别方法、系统及终端
CN111260487A (zh) * 2020-01-20 2020-06-09 北京中科泽达科技有限公司 一种车险理赔的风险控制方法及装置
CN113255833A (zh) * 2021-06-24 2021-08-13 平安科技（深圳）有限公司 车辆定损方法、装置、设备及存储介质

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116452007A (zh) * 2023-06-15 2023-07-18 深圳市迪博企业风险管理技术有限公司 一种基于胶囊网络的企业税收合规风险评估方法
CN116452007B (zh) * 2023-06-15 2023-09-19 深圳市迪博企业风险管理技术有限公司 一种基于胶囊网络的企业税收合规风险评估方法
CN116452008A (zh) * 2023-06-16 2023-07-18 山东四季车网络科技有限公司 一种基于多项式建模的二手车风险预测方法及系统
CN116452008B (zh) * 2023-06-16 2023-08-29 山东四季车网络科技有限公司 一种基于多项式建模的二手车风险预测方法及系统

Also Published As

Publication number Publication date
CN113255833B (zh) 2021-10-12
CN113255833A (zh) 2021-08-13

Similar Documents

Publication Publication Date Title
WO2022267456A1 (zh) 车辆定损方法、装置、设备及存储介质
US20160078361A1 (en) Optimized training of linear machine learning models
CN111652280B (zh) 基于行为的目标物数据分析方法、装置及存储介质
CN110175697B (zh) 一种不良事件风险预测系统及方法
CN110992167A (zh) 银行客户业务意图识别方法及装置
CN111464583B (zh) 计算资源分配方法、装置、服务器及存储介质
CN111242230A (zh) 基于人工智能的图像处理方法及图像分类模型训练方法
CN111507470A (zh) 一种异常账户的识别方法及装置
CN113190599B (zh) 应用用户行为数据的处理方法、装置、设备及存储介质
CN113282630B (zh) 基于接口切换的数据查询方法及装置
CN110995459A (zh) 异常对象识别方法、装置、介质及电子设备
CN113283675B (zh) 指标数据分析方法、装置、设备及存储介质
CN115147092A (zh) 资源审批方法、随机森林模型的训练方法及装置
CN108428001B (zh) 信用分值预测方法及装置
CN113568759B (zh) 一种基于云计算的大数据处理方法及其系统
CN107844490B (zh) 一种数据库的分库方法及装置
CN112784273A (zh) 一种sql风险识别方法、装置及设备
CN115730507A (zh) 模型引擎的构建、核函数处理方法、设备及存储介质
CN113434273B (zh) 数据处理方法、装置、系统及存储介质
CN117056591A (zh) 基于动态预测的电力缴费渠道智能推荐方法及系统
CN113269179B (zh) 数据处理方法、装置、设备及存储介质
CN111523685B (zh) 基于主动学习的降低性能建模开销的方法
Chen et al. Using deep learning to predict and optimize hadoop data analytic service in a cloud platform
CN114048512A (zh) 一种处理敏感数据的方法及装置
CN113205442A (zh) 基于区块链的电子政务数据反馈管理方法及装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22826967

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE